Institut de recherche et documentation en économie de la santé

Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature Laure Com-Ruelle, Zeynep Or, Thomas Renaud. Avec la collaboration de Chantal Ambroise et Anna Marek

Rapport n° 545 (biblio n° 1734) - Décembre 2008 CONSEIL D’ADMINISTRATION

Président François Joliclerc Trésorier Jean-Marie Langlois Membres Hubert Allemand, Laure Com-Ruelle, Yves Humez, Jean-Marie Langlois, Daniel Lenoir, Dominique Liger, Christine Meyer, Julien Mousquès, Bernard Salengro, Georges Simoni, Jean-Marie Thomas, Frédéric Van Roekeghem

CONSEIL SCIENTIFIQUE

Président François Heran Membres Didier Blanchet, Marc Brodin, Martine Bungener, Mathieu Cousineau, Gérard De Pouvourville, Guy Desplanques, Brigitte Dormont, Pierre-Yves Geoffard, Pierre-Jean Lancry, Annette Leclerc, Pierre Lombrail, Andrée Mizrahi, Arié Mizrahi, Lucile Olier, Dominique Polton, Philippe Ricordeau, Lise Rochaix, Simone Sandier, Alain Trognon, Philippe Ulmann

EQUIPE DE RECHERCHE

Directrice Chantal Cases Directrice adjointe Catherine Sermet Directeurs de recherche Yann Bourgueil, Laure Com-Ruelle, Philippe Le Fur

Maîtres de recherche Thierry Debrand, Paul Dourgnon, Nathalie Grandfils, Véronique Lucas-Gabrielli, Georges Menahem, Julien Mousques, Zeynep Or, Marc Perronnin. Chargés de recherche Anissa Afrite, Anne Aligon, Caroline Allonier, Magali Coldefy, Bidénam Kambia-Chopin, Pascale Lengagne, Aurélie Pierre, Thomas Renaud, Thierry Rochereau, Jérôme Silva, Nicolas Sirven Chercheur associé Caroline Despres, Michel Grignon, Florence Jusot, Michel Naiditch Attachées de recherche Claire Barangé, Martine Broïdo, Nicolas Célant, Violaine Eudier, Stéphanie Guillaume, Nelly Le Guen, Frédérique Ruchon Responsable de la documentation Marie-Odile Safon Documentalistes Suzanne Chriqui, Véronique Suhard Responsable des publications Nathalie Meunier Informaticiens Bruno Dervillez, Jacques Harrouin Secrétaire générale Dominique Goldfarb Cadre administratif Isabelle Henri Assistante de direction Catherine Banchereau Secrétaires-maquettistes Khadidja Ben Larbi, Franck-Séverin Clérembault, Aude Sirvain Secrétaire de la diffusion Sandrine Bequignon Assistant documentation/diffusion Damien Le Torrec INSTITUT DE RECHERCHE ET DOCUMENTATION EN ECONOMIE DE LA SANTE

Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature

Laure Com-Ruelle

Zeynep Or

Thomas Renaud

Avec la collaboration de Chantal Ambroise et Anna Marek

Rapport n° 545 (biblio n° 1734) Décembre 2008

Les noms d’auteurs apparaissent par ordre alphabétique. Toute reproduction de textes ou tableaux est autorisée sous réserve de l’indication de la source et de l’auteur. En cas de reproduction du texte intégral ou de plus de 10 pages, le directeur de l’IRDES devra être informé préalablement.

IRDES - Association Loi de 1901 - 10 rue Vauvenargues - 75018 - Tél. Secrétariat : 01.53.93.43.05 - Fax : 01.53.93.43.50 I.S.B.N. : 978-2-87812-349-4. Sommaire

- 5 - Sommaire

Synthèse...... 9

Introduction...... 13

Partie 1 Lien entre volume d’activité et qualité : une revue systématique de la littérature ...... 17

1.1. Introduction ...... 19

1.2. Résumé des revues systématiques antérieures ...... 19

1.3. Méthodologie de la revue IRDES ...... 20

1.3.1. Identification des études...... 20

1.3.2. Critères d’inclusion et d’exclusion ...... 20

1.3.3. Protocole/Grille d’analyse...... 21

1.4. Principaux résultats ...... 21

1.4.1. Caractéristiques des articles sélectionnés ...... 21

1.4.2. Soins et interventions étudiés ...... 23

1.4.3. Définition de la qualité ...... 27

1.4.4. Définition du volume d’activité ...... 29

1.4.5. Ajustement par les caractéristiques cliniques du patient (case-mix) ...... 29

Partie 2 Causalité entre volume d’activité et qualité : éléments théoriques ...... 31

2.1. Introduction ...... 33

2.2. Les économies d’échelle ...... 33

2.2.1. Les économies d’échelle externes ...... 34

2.2.2. Les économies d’échelle internes ...... 34

2.3. Théorie de la courbe d’apprentissage (learning by doing)...... 35

2.4. L’effet d’apprentissage dans le contexte hospitalier ...... 36

2.4.1. Comment le volume engendre-t-il la qualité des soins ? ...... 36

2.4.2. Effet d’apprentissage ou renvoi sélectif ? ...... 37

Conclusion ...... 39

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 6 - Bibliographie...... 43

Liste des tableaux et graphiques ...... 57

Annexes...... 61

Annexe 1 - Stratégie de recherche bibliographique ...... 63

Annexe 2 - Grille d’analyse ...... 67

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 7 -

Remerciements

Les auteurs tiennent à remercier :

– Michel Naiditch (IRDES) et Fréderic Bousquet (HAS) pour leurs nombreux commentaires constructifs ;

– Martine Broïdo pour sa relecture particulièrement vigilante ;

– Khadidja Ben Larbi pour la mise en page de ce rapport.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008

Synthèse

- 11 -

Synthèse

Contexte La grande majorité des recherches retenues porte sur des interventions chirurgicales (85 %), Au cours des deux dernières décennies, les principalement liées au cancer. Le reste des études portant sur la relation entre le volume études s’intéresse à un ensemble hétérogène de d’activité et la qualité des soins dans le secteur soins, regroupant à la fois des soins médicaux de la santé ont suscité une vive controverse, et l’ensemble des prestations délivrées par des notamment parce qu’elles ont souvent servi services et unités de soins spéci ques (service d’argument à la concentration de l’offre de soins fi d’obstétrique, unité de soins intensifs…). Plus hospitaliers, qui est présentée comme un moyen l’intervention est spéci que et complexe, plus la d’améliorer la qualité des soins. En fi corrélation volume-qualité est af rmée. Pour la également, la question du maintien d’un juste fi plupart des interventions, il n’existe pas de seuil équilibre entre la concentration de l’activité d’activité unanimement accepté. De plus, certaines hospitalière dans de grands centres et le maintien études montrent que la relation volume-qualité des soins de proximité se pose de plus en plus. devient marginale au-delà d’un seuil qui peut être Il est donc important de vérifier la pertinence des relativement bas. En n, l’in uence du volume sur seuils d’activité dans un contexte où la maîtrise fi fl la qualité peut diminuer, voire disparaître au cours des dépenses de santé et les contraintes de la du temps pour certaines procédures, notamment à démographie médicale pèsent sur les décisions mesure que la maîtrise de cette procédure s’étend d’implantation des plateaux techniques de et que celle-ci se banalise. différents niveaux.

L’effet d’apprentissage au niveau individuel Objectifs (chirurgien) mais aussi au niveau de l’hôpital Ce travail de recherche vise à fournir une analyse (transfert de connaissances, mode d’organisation) synthétique de l’ensemble des connaissances semble expliquer une grande partie de cette disponibles à ce jour sur les liens entre le volume corrélation. Cet effet d’apprentissage implique d’activité et la qualité des soins. Il propose en en effet que la qualité des soins dispensés par premier lieu une analyse systématique et critique les médecins et les hôpitaux augmente lorsqu’ils des études empiriques qui ont tenté d’établir un soignent plus de patients, grâce à une expérience lien entre le volume d’activité et la qualité des accrue. Cela suggère que la relation volume- soins dans les établissements de santé. Ainsi, qualité résulte d’économies d’échelle internes une revue systématique de la littérature des dix propres à l’établissement. dernières années a été réalisée afin d’identifier les interventions chirurgicales et les autres types Toutefois, dans certains cas, l’hypothèse alternative de soins ayant fait l’objet d’analyses sur le rapport du « renvoi sélectif », selon laquelle les patients volume/qualité. Au total, 175 articles ont été sont orientés vers les hôpitaux ayant de meilleurs évalués selon un protocole standard. résultats, ne peut être réfutée. Cela suppose que c’est la qualité qui engendre le volume d’activité En deuxième lieu, ce travail cherche à expliquer plutôt que l’inverse. pourquoi et comment le volume peut générer la qualité, en identifiant les mécanismes qui Ce lien de causalité entre volume et qualité sous-tendent ce lien. Les concepts théoriques doit par ailleurs être nuancé : les résultats sont relatifs à l’économie industrielle ont été mobilisés sensibles à la nature des procédures et interven- pour mieux comprendre, d’une part, le sens de tions analysées, ainsi qu’aux seuils d’activité la causalité entre volume et amélioration de la retenus et à des évolutions temporelles. qualité et, d’autre part, les dynamiques sous- jacentes expliquant cette causalité.

Principaux résultats La revue de la littérature montre que pour certaines procédures et interventions, en particulier pour la chirurgie complexe, la possibilité d’amé- liorer la qualité des soins lorsque le volume d’activité augmente est réelle.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008

Introduction

- 15 -

Introduction

Au cours des deux dernières décennies, les Il existe une quantité abondante de recherches, études portant sur la relation entre le volume et la provenant pour la plupart des États-Unis, qui qualité des soins dans le secteur de la santé ont suggèrent qu’un volume d’activité clinique plus suscité une vive controverse, notamment parce élevé est associé à de meilleurs résultats des 4 qu’elles ont souvent servi d’argument à la concen- soins . Ces études portent sur une grande variété tration de l’offre de soins hospitaliers, qui est d’interventions chirurgicales, de diagnostics ou présentée comme un moyen d’améliorer la qualité de programmes cliniques et analysent, dans des des soins. contextes de soins variés, différentes populations avec des méthodes et des indicateurs de qualité résultats divers. Dans les années quatre-vingt-dix, En France, un rapport récent1 réalisé à la demande quelques revues de littérature ont été publiées du ministre de la Santé et des Solidarités recom- sur ces questions, notamment en anglais, mais mande l’introduction de seuils d’activité chirurgicale jusqu’à présent aucune revue systématique n’était et préconise la fermeture « sans délai » de 113 disponible en français. services de chirurgie ayant un faible volume d’activité. Le rapport suggère que ces services ne sont pas capables de garantir une qualité et À la demande de la Haute Autorité de Santé (HAS) une sécurité suffisantes. De son côté, l’Institut en 2007, l’IRDES a réalisé un examen complet de national du Cancer (INCa), face à la nécessité la littérature portant sur la relation entre le volume de réfléchir à des critères d’autorisation en d’activité dans les établissements de santé et les cancérologie2, vient de définir des seuils minimaux résultats des soins. Une revue systématique de pour autoriser un certain nombre d’interventions. la littérature au cours des dix dernières années En ce qui concerne la chirurgie cardiaque, qui est a été effectuée afin d’identifier les interventions une des cinq activités de soins pour lesquelles chirurgicales et les autres types de soins ayant fait les Agences régionales de l’hospitalisation (ARH) l’objet d’analyses sur le rapport volume/qualité. doivent élaborer un schéma interrégional d’organi- Au total, 175 articles ont été évalués selon un sation sanitaire (SROS), l’arrêté du 24 janvier 2006 protocole standard. Par ailleurs, nous avons étudié fixe l’activité minimale annuelle par an et par site les concepts théoriques nécessaires pour mieux de chirurgie à 400 interventions majeures sur des comprendre le lien entre volume d’activité et patients adultes et à 150 actes pour la chirurgie résultats des soins, en nous référant à l’économie pédiatrique3. industrielle.

Ces seuils auront une influence importante sur la Cette revue de la littérature comporte deux parties restructuration de l’offre hospitalière en chirurgie. distinctes. Une première partie propose une Pourtant, la question du maintien d’un juste analyse systématique et critique des études équilibre entre la concentration de l’activité empiriques qui ont tenté d’établir un lien entre le hospitalière dans de grands centres et le maintien volume d’activité et la qualité des soins dans les des soins de proximité se pose de plus en plus. établissements de santé. L’objectif spécifique de Il est donc important de vérifier la pertinence des cette partie est d’identifier : seuils d’activité dans un contexte où la maîtrise des dépenses de santé et les contraintes de la – les champs d’activité, les conditions cliniques et démographie médicale pèsent sur les décisions les interventions (chirurgicales ou médicales) d’implantation des plateaux techniques de pour lesquels le lien entre volume d’activité et différents niveaux. qualité a été étudié et ceux pour lesquels cette relation est établie ;

– les indicateurs utilisés pour mesurer la qualité des soins et le volume d’activité ;

1 Rapport présenté par Guy Vallancien, L’évaluation de la sécurité, de la qualité et de la continuité des soins chirurgicaux dans des petits hôpitaux publics en France, avril 2006.

2 La mesure 36 du plan cancer prévoit des critères d’agrément pour la pratique de la cancérologie dans les établissements de soins publics et privés. Le décret du 06 mai 2005 inclut le traitement du cancer dans les activités soumises à autorisation. 4 Voir par exemple, Ho et al. (2006) ; Geubbels et al. (2005) ; 3 Circulaire DHOS/O4 n° 2006-293 du 3 juillet 2006 relative à Birkmeyer et al. (2002); Hillner et al. (2000) ; Thiemann et al. l’activité de soins de chirurgie cardiaque. (1999) ; Hannan et al. (1992).

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 16 -

– les méthodes statistiques employées dans – à identifier les hypothèses sous-jacentes d’un ces études, pour établir leur pertinence avec lien volume-qualité dans la littérature en se des critères communs (échantillon utilisé, rapportant à l’économie industrielle ; indicateur(s) de qualité employé(s), contrôle du case-mix, etc.). – à évaluer dans quelle mesure et comment ces théories/hypothèses sont adaptées (ou Une seconde partie est consacrée à l’analyse adaptables) pour l’analyse des liens volume- théorique des mécanismes sous-jacents de qualité dans le domaine de la santé. l’association entre le volume et la qualité. Cette partie vise notamment :

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 Partie 1 Lien entre volume d’activité et qualité : une revue systématique de la littérature

- 19 -

1. Lien entre volume d’activité et qualité : une revue systématique de la littérature

1.1. Introduction publié un examen systématique des études publiées jusqu’en 1996. Cette revue de York Cette première partie présente les méthodes et (Sowden et al., 1997), qui est devenue une les résultats principaux de la revue systématique référence majeure dans ce domaine, tire trois de l’IRDES sur les liens entre le volume d’activité conclusions principales sur la relation volume- et la qualité des soins au sein des établissements qualité : de santé. Les principales questions adressées par – Contrôle du case-mix: la plupart des recherches cette revue sont les suivantes : ne contrôlent pas suffisamment les caractéris- tiques cliniques des patients traités dans les – Quels sont les domaines d’activité (interventions) différents établissements ; le plus souvent étudiés ? – Causalité : aucune étude ne montre que – Quels sont les indicateurs utilisés pour mesurer l’augmentation de l’activité dans le temps la qualité et le volume ? conduit à l’amélioration des résultats. Il est donc difficile de conclure, à partir des résultats – Dans quelle mesure les résultats convergent-ils des études transversales (montrant que les pour une même intervention/procédure ? établissements à volume d’activité élevé ont de meilleurs résultats), que ce résultat est – Quelle est la pertinence des méthodes statistiques transposable à d’autres établissements et que utilisées ? l’expansion de l’activité y améliorera la qualité ;

Cette revue systématique couvre les études – Seuil d’activité : la plupart des études montrent publiées de 1997 à 2007. Le fait de se concentrer que si l’augmentation d’activité engendre un sur les études les plus récentes est important gain de qualité, le niveau du seuil demandé est dans un contexte où la qualité des données plutôt faible. Au-delà d’un certain seuil d’activité, et les méthodes statistiques utilisées évoluent les augmentations de volume peuvent être rapidement. Avant de présenter les résultats réalisées au sein d’un même établissement, de cette revue, nous faisons un point rapide sur sans changements structurels majeurs, les enseignements des revues systématiques mais en redéfinissant le partage du travail à antérieures afin de positionner nos résultats par l’intérieur des départements/services d’un rapport à ces travaux. Ainsi, nous résumons dans établissement donné. la section suivante les résultats de quatre grandes revues réalisées dans les dix dernières années. Ainsi, malgré plusieurs travaux considérés comme Nous présentons ensuite la stratégie de recherche pertinents, la revue de York conclut essentielle- et les méthodes employées dans notre revue ment que la plupart des études disponibles sont systématique (section 1.3), suivies par le résumé contestables du point de vue méthodologique synthétique des résultats (section 1.4). La totalité et qu’elles ont peu de valeur pour orienter les des articles étudiés est présentée dans l’Annexe 2, décisions concernant la restructuration des à l’aide d’une grille d’analyse commune. services de soins.

Les résultats des autres revues systématiques 1.2. Résumé des revues systématiques suivant la revue de York vont dans le même sens. antérieures Slim et Chipponi (2002), en étudiant la littérature entre 1985 et 2001 sur un certain nombre d’inter- Depuis les années quatre-vingt, un grand nombre ventions chirurgicales, notent que peu d’études d’études ont tenté d’établir une relation significative s’appuient sur un bon ajustement des caractéris- entre les volumes de cas traités, pour diverses tiques des patients pour comparer les centres de interventions chirurgicales, et les résultats pour soins ou les chirurgiens. Ils concluent que les biais les patients. Si la plupart des études semblent méthodologiques ne permettent pas une conclusion indiquer que les patients soignés dans des définitive, même si une relation positive est établissements à volume élevé ont un risque de suggérée, notamment pour les centres de soins. mortalité plus faible, la fiabilité des études est En ce qui concerne le volume d’activité du souvent mise en cause. chirurgien et ses résultats, ils estiment que la relation est moins bien établie. En 1997, le Centre national de revues et dissémi- nation à l’Université de York (NHS Center for En évaluant les études publiées entre 1980 Reviews and Dissemination) en Angleterre a et 2000, Ethan et al. (2002) concluent qu’un

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 20 - volume d’activité élevé est associé à de électronique et avons exclu les articles dont le meilleurs résultats pour un grand nombre sujet principal n’était pas l’évaluation des résultats d’interventions mais que l’ampleur de cette en fonction du volume de soins. Notre investigation relation varie largement. Les auteurs notent a été complétée par une recherche manuelle des toutefois qu’en raison des problèmes métho- bibliographies. dologiques liés à la plupart de ces études, il Après élimination des doublons, nous avons pour orienter les politiques régionales. recensé au total 226 références entrant dans le cadre de notre étude. Dans cette liste, 29 articles - ont été exclus de la procédure de relecture car dée et publiée par l’Institut Canadien d’informa- ils ne correspondaient pas exactement à notre problématique « volume-qualité » ou ne compor- résultats de l’étude de York: « […] on a peu sou- taient pas de résultats empiriques. En outre, 22 vent pris en compte les risques préexistants et articles n’ont pas été inclus dans notre revue les autres caractéristiques ayant des répercus- puisque nous n’avons pas réussi à les obtenir sions sur les résultats pour les patients comme dans les délais de notre étude5 l’âge, la gravité des maladies et les autres problè- avons examiné 175 articles en détail. mes de santé. » . Les auteurs notent également chaque étude en raison de différences existant 1.3.2. Critères d’inclusion et d’exclusion Seules les études publiées ces dix dernières volume, et les mesures des résultats utilisées. années, présentant des résultats empiriques et Ainsi, même si les établissements ayant un grand ayant un questionnement direct sur le rapport volume d’activité obtiennent souvent de meilleurs entre le volume d’activité d’un établissement (ou résultats, ils concluent que la relation entre le d’un praticien) et les résultats de soins obtenus, volume et les résultats n’est pas explicite. ont été retenues.

Nous avons inclus un large spectre d’études 1.3. Méthodologie de la revue IRDES portant sur des données cliniques, des données administratives, des cohortes prospectives/ rétrospectives et des analyses longitudinales. Les 1.3.1. Identification des études études comparant les résultats des patients traités dans des établissements avec différents niveaux Nous avons recherché de manière systématique de volume et les études de type « avant/après » les études publiées entre 1997 et 2007 portant sur sont incluses. Nous n’avons pas fait de restriction des procédures chirurgicales, des interventions concernant les variables de « qualité » ou de et/ou d’autres types de soins ou services de santé résultats cliniques. Toutes les variables utilisées dont les résultats ont été étudiés à partir d’une comme outcome (y compris la morbidité, la mesure du volume de soins. satisfaction, la qualité de vie, etc.) sont conservées dans le champ analysé. La recherche bibliographique a été effectuée à partir des mots-clés sur les bases de données Nous avons explicitement cherché les études électroniques suivantes (1997-2007): qui se penchent sur le volume d’activité des médecins (chirurgiens), ainsi que sur celui des => Medline, Pubmed, Embase (Elsevir), Pascal établissements de santé (Cf. annexe 1). (CNRS-INIST France), Bdsp (Banque de données santé publique), bases IRDES. Les articles ne présentant pas de résultats empiriques (discussion, revues, éditoriaux, etc.) La stratégie de recherche est décrite en détail en n’ont pas été retenus. Les études utilisant les Annexe 1. Une revue de la littérature préliminaire mêmes données et méthodes pour une intervention donnée (duplication de publication) ne sont Nous avons cherché à éviter les restrictions a présentées qu’une seule fois. priori concernant les interventions et les types de soins (ce qui était le cas pour la plupart des revues études portant même sur des soins de ville. 5 Dans la mesure où certains articles constituent des études régionales (une seule région) ou observationnelles provenant en grande partie des Etats-Unis, ils sont publiés dans des revues Nous avons parcouru les titres et les résumés de près de 1000 articles repérés par la recherche France.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 21 -

Les études qui n’interrogent pas directement le – variables à l’échelle des patients pour contrôler rapport volume-qualité mais s’intéressent à l’impact le case-mix (morbidité, variables démographi- des différents modes d’intervention ou traitement ques, socio-économiques, etc.) ; (par exemple anticancéreux), en contrôlant par le volume de l’établissement, ont été exclues. Pour – résultats des soins obtenus. restreindre la portée de la recherche électronique, nous avons cherché les mots-clés uniquement Tous les articles ainsi évalués sont présentés dans dans les titres et les abstracts. l’Annexe 2.

1.3.3. Protocole/Grille d’analyse 1.4. Principaux résultats

Les études correspondant à nos critères ont été Au total, nous avons examiné 175 articles évaluées avec un protocole standard. Pour chaque complets. Nous décrivons ci-dessous les étude, nous avons décrit les éléments suivants : principales caractéristiques de ces 175 articles, – pays ou région(s) étudiée(s) ; en distinguant les interventions étudiées, les indicateurs de qualité et de volume employés, les – période(s) étudiée(s) ; variables de contrôle de case-mix de et les résultats obtenus. – type de données mobilisées (administratives, cliniques, appariements, registres) ; 1.4.1. Caractéristiques des articles sélec- – taille de l’échantillon (nombre d’établissements, tionnés de chirurgiens, de patients…) ; Année de publication – intervention(s) étudiée(s) et codes médicaux (diagnostics, actes ou autres) utilisés pour Les articles étudiés ont été publiés entre 1997 et définir l’intervention (si possible) ; le printemps 2007. Le plus grand nombre d’articles a été publié en 2005 et 2006 (16 % et 15 % – indicateurs d’outcome employés pour mesurer respectivement) ; relativement peu d’articles ont « la qualité » ; été publiés dans les années 1997 et 1998. Ceci reflète notre stratégie de privilégier les études les – indicateurs de volume (par établissement et/ou plus récentes et peut expliquer la différence entre par chirurgien) ; nos conclusions et les enseignements des revues précédentes, notamment en ce qui concerne les – méthodes statistiques et économétriques ; données et méthodes utilisées (Cf. graphique 1). Graphique 1 Répartition des études par année de publication

Nombre d'articles

30

25

20

15

10

5

0 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 22 -

Zone géographique cliniques et a fortiori dans les données cliniques pures (comme les registres) qui sont générale- La majorité des études ont été menées aux ment beaucoup plus détaillées au point de vue États-Unis (126 études, soit 72 %) et ensuite au clinique. Canada (16 études). Par rapport aux revues systématiques de la littérature antérieure, notre 44 études reposent sur des données administra- étude contient une part relativement importante tives hospitalières, chaînées avec d’autres bases d’études non-américaines. Les études réalisées de données cliniques (ou, à défaut, fournissant des en Europe représentent 12 % du total et celles données de morbidité détaillées). On en compte réalisées dans le Pacifique occidental plus de 26 qui sont basées sur des registres spécifiques : 6 %. Une étude internationale compare la mortalité elles offrent évidemment un meilleur contrôle hospitalière pour les interventions coronaires des caractéristiques des patients avec de très percutanées à partir de 4 registres européens, nombreuses variables cliniques bien ciblées. 3 registres japonais, 4 registres américains et 2 6 publications sont fondées sur des enquêtes registres internationaux. ponctuelles ad hoc et 5 sur une base prospective avec un suivi de cohorte. Tableau 1 Répartition des études par pays Nous avons également repéré une analyse utilisant Nombre une étude de cohorte intégrée dans un essai clinique Pays de l’étude % d’articles randomisé. Etats-Unis 126 72,0 La plupart des études sont menées de façon 16 9,1 Canada rétrospective, mais certaines proposent un suivi International 1 0,6 prospectif, notamment les études de cohorte. Europe 21 12,0 Certaines s’appuient sur un suivi mixte, combinant Allemagne 4 2,3 ces deux aspects rétrospectif et prospectif. Angleterre 9 5,1 Espagne 1 0,6 Les méthodes statistiques utilisées dans ces études sont diverses mais généralement bien France 2 1,1 adaptées, comprenant des analyses uni-variées Italie 1 0,6 et bi-variées (avec des tests du Khi²), des tests Pays-Bas 3 1,7 statistiques variés (T-test, test de Wilcoxon ou Suède 1 0,6 méthode rank-sum) ou des régressions simples Pacifique occidental 11 6,3 (logistique, linéaire). Australie 4 2,3 Certaines études s’appuient sur des modèles Japon 3 1,7 statistiques (analyse de la variance, régression 1 0,6 Corée logistique multiple) avec des ajustements sur des Taiwan 3 1,7 plus ou moins nombreuses variables : le profil TOTAL 175 100,0 sociodémographique et socio-économique du patient, la sévérité de la maladie (stade de gravité d’une tumeur par exemple), les comorbidités, les caractéristiques des procédures chirurgicales, Sources de données et méthodes employées les autres traitements éventuellement administrés, etc. Un peu plus de la moitié des études (53 %) sont basées sur des données administratives (Cf. 3 études sur 4 portaient sur plusieurs années (de Tableau 2), la plupart étant représentatives à 2 à 10 ans). Ceci permet l’utilisation de techniques l’échelle nationale. Presque toutes ajustent par économétriques plus élaborées et plus pertinentes des indicateurs de comorbidité et les caractéris- dans certaines études prenant en compte la tiques des patients (âge, sexe). De nombreuses temporalité (économétrie de panel) ou certaines études sur données administratives introduisent caractéristiques des établissements hospitaliers également un contrôle du statut socio-économique (méthodes multi-niveaux). De manière générale, des patients. ces études comprennent une multitude de techni- ques et d’approches économétriques différentes et plutôt innovantes. Le type de données utilisées détermine la qualité de la mesure et du contrôle du case-mix : celui-ci s’améliore dans les publications basées sur un appariement de données administratives et

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 23 -

Tableau 2 Les interventions chirurgicales Nature des données utilisées dans la littérature

Nombre de Concernant les interventions chirurgicales, sur Type de données mobilisées publications 147 articles examinés, près d’un quart porte sur plusieurs interventions : au total, les articles Données administratives 81 retenus analysent 185 interventions chirurgicales Appariement de données cliniques et 44 distinctes. Trois articles ont examiné un ensemble administratives hétérogène d’interventions chirurgicales, l’un Données cliniques 50 d’entre eux analysant jusqu’à 16 interventions Suivis de cohorte/essais cliniques 12 différentes. Les pourcentages indiqués ci-des- Etude de cohorte incluse dans un sous feront référence au nombre d’articles 1 essai clinique randomisé examinés alors que le Tableau 3 présente Enquêtes ponctuelles 6 (sous le libellé « interventions chirurgicales ») Registres 26 le nombre total d’interventions étudiées par catégorie et les pourcentages qui s’y réfèrent. Registres avec suivi prospectif 5 a. Chirurgies du cancer

1.4.2. Soins et interventions étudiés La chirurgie du cancer est un objet d’analyse fréquent dans la littérature : 58 études parmi La grande majorité des recherches publiées celles retenues portent sur ce sujet et analysent portaient sur des interventions chirurgicales (85 %). un total de 90 interventions distinctes sur le cancer. Le reste des études s’intéresse à un ensemble hétérogène de soins, regroupant à la fois des En termes de résultats, seule une faible minorité soins médicaux et l’ensemble des prestations ne trouve pas d’association entre volume et qualité délivrées par des services et unités de soins spéci- (5 études). 86 % des études analysées révèlent fiques (service obstétrique, unité de soins intensifs, une relation positive entre volume et qualité. etc.). Enfin, 3 études consacrées aux soins de ville 3 études concluent à une association plus mitigée ont également été conservées dans le champ de dépendant fortement de la procédure analysée et/ notre revue systématique. ou de l’indicateur de qualité retenu.

Tableau 3 Sur le plan anatomique, les publications portent Nature des soins et des interventions étudiés le plus souvent sur un seul organe et une seule dans la littérature procédure chirurgicale, essentiellement sur Nombre Pourcentage l’appareil digestif (55 %) et dans une moindre Interventions chirurgicales dont : 147 84,0 mesure sur l’appareil urogénital (21 %). Mais une part non négligeable d’entre elles rassemble au Interventions liées à un 90 51,4 cancer/tumeur sein d’une même étude plusieurs localisations et, par conséquent, plusieurs procédures chirurgica- Interventions cardio-vasculaires 67 38,3 les (19 %). Plus précisément, la répartition des Interventions orthopédiques 18 10,3 études selon les localisations anatomiques des Autres 10 5,7 tumeurs est la suivante : Centres de soins médicaux dont : 14 8,0 Soins intensifs 1 0,6 – appareil digestif (55 %) : 8 études sur les résec- Traumatologie 10 5,7 tions de l’œsophage, 2 sur les pancréatectomies, Urgences 2 1,1 4 sur les pancréatico-duodénomectomies, 2 Tous soins confondus 1 0,6 sur les résections hépatiques, 8 interventions Soins médicaux à l’hôpital dont : 6 3,4 sur le côlon et/ou le rectum, 1 étude portant à la fois sur les gastrectomies et les colectomies ; Cardiologie 3 1,7 Respiratoire 2 1,1 – appareil urogénital (21 %) : 6 analyses sur les Néphrologie 1 0,6 interventions pour cancer du sein, 1 sur les Soins obstétriques dont : 5 2,9 ovaires, 1 sur l’utérus, 2 sur la prostate et 2 sur Accouchement non compliqué 1 0,6 la vessie ; Naissance à risque (sous-poids) 2 1,1 – 1 étude portant sur les pou- Soins intensifs pédiatriques/ 2 1,1 poumons (2 %) : néonataux mons (thoracotomie pour cancer du poumon) ; Soins de ville 3 1,7 Total 175 100,0

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 24 -

– autres (3 %) : 2 études portant sur des cancers Tableau 4 autres, 1 sur le cancer de la thyroïde et 1 sur Nature des interventions liées au cancer un cancer du cerveau (au niveau de la glande étudiées dans la littérature pituitaire) ; Nombre % Appareil digestif, dont : 32 55,2 – mixtes : cancers différents (9 %) : 5 études Oesophagectomie et/ou résection du cardia 8 13,8 intègrent des cancers de localisation différentes cités ci-dessus ; Estomac 2 3,4 Pancréas 8 13,8 – mixtes : cancers et autres tumeur (9 %) : Foie - Résection hépatique 3 5,2 5 autres études intègrent à la fois des cancers Côlon, rectum 10 17,2 de localisation diverse et des interventions hors Mixte (Gastrectomie, colectomie) 1 1,7 cancer, portant le plus souvent sur l’appareil cardiovasculaire mais aussi orthopédique. Urogénital, dont : 12 20,7 Sein 6 10,3 Cette répartition n’est pas représentative des Ovaires 1 1,7 interventions chirurgicales effectuées mais elle Utérus 1 1,7 montre cependant l’intérêt particulier qui existe pour les procédures relatives aux cancers dont Prostate (Prostatectomie radicale ou non) 2 3,4 l’incidence est très importante. Vessie (Cystectomie radicale) 2 3,4 Bronches, poumons 1 1,7 Sur le plan géographique, les États-Unis rassem- Autres 2 3,4 blent à eux seuls 60 % des études sur la chirurgie Etudes mixtes dont : 11 19,0 du cancer, suivis du Canada (14 %), du Royaume- Différentes localisations de cancers 6 10,3 Uni (3 études), de la Hollande (3), de l’Allemagne Cancers et autres interventions non (2), du Japon (2), et enfin de l’Espagne, de la Suède, liées au cancer 5 8,6 de l’Australie et de la Corée (chacun 1). Total 58 100,0

La plupart des études font état d’ajustements ou de Le nombre de patients inclus ou observés est très contrôles des résultats par un socle minimum de variable d’une étude à l’autre : de moins de 200 variables sociodémographiques et de comorbidité. (oesophagogastrectomie) jusqu’à près de 350 000 Certaines contrôlent également par le statut de dans le cas d’une étude mixte sur plusieurs types l’hôpital ou du chirurgien vis-à-vis de certaines de cancers et d’interventions, répartis dans un activités, tel que le degré d’enseignement dispensé. plus ou moins grand nombre d’hôpitaux.

S’agissant des cinq études qui ne repèrent aucune Les techniques chirurgicales employées peuvent relation entre volume et qualité, il s’avère qu’elles varier d’un malade à l’autre, notamment en fonction portent toutes sur des tumeurs de l’appareil du stade de sévérité de la tumeur, mais aussi d’un digestif et que trois d’entre elles s’appuient sur des établissement à l’autre et/ou d’un chirurgien à effectifs très faibles (Cf. Etudes n° 23, 117 et 162 l’autre. Le recrutement des patients opérés n’est dans l’Annexe 2). Par ailleurs, cette absence de pas non plus toujours homogène selon les zones résultat renvoie peut-être à la spécificité du lien géographiques. Dans le cas du cancer, il faut garder volume-qualité selon la nature de la maladie et la à l’esprit que souvent l’intervention chirurgicale procédure analysée, puisque deux études récentes n’est qu’une étape dans un processus de traitement et plus robustes sur le traitement du cancer plus global, relayée par exemple par de la chimio- colorectal ne mettent en évidence aucune corré- thérapie et/ou de la radiothérapie, ce qui n’est lation entre le volume d’intervention d’une part, et pas sans influence sur les indicateurs de qualité le nombre de rechutes et la mortalité d’autre part utilisés. En effet, la pratique de ces thérapies (n° 48 et 76). complémentaires doit se conformer aux recom- mandations de bonnes pratiques. Ainsi, pour Parmi les trois études présentant des résultats certains cancers étudiés ici, il est capital que plus contrastés, l’une concerne la résection l’analyse tienne compte de l’existence de ces pancréatique (n° 153), la seconde porte sur le traitements complémentaires lorsqu’ils sont indiqués. cancer de l’estomac (n° 27) et la troisième (n° 30) couvre cinq types de cancer (côlon, œsophage, foie, sein, poumons). b. Interventions cardio-vasculaires 67 études portent sur une ou plusieurs interven- tions cardio-vasculaires. L’intervention chirurgicale la plus souvent étudiée est le pontage aorto- coronarien (PAC). Le PAC est une intervention

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 25 -

à cœur ouvert, devenue relativement commune, de mortalité et de complication sont liés de manière coûteuse, présentant un risque élevé de mortalité significative au volume d’activité pour les inter- et qui nécessite une compétence technique ventions urgentes. Une autre publication (n° 141) importante. compare les taux de mortalité des hôpitaux classés par groupes selon des seuils d’activité utilisés par 18 études évaluent l’effet du volume de pontages Leapfrog6 aux Etats-Unis : au-delà de l’existence sur les résultats obtenus. A l’exception de deux réelle d’un lien volume-qualité, cette étude montre d’entre elles, toutes ces études montrent une qu’il existe des variations intra-classes importantes relation positive entre le volume d’activité des (c’est-à-dire des variations à l’intérieur de ces établissements et des chirurgiens et de meilleurs groupes d’hôpitaux définis). résultats, mesurés le plus souvent par la mortalité à l’hôpital et/ou à 30 jours. Les 9 études sur la réparation de l’anévrisme de l’aorte abdominale (AAA), qui est une intervention Parmi ces études, l’une d’entre elles observe les extrêmement délicate nécessitant une greffe dans résultats en termes d’infections nosocomiales l’aorte, trouvent toutes une relation négative entre après l’opération (n° 40), une autre compare le le volume d’activité et la mortalité. temps d’attente avant l’opération dans différents établissements (n° 45) et une troisième intègre La grande majorité des études examinant diffé- également les durées de séjours (n° 74). Près rentes interventions coronaires sont unanimes de la moitié des publications évalue l’activité des dans leurs conclusions : il existe un impact du chirurgiens en plus de (ou au lieu de) l’activité volume d’activité, non seulement des établisse- des établissements. Elles concluent toutes que ments, mais aussi des chirurgiens, qui permet le volume d’activité des chirurgiens a un impact de réduire la mortalité et d’augmenter la survie significatif sur la qualité. Une seule d’entre elles des patients. Ainsi, une étude (n° 10) démontre suggère que le volume d’activité du chirurgien n’a que, pour un chirurgien, c’est son volume annuel pas d’effet sur les résultats des PAC (réalisés à d’activité qui prend le pas sur son nombre d’années « cœur battant ») dans les établissements à haut d’expérience. Une autre étude, (n° 29) axée sur volume d’activité (n° 13). Mais cette étude trouve le remplacement de la valvule sigmoïde chez les une relation négative entre le volume d’activité personnes âgées, suggère que les établissements des chirurgiens (même dans les établissements à bas volume suivent moins bien les recomman- à forte activité) et la mortalité pour les pontages dations cliniques et les évolutions technologiques réalisés avec « circulation extracorporelle ». les plus récentes.

Tableau 5 Une étude japonaise n’observe aucun lien entre Nature des interventions cardio-vasculaires le nombre d’angioplasties réalisées pour infarctus étudiées dans la littérature du myocarde aigu (IMA) dans les établissements, Nombre % et les taux de mortalité (n° 127). La seule autre étude suggérant que le volume d’interventions Pontage aortocoronarien 18 26,9 coronaires par chirurgien n’a pas d’impact sur les Intervention coronaire percutanée 11 16,4 résultats vient des Etats-Unis, mais elle est basée Réparation de l’anévrisme de l’aorte sur un échantillon très réduit de 9 chirurgiens. abdominale 10 14,9

Angioplastie et thrombolyse 7 10,4 Une étude internationale compare les résultats Endartériectomie de la carotide 7 10,4 (mortalité à l’hôpital) des interventions coronaires Remplacement de la valvule sigmoïde 4 6,0 percutanées à partir de 4 registres européens, Angioplastie coronaire percutanée 3 registres japonais, 4 registres américains et 2 transluminale 4 6,0 registres internationaux (n° 126). Elle observe Pose de stent coronaire 2 3,0 que, dans chaque pays, il existe une relation inverse entre le volume d’interventions réalisé et Autres 4 6,0 la mortalité. En parallèle, l’étude montre que, dans Total 67 100 l’ensemble, la qualité des soins est meilleure au Japon, où même les établissements ayant une Une étude française exploitant les données de faible activité ont des résultats comparables aux registre de l’Agence Régionale d’Hospitalisation meilleurs établissements des autres pays. d’Ile-de-France suggère qu’il n’y a pas de relation entre le volume de pontages de l’établissement et le taux de mortalité pour les patients à bas 6 Leapfrog est un groupement d’employeurs qui achètent des soins risques (cas non-urgents et non compliqués), mais de santé aux États-Unis. Il vise à améliorer la qualité, la sécurité que les taux de complication y sont plus élevés et la transparence des soins achetés pour les Américains. Pour choisir les hôpitaux et juger de leur qualité, un des critères utilisés (n° 39). En revanche, l’étude montre que les taux est le volume d’activité concernant les interventions chirurgicales.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 26 - c. Interventions orthopédiques Tableau 7 Nature des autres interventions chirurgicales 18 articles évaluent diverses procédures chirurgi- étudiées dans la littérature cales orthopédiques, notamment la prothèse totale Nombre % de la hanche et du genou. La quasi-totalité confirme Chirurgie bariatrique 3 20,0 que les résultats en termes de mortalité, de réadmission, de complications postopératoires Cataracte 3 20,0 (telles que dislocations, embolie, infection, etc.) et Incontinence urinaire 1 6,7 de durée de séjour après opération s’améliorent Interventions gastro-intestinales 1 6,7 avec le volume d’activité des établissements et/ou complexes des chirurgiens (Cf. Tableau 6). Reconstruction pelvienne 1 6,7 Chirurgie pédiatrique 3 20,0 La plupart des études examinent à la fois le Dérivation ventriculo-péritonéale 1 6,7 volume d’activité spécifique de l’établissement et Hémorragie sous-arachnoïdienne 1 6,7 celui des chirurgiens. L’une d’entre elles (n° 37) Cholecystectomie laparoscopique 1 6,7 suggère que si le nombre (volume) de prothèses de hanche réalisées par le chirurgien est un Total 15 100 déterminant significatif des résultats, ce n’est pas le cas du volume global d’activité orthopédique de Parmi ces études, trois analysent l’impact de l’établissement. l’activité du chirurgien sur les résultats après une chirurgie de la cataracte : Une autre étude (n° 95) observe que la relation entre le volume des interventions et la qualité n’est – La première, une étude américaine (n° 175), pas systématique et que, dans le contexte améri- observe qu’il existe une relation positive et cain, le type de HMO7 ou d’assurance des patients significative entre le volume d’activité des peut avoir plus d’influence sur les résultats. chirurgiens et de meilleurs résultats (taux de complications) et que cette relation se confirme Tableau 6 même pour des seuils d’activité très élevés Nature des interventions orthopédiques (> 1 000 actes par an). étudiées dans la littérature – Une deuxième étude, anglaise (n° 15), trouve Nombre % des résultats similaires. Arthroplastie totale de la hanche 5 27,8 Arthroplastie totale du genou 4 22,2 – Toutefois, une autre étude menée par les Arthroplastie totale de l’épaule 3 16,7 mêmes auteurs (n° 14), réfute ces résultats en Intervention majeure sur la hanche introduisant un meilleur contrôle du case-mix. et le genou 1 5,6 Les auteurs soulignent l’importance de bien Fracture de la hanche 3 16,7 contrôler les facteurs de risque pour attester Scoliose pédiatrique 1 5,6 de la fiabilité d’une étude volume-qualité sur la chirurgie de la cataracte. Réparation du poignet 1 5,6 Total 18 100 La plupart des autres études sur les interventions listées dans le Tableau 7 confirment le lien signifi- catif et positif entre volume d’activité et qualité des résultats obtenus, sauf pour la pancréatectomie pédiatrique (n° 162). Notons toutefois qu’en d. Autres interventions chirurgicales raison de la rareté de cette opération, les résultats Nous avons également recensé 15 autres de cette étude ne s’appuient que sur cinq cas. interventions chirurgicales étudiées dans la littéra- ture qui ne se rattachent à aucune des catégories Soins médicaux et prises en charge de services précédentes (Cf. Tableau 7). spécifiques

Certaines études ne portent pas sur des interven- tions bien ciblées, mais sur des soins non chirurgi- caux ou sur des prises en charges globales assurées par des services et des unités de soins spéci ques. Au total, nous avons repéré 28 7 Le Health Maintenance Organization (HMO) constitue une forme fi d’organisation de managed care qui contracte directement avec études qui entrent dans ce cadre. les producteurs de soins (sur une base tarifaire et de contenu des prestations de soins délivrées) et qui couvre les assurés à la condi- tion que ceux-ci ne consultent que les producteurs de soins agréés par le HMO.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 27 -

Cinq études portent sur des procédures obstétri- contexte plus large. La première étude porte sur ques et sur la prise en charge des nouveau-nés la relation entre le volume d’activité (visites en am- prématurés. Les études analysant l’activité dans bulatoire et séjours) des centres psychiatriques les centres de soins intensifs pour les prématurés et la qualité des soins fournis mesurée par les concluent en général que la mortalité néonatale indicateurs HEDIS8 : elle établit une relation diminue avec l’activité du centre ; seule une étude positive (n° 8). anglaise apporte des conclusions différentes. Sur deux études observant la relation entre le volume La deuxième étude évalue le nombre de patients d’accouchements des femmes enceintes à faible traités par médecin généraliste et conclut que risque et les taux d’intervention pour complica- les patients âgés diabétiques consultant des tions, ainsi que la mortalité néonatale, la première médecins ayant une activité importante ont plus (australienne) ne constate aucune corrélation, de chances de recevoir des soins de bonne qualité tandis que l’autre (allemande) observe des taux (n° 4). En revanche, une troisième étude américaine de mortalité plus élevés dans les cliniques à faible arrive à la conclusion inverse : elle montre que volume d’activité. Précisons que, contrairement à les médecins ayant une patientèle importante l’étude allemande, l’étude australienne ne contrôle réalisent leurs consultations plus rapidement, ce pas les facteurs de risques. qui a un effet néfaste sur la qualité, mesurée par la satisfaction des patients et le taux des soins Nous avons isolé 6 études qui s’intéressent préventifs fournis (n° 105). davantage à la prise en charge globale à l’hôpital des pathologies (principalement cardiaques mais Effet pervers d’une trop forte activité aussi respiratoires) qu’à des interventions spécifi- ques. Elles suggèrent toutes que les patients traités Par ailleurs, nous avons repéré 3 études (n° 172, dans des établissements à haut volume ou par 173 et 174) démontrant qu’un volume d’activité trop des médecins à forte activité (expérimentés) ont élevé pouvait avoir des répercussions néfastes des taux de survie plus élevés. Une autre étude sur la qualité des soins et sur les résultats porte sur la prise en charge d’une maladie rare sanitaires. Une première étude relève une relation (systémique lupus erythematosus) et suggère positive entre le volume des patients traités par également que l’expérience des médecins (mesu- infirmier et la survenue d’événements indésirables. rée par le volume de patients traités) est un facteur Deux autres études examinent l’activité des important de réduction de la mortalité. services d’urgence et montrent que la mortalité hospitalière y est plus élevée dans les périodes 14 études évaluent le volume d’activité de services de suractivité. Notons que ces études cherchent à ou d’unités spécifiques de prestation de soins, établir l’impact d’une « surcharge » (par exemple notamment les centres de traumatologie. Les un taux d’occupation supérieur à 100 %) sur les résultats concernant ces centres de traumatologie résultats et emploient des méthodes différentes ne semblent pas tranchés puisque la moitié des de la plupart des études traditionnelles sur le lien études mettent en évidence une relation signifi- volume-qualité. Nous les avons incluses ici à titre cative entre le volume d’activité des centres et la illustratif car elles permettent de remettre en qualité, tandis que l’autre moitié observe que le vo- perspective la relation volume-qualité, en pointant lume n’a pas d’effet sur les résultats. Cependant, les risques potentiels d’une suractivité. même les études validant la corrélation volume- qualité tempèrent cette conclusion en indiquant que cette relation est surtout vérifiée pour les cas les 1.4.3. Dé nition de la qualité plus compliqués. Enfin, une publication portant sur fi les unités de soins intensifs pour trois diagnostics principaux, montre que le volume a un effet Le taux de mortalité est l’indicateur de résultat le significatif pour les patients souffrant d’affections plus couramment retenu. Le plus souvent, il s’agit respiratoires compliquées ou gastro-intestinales, du taux de mortalité à l’hôpital ou bien à 30 jours. mais pas pour ceux ayant des problèmes De nombreuses études s’intéressent également à neurologiques. la survie du patient sur des périodes plus longues, de 1 à 5 ans (voire 10 ans). Une autre mesure fréquemment utilisée, si l’intervention ou les Soins de ville soins étudiés s’y prêtent, est la durée du séjour d’hospitalisation. La survenue de réadmissions Citons également 3 études fondées sur l’analyse du volume d’activité en soins de ville. Même si ces études n’entrent pas dans notre champ d’analyse 8 HEDIS (Health Employer Data and Information Set), qui couvre 73 millions d’Américains, est la plus importante base de données stricto sensu, elles ont été conservées dans la aux Etats-Unis. Elle fournit de nombreux indicateurs pour mesurer la revue de la littérature dans l’objectif de replacer performance des soins de santé mentale, notamment en termes de les résultats des études hospitalières dans un suivi après hospitalisation et de pertinence des traitements médica- menteux (antidépresseurs).

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 28 - non programmées, de ré-interventions ou de – le taux de survie globale (à 5 ans, 6 ans, 10 complications postopératoires est aussi fréquem- ans et à plus long terme) ou le taux de survie à ment étudiée. Certaines études s’intéressent, 5 ans sans rechute ; enfin, aux infections ou aux autres événements indésirables survenus à l’hôpital pendant ou après – le taux d’apparitions de certaines séquelles les soins. (ex. colostomie permanente) ;

Les indicateurs de qualité utilisés concernent, par – le temps de survie avant décès par cancer ou ordre de fréquence : quelle qu’en soit la cause ;

– la mortalité hospitalière, la mortalité à 30 jours, – le taux d’échec du chirurgien. 60 jours et à 2 ans ; De manière générale, on constate que les – les complications postopératoires et les décès indicateurs utilisés pour évaluer les résultats sont par complications postopératoires ; de plus en plus variés et le plus souvent adaptés aux procédures spécifiques étudiées. Par exemple, – la durée de séjour totale ou celle en soins les études sur des interventions complexes intensifs ; présentant un risque de mortalité élevé se concentrent sur les taux de mortalité tandis que – le taux de réadmission à l’hôpital ; toutes les études sur les interventions orthopé- diques, telles que la prothèse totale de la han- – le taux de ré-intervention ; che ou du genou, examinent non seulement les taux de mortalité mais aussi plusieurs indicateurs – le taux de rechute ; de complications postopératoires à long terme (Tableau 8).

Tableau 8 Indicateurs de résultats utilisés selon les interventions étudiées Chirurgie Chirurgie Centres de du cancer cardiaque Chirurgie orthophédique traumatologie Obstétrique

% des études mobilisant cet indicateur de résultats Mortalité hospitalière 60,3 71,6 11,1 80,0 60,0 Mortalité Mortalité à 30 à 90 jours 24,1 19,4 55,6 - - Taux de survie à long terme (6 mois à 10 ans) 29,3 - - 40,0 -

Réadmissions/ réinterventions à 30 j 10,3 1,5 - 10,0 - Réhospitalisation Réadmissions/ -- 27,8 - - réintervention de 60 j à 5 ans

Complications post-opératoires 29,3 25,4 50,0 - - Durée moyenne de séjour (DMS) 25,9 10,4 44,4 40,0 20,0 Coût hospitalier 10,3 6,0 16,7 - 20,0 Autre Résultats cliniques du patient 0,0 17,9 11,1 - 20,0 Infections nosocomiales 0,0 3,0 27,8 - 20,0 Autres 8,6 7,5 - - -

Guide de lecture : 60,3 % des études portant sur la chirurgie du cancer ont utilisé la mortalité hospitalière comme indicateur de résultat. La plupart des études utilisent plusieurs indicateurs de résultats conjointement, ce qui explique que les sommes des pourcentages en colonne soient supérieurs à 100 %.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 29 -

1.4.4. Définition du volume d’activité ment acceptés. Par exemple, plusieurs études américaines ont remis en question la pertinence des L’indicateur de volume utilisé préférentiellement découpages a priori et la validité des seuils de est le nombre de séjours spécifiquement lié à volume utilisés comme indicateurs de qualité par l’intervention étudiée, et mesuré au niveau de le groupe Leapfrog aux Etats-Unis. Ces études l’hôpital et/ou du chirurgien. Parfois, on y adjoint le indiquent que les variations de mortalité intra-classe volume de consultations réalisées. demeurent importantes dans les classes créées par Leapfrog. Variable continue ou variable en classes Volume d’activité mesuré par établissement Il existe deux façons de considérer le volume : ou par médecin ? comme une variable continue ou comme une variable en classes. Considérer le volume comme La plupart des études se sont concentrées sur une variable continue implique que les résultats le volume d’activité global, mesuré au niveau de cliniques varient de manière continue (voire linéaire) l’établissement de soins. Cependant, plus d’un selon les différentes valeurs du volume d’activité. tiers des études (63) s’intéressait également au Dans une grande majorité des études, le volume nombre d’interventions réalisée par les praticiens, est considéré comme une variable nominale, qui chirurgiens notamment. Ces études démontrent regroupe les établissements/chirurgiens dans des majoritairement (96 %) une relation positive entre catégories définies par des seuils de volumes. la qualité des résultats et le volume d’activité de Ces seuils de volume et les catégories qui en ceux-ci. découlent peuvent être fixés a priori ou a posteriori. La définition a posteriori des catégories est parfois Il est clair que l’activité d’un établissement et de critiquée, avec l’argument que l’analyste peut ses chirurgiens sont corrélées. Il n’existe toutefois choisir des points de découpage pour définir des pas de conclusions définitives sur l’importance catégories de volume qui optimisent la corrélation relative de chacune dans le niveau de résultats. volume-résultats. Définir des catégories de volume En effet, peu d’études ont essayé de distinguer, en a priori, avant l’observation des données, est utilisant par exemple des modèles hiérarchiques, censé garantir une plus grande objectivité. Par l’effet du volume de l’établissement de l’effet de ailleurs, notons que dans la plupart des études l’activité du chirurgien. De manière générale, ces examinées ici, les seuils sont définis de façon études suggèrent, d’une part, que le volume à créer un nombre limité de classes de volume d’activité de l’établissement et du chirurgien sont d’importance égale à partir de la distribution tous les deux significatifs même lorsqu’ils sont observée (de 2 à 4 le plus souvent, fondées sur la contrôlés simultanément et, d’autre part, que médiane ou les quartiles). Le nombre de groupes l’impact spécifique du volume d’activité des chirur- peut monter jusqu’à 10 quand le découpage se giens persiste même dans les établissements à très fonde sur les déciles de la distribution. forte activité. Cela étant, l’importance relative du volume d’activité de l’établissement ou du chirurgien Le découpage en classes de volume d’activité, semble varier d’une procédure à l’autre et en d’une part, permet une plus grande souplesse fonction de l’indicateur de résultat utilisé. dans la forme du lien testé entre volume et qualité et, d’autre part, facilite la lecture des résultats et leur interprétation par les utilisateurs. Seulement 1.4.5. Ajustement par les caractéristiques deux études ont traité le volume uniquement cliniques du patient (case-mix) comme une variable continue, tandis qu’un quart des études ont utilisé la combinaison des deux Les caractéristiques des patients, telles que l’âge, approches : par exemple en modélisant d’abord le le sexe, la gravité de la maladie, les comorbidités volume en tant que variable continue pour établir existantes, le statut socio-économique, etc. le lien volume-qualité, puis en introduisant des influent sur les résultats d’une intervention ou d’un classes de volume pour évaluer les niveaux de traitement. Dans les études qui tentent d’expliquer risque. les variations des résultats obtenus en fonction du volume d’activité, il est important de s’assurer que De manière générale, on peut donc dire que la ces caractéristiques n’influencent pas la relation définition de « haut » ou « bas » volume varie volume-qualité de manière indue. La revue de d’une étude à l’autre et dépend largement de la littérature de York (Sowden et al., 1997) avait l’intervention/procédure étudiée. Il est très difficile notamment critiqué la qualité de l’ajustement par de fournir un résumé des seuils utilisés par les caractéristiques cliniques des patients (ca- intervention en raison de la multitude des méthodes se-mix) dans les études évaluées : «... la plupart employées. Au final, pour la plupart des interven- des recherches ne contrôlent pas suffisamment tions, il n’existe pas de seuils d’activité unanime-

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 30 - les caractéristiques cliniques des patients traités Signalons que 40 études introduisent également dans différents établissements ». des variables socio-économiques relatives aux patients, en s’appuyant généralement sur l’appa- De ce point de vue, on constate que les études riement entre différentes bases de données. Les réalisées au cours des dix dernières années se études fondées sur des registres de pathologies sont beaucoup améliorées. Nous n’avons repéré spécifiques présentent dans l’ensemble un très aucune étude hospitalière qui ne contrôle pas les bon contrôle de la morbidité, assis sur un large facteurs de risque basiques des patients. Au mi- spectre d’indicateurs cliniques. Par exemple, les nimum (dans les études basées sur des données études sur les interventions pour cancer ajustent administratives), les résultats sont ajustés par par le type de tumeur, le stade de gravité et l’état l’âge, le sexe, ainsi que par un indice de comorbi- de santé général des patients (tension artérielle, dité (tel que l’indice de Charlson). Mais la plupart diabète, problèmes cardiaques, etc.). des études introduisent d’autres types de contrôle clinique, en testant différents indices de gravité et de morbidité en plus de l’indice de Charlson.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 Partie 2 Causalité entre volume d’activité et qualité : éléments théoriques

- 33 -

2. Causalité entre volume d’activité et qualité : éléments théoriques

2.1. Introduction semble être particulièrement pertinente pour appréhender la façon dont le volume influence la Malgré l’existence d’un important corpus de qualité (section 2.3). Nous introduisons d’abord recherches empiriques sur le lien entre les volumes les aspects théoriques de la courbe d’apprentis- d’activité de soins spécifiques et les résultats en sage, puis nous cherchons à identifier la façon termes de santé des patients (Cf. partie 1), peu dont le volume est lié aux coûts et aux résultats, d’études se sont concentrées sur l’explication de en exploitant les résultats empiriques provenant ce lien. Deux questions sont essentielles pour d’autres secteurs d’activité. Dans une dernière comprendre le lien volume-qualité : d’une part, le section, nous proposons une synthèse des ensei- sens de la causalité entre volume et amélioration gnements pouvant être tirés des études réalisées de la qualité et d’autre part, les dynamiques sous- sur l’hôpital, en observant les études empiriques jacentes expliquant cette causalité. Si le volume questionnant le sens de la causalité entre volume génère de la qualité, il est important de comprendre et qualité et l’importance de l’effet d’apprentissage, pourquoi et comment en identifiant les mécanis- ainsi que les mécanismes sous-jacents (section mes qui sous-tendent ce lien. 2.4).

En premier lieu, dans le contexte du marché hospitalier, le sens de la causalité dans la relation 2.2. Les économies d’échelle observée entre volume et qualité doit être questionné. Deux hypothèses principales ont été Dans le cas d’une entreprise, on dit qu’il existe mises en avant : des économies d’échelle sur une gamme de produits si les coûts moyens de long terme baissent – La théorie des effets d’apprentissage (ou «la à la suite d’une augmentation dans l’échelle de pratique rend parfait»), selon laquelle la qualité production. La théorie économique néoclassique des soins dispensés par les médecins et les postule souvent une relation entre les coûts de hôpitaux augmente lorsqu’ils soignent plus de long terme et la taille de l’entreprise, qui suit une patients (grâce à une expérience accrue). Cela courbe en U. Pour les hôpitaux, qui sont des suggère notamment que la relation volume- entreprises produisant une multiplicité de biens qualité résulte d’économies d’échelle ; et de services, de nouveaux concepts ont été introduits afin de distinguer les économies d’échelle – La théorie des « renvois sélectifs », selon d’ensemble (ray scale) des économies spécifi- laquelle on oriente plus volontiers les patients quement reliées à un produit donné. De manière vers des médecins et des établissements générale, ce terme englobe tous les facteurs qui de bonne réputation qui offrent des soins de expliquent que les coûts unitaires de production qualité. Contrairement aux effets d’apprentis- diminuent lorsque la taille d’une entreprise ou de sage, la théorie des renvois sélectifs suppose l’un de ses services augmente. que c’est la qualité qui engendre le volume plutôt que l’inverse. Enfin, le concept « d’économies de gamme » La plupart des travaux théoriques et empiriques désigne les réductions dans les coûts de consacrés au lien volume-qualité proviennent production réalisées lorsqu’une entreprise élargit de l’économie industrielle et utilisent le concept sa gamme de produits et de services (production « d’économies d’échelle » pour expliquer le lien conjointe). Il faut pour cela qu’elle puisse utiliser volume-qualité. Les sources d’économies des inputs identiques, les mêmes installations d’échelle peuvent être statiques ou dynamiques. et le même personnel pour produire plusieurs Les effets d’apprentissage, correspondant à produits ou services. une production plus efficiente avec l’expérience accumulée, ont été particulièrement bien démontrés Globalement, deux sources d’économies d’échelle dans un grand nombre de secteurs d’activité. sont distinguées : les économies d’échelle internes et les économies d’échelle externes. Ces Cette partie est organisée de la manière suivante : deux termes sont utilisés différemment par les nous présentons tout d’abord (section 2.2) le divers auteurs selon le degré d’agrégation choisi concept d’économies d’échelle, nécessaire pour pour séparer les deux niveaux. Par exemple, dans comprendre les questions qui se rapportent le domaine hospitalier, les économies d’échelle aux volumes d’activité des établissements internes se référeront souvent à l’hôpital, et les de soins. Nous développons ensuite la économies d’échelle externes à la région (ou au littérature sur les effets d’apprentissage, qui secteur sanitaire).

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 34 -

Pour comprendre comment le volume d’activité facteurs sont contrôlés par ailleurs, bien qu’il peut améliorer les résultats sanitaires dans les semble extrêmement difficile d’ajuster toutes les établissements de soins, la littérature sur les différences de case-mix et de qualité des soins économies d’échelle internes semble particulière- entre les différents hôpitaux. ment intéressante. Cependant, il est impossible d’évoquer les économies internes sans considérer Plus récemment, des travaux – limités mais également les répercussions que pourrait avoir néanmoins en plein développement – ont examiné la taille de l’établissement sur d’éventuelles l’impact sur la qualité des soins de la concentration économies d’échelle externes. Nous introduisons ou du niveau de concurrence dans le marché ici brièvement les économies d’échelle externes hospitalier. Gaynor (2004, 2006) présente un bon dans la mesure où elles occupent une place résumé de la littérature théorique et empirique importante dans la littérature en économie de la concernant la concurrence sur le marché de la santé et qu’il est indispensable de comprendre santé, et ses impacts sur la qualité et le bien- leur influence sur la performance d’un établis- être social. Il souligne que la théorie montre que sement de santé donné. Par la suite, nous nous la concurrence contribue à augmenter la qualité restreindrons aux aspects internes des économies lorsque les prix sont fixés (régulés), comme c’est d’échelle, c’est-à-dire aux économies d’échelle le cas de la plupart des services sur le marché réalisées au niveau d’un établissement donné. hospitalier français. Toutefois, les démonstrations empiriques semblent parfois contradictoires et invitent à ne pas tirer de conclusions définitives 2.2.1. Les économies d’échelle externes sur l’impact de la structure du marché sur les indicateurs de qualité, notamment sur la mortalité à l’hôpital. Les limites méthodologiques de ces Les économies d’échelle externes sont constituées études sont bien soulignées par Gaynor. Retenons par les avantages qu’une entreprise tire du pouvoir surtout que cette littérature restreinte provient qu’elle exerce sur son environnement (Vettori, majoritairement des États-Unis et qu’elle suggère, 2000). De ce fait, elles découlent de l’action de globalement, que la concurrence contribue à plusieurs agents économiques. Une entreprise de augmenter la qualité. grande taille qui aurait une situation de monopole ou une situation de leader pourrait, si la concurrence est mal organisée, jouir du pouvoir de fixer les prix 2.2.2. Les économies d’échelle internes sur le marché. Ceci renvoie à la question de la taille de l’entreprise, appréhendée de façon absolue Les économies d’échelle internes résultent des ou relative (par rapport à la concurrence locale), avantages procurés par la taille au niveau de la et fait donc référence à un éventuel « pouvoir de gestion et de la production pour une entreprise monopole ». donnée (ou un seul service). Cela fait référence à toutes les améliorations de gestion rendues Les économies d’échelle externes ont une influence possibles par l’accroissement de la taille, d’une sur la performance des établissements et d’un part avec la standardisation, la mécanisation, etc. secteur car elles influencent le comportement et d’autre part par l’amélioration de l’organisation de la concurrence et les relations de pouvoir qui et la planification (production « à temps », automa- prévalent. Elles touchent directement à l’analyse tisation, mais aussi partage du temps de travail, de la concentration sectorielle. informatisation, etc.). Ceci démontre l’importance de la technique de production lorsque le coût de Le marché hospitalier a traditionnellement constitué production croît plus lentement que la capacité un champ d’application important pour les de production générée, mais aussi l’importance politiques antitrust et la littérature sur les écono- de l’apprentissage lorsque le nombre d’erreurs mies/des économies d’échelle et de gamme dans diminue avec la quantité produite et que la le secteur hospitalier est abondante. Il existe, par production est mieux répartie. exemple, de nombreux travaux sur l’existence d’économies d’échelles et sur leur importance On peut alors distinguer deux sources d’économies dans la production afin de déterminer la meilleure d’échelle internes : statiques et dynamiques. configuration pour l’ensemble des services Les économies d’échelle internes statiques sont hospitaliers, ainsi que le niveau d’activité optimal directement liées au processus de production dans les hôpitaux. La revue de York (Sowden et et permettent une réduction des coûts unitaires al., 1997) propose une évaluation synthétique par une augmentation de la production à un de cette littérature et conclut que les économies moment donné (l’élasticité des coûts par rapport à la d’échelle semblent concerner davantage les production est inférieure à 1). Les coûts unitaires petits hôpitaux. De plus, les auteurs rappellent diminuent au moment t à cause d’une baisse des que des estimations fiables d’économies d’échelle coûts marginaux ou à cause de l’existence de ne peuvent être obtenues que si tous les autres coûts fixes de production.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 35 -

Les économies d’échelle internes dynamiques La forme fonctionnelle la plus appropriée pour la mettent en rapport la diminution des coûts unitaires courbe d’apprentissage n’a pas véritablement été avec l’augmentation de la production cumulée, testée, mais un grand nombre d’études ont montré ce qu’on appelle également des effets que la relation log-linéaire était bien adaptée. d’apprentissage. Les économies d’échelle dyna- La courbe d’apprentissage a été utilisée comme miques peuvent aussi provenir d’améliorations de fondement dans des stratégies de fabrication, des l’organisation du travail, de progrès technolo- processus marketing, mais aussi pour mettre en giques et de la capacité des employés à mieux œuvre des politiques antitrust. travailler (Vettori, 2000). Si la présence d’une courbe d’apprentissage est bien documentée dans un grand nombre Bien qu’en théorie l’effet d’apprentissage ne soit d’industries, les dynamiques du lien entre volume que l’une des sources possibles d’économies (production cumulée) et apprentissage sont d’échelle internes, c’est sur ce sujet spécifique moins bien établies. L’apprentissage pourrait se que la littérature a été la plus riche et la plus développer à travers l’expérience accumulée par essentielle pour comprendre les mécanismes du les travailleurs ou par l’encadrement (manage- rapport de causalité entre volume et qualité. ment), une meilleure coordination des processus de production, mais aussi à travers un investis- 2.3. Théorie de la courbe sement croissant dans des équipements plus performants. d’apprentissage (learning by doing) À la fin des années soixante, les travaux du Le terme général de fonction d’apprentissage, ou Boston Consulting Group ont démontré la grande fonction de progrès, a été introduit dans la littérature utilité du concept de la courbe d’apprentissage à partir d’études plus anciennes portant sur (ou « courbe d’expérience » comme les auteurs l’industrie aéronautique, la construction navale et l’ont qualifiée) et ont suggéré que cette courbe l’industrie lourde. Dans les années trente, Wright d’apprentissage était la résultante de l’apprentis- (1936) a observé que le coût direct généré par la sage au travail, de l’amélioration des processus production d’une cellule d’avion diminuait avec la de production (organisation et coordination du quantité cumulée de cellules d’avions produites. travail), de la standardisation de la production Ce résultat a été largement corroboré par d’autres et d’économies d’échelle statiques. Ils n’ont pas études réalisées dans les années quarante dans essayé de distinguer le rôle de chacun de ces le secteur de l’industrie aéronautique - mais aussi différents éléments. Mais plusieurs études ont par la suite dans une variété de secteurs indus- par la suite essayé de différencier les écono- triels – et qui ont démontré que le coût marginal mies d’échelle statiques - qui postulent que le de production décroît avec l’expérience cumulée volume actuel affecte les résultats actuels - des (Hirsch, 1952 ; Hirsmann, 1956 ; Liebermann, effets dynamiques d’apprentissage qui postulent 1984). Cette chute du coût marginal est connue que le volume actuel affecte à la fois les résultats sous le nom d’effet d’apprentissage (learning actuels et futurs. La plupart des études industriel- by doing). Des études plus récentes ont égale- les suggèrent que les économies d’échelle stati- ment mis en évidence l’existence d’une courbe ques sont significatives, mais restent beaucoup d’apprentissage dans les sociétés de services moins importantes que les économies générées (Darr et al.1995). par les effets d’apprentissage (Stabaugh et Town- send, 1975 ; Lieberman 1984). Lieberman (1984) a La forme conventionnelle de la courbe d’appren- également démontré qu’il existe des différences, tissage est : faibles mais néanmoins significatives, dans la pente de la courbe d’apprentissage entre diffé- rents produits. L’effet de l’apprentissage est Y = αX-b (1) d’autant plus important (c’est-à-dire que la pente de la courbe d’apprentissage est où Y désigne le coût direct pour produire la Xème d’autant plus grande) que les dépenses unité. X désigne la production cumulée (volume), en recherche et développement, ainsi que α est le coût de production de la première unité, l’investissement en capital (humain et techni- et b « l’élasticité d’apprentissage » qui définit que) dans une entreprise sont élevés. le taux de réduction des coûts à mesure que le volume (production cumulée) augmente. Dans Toutefois, certaines études ont démontré qu’il cette équation, la production cumulée X est une existe des limites aux potentialités des effets variable proxy de la connaissance acquise à d’apprentissage ceux-ci sont notamment condi- travers la production. tionnés à la capacité de l’encadrement (management) à tirer le meilleur parti des ressources mises à sa disposition (Kimberly, 1981 et Huber, 1991).

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 36 -

De plus, l’hypothèse selon laquelle l’apprentissage 1987). Il est probable que certains patients (ou leurs est associé au volume implique que les effets de médecins) choisissent des hôpitaux en fonction l’apprentissage et l’expérience ainsi accumulée de leurs résultats et de leur réputation plu- persistent au cours du temps. Or, les entreprises tôt que ceux qui sont situés à proximité ; cela et les organisations peuvent perdre (partielle- aurait pour conséquence d’augmenter le volume ment) le gain de l’apprentissage et donc oublier des patients dans les meilleurs hôpitaux. Ce l’expérience accumulée en raison du turnover des concept est baptisé « renvoi sélectif ». On s’attend employés, des périodes d’inactivité ou de l’incapacité à ce que ce renvoi sélectif soit plus massif pour à institutionnaliser le capital de connaissances des situations non urgentes et nécessitant acquis (Argote et al., 1990 ; Besanko et al., 2007). l’adressage d’un médecin, mais qui présentent des Ainsi, la recherche académique semble maintenant risques et/ou des complications élevés. avoir établi que les taux d’apprentissage peuvent différer selon les situations d’apprentissage et en Le sens de la causalité a des implications fonction des biens et services produits. importantes pour la mise en œuvre des politiques de planification dans le secteur hospitalier. Si le lien existant entre volume et qualité est majoritai- rement dû à des effets d’apprentissage, la 2.4. L’effet d’apprentissage dans le centralisation de certains services serait notam- contexte hospitalier ment susceptible de présenter un intérêt général. Si l’explication principale est plutôt le renvoi sélectif, Par définition, la courbe d’apprentissage s’applique la régionalisation pourrait entraîner une réduction aux coûts de production. Or, dans la littérature de la concurrence, sans améliorer la qualité médicale, l’apprentissage est mesuré en termes observée. En outre, la forme fonctionnelle de cette d’amélioration des résultats (c’est-à-dire en général relation « volume-qualité » ainsi que les dynami- une mortalité plus faible ou moins d’événements ques d’apprentissage, c’est-à-dire la façon dont indésirables), et non en termes de coûts des l’apprentissage se développe avec l’expérience soins. et la façon dont il est transféré, constituent des questions importantes pour les établissements de La plupart des études initiales sur la relation santé et en termes de planification. entre le volume d’activité chirurgical et les résultats montrent une relation négative entre nombre de patients opérés et mortalité. Ces résultats sont 2.4.1. Comment le volume engendre-t-il souvent interprétés comme la confirmation de la qualité des soins ? l’existence d’une courbe d’apprentissage (Luft et al., 1979) et donc la justification de concentrer Tout d’abord, nous ne savons pas avec certains services de soins. Il est, en effet, certitude si la relation entre le volume d’activité et la raisonnable de considérer qu’une équipe chirurgi- qualité est linéaire, simplement continue ou définie cale qui effectue une cinquantaine d’interventions par un (ou des) point(s) d’inflexion. Le graphique 2 cardiaques par mois est plus compétente qu’une ci-dessous présente quelques formes possibles. équipe qui n’effectue qu’une ou deux interventions Par exemple, introduire un seuil d’activité pour par mois. orienter les patients suppose qu’il existe un seul point de coupure (graphique 2.c). Si l’idée d’intro- Cela étant, dans le cas du secteur hospitalier, duire une seule référence de volume pour certains auteurs ont avancé que le lien observé distinguer la qualité des hôpitaux est attractive, entre volume et qualité pouvait refléter un phé- cela peut être trop simpliste. Les études prove- nomène d’adressage spécifique dans lequel les nant des autres secteurs d’activité montrent par patients sont orientés vers les hôpitaux ayant de ailleurs que la forme la plus appropriée pour la meilleurs résultats (Dranove, 1998 ; Luft et al., courbe d’apprentissage est la forme log-linéaire (graphique 2.b).

Graphique 2 Les formes envisageables de la relation « volume d’activité - mortalité » a b c Mortalité Mortalité Mortalité

Volume d'activité Volume d'activité Volume d'activité

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 37 -

Par ailleurs, la différence entre les économies au cours de la période était davantage due à la d’échelle statiques et les économies d’échelle diffusion technologique qu’à des effets d’apprentis- dynamiques (effets d’apprentissage) est importante sage. Elle suggère ainsi que la régionalisation de pour comprendre cette relation. En ce qui concerne l’intervention ACPT peut entraîner des économies les économies d’échelle statiques, le lieu physique en termes de coûts par patient, mais seulement (hôpital) dans lequel volume est concentré importe de faibles améliorations dans les résultats peu : un volume d’activité élevé générera de bons sanitaires. résultats quel que soit l’hôpital concerné. Pour ce qui est de l’effet d’apprentissage, le lieu ou le mode d’organisation de l’hôpital importe davantage : 2.4.2. Effet d’apprentissage ou renvoi transférer l’activité d’une organisation (hôpital) à sélectif ? une autre peut entraîner la perte de l’expérience acquise au sein du premier hôpital. De plus, les bénéfices tirés de la concentration du volume dans L’étude de Luft et al. (1987) est la première à un seul hôpital se développeraient au fil du temps. avoir testé explicitement l’impact du renvoi sélectif Cela implique que les bénéfices nets d’une fusion par rapport à celui de l’effet d’apprentissage des établissements hospitaliers seraient moins (dénommé par les auteurs « la pratique rend importants dans les approches de régulation parfait »). En exploitant les données de mortalité régionale. liée aux 17 interventions/diagnostics en 1972 dans 900 établissements américains, ils modélisent simultanément l’impact du volume sur la Dans l’une des rares études analysant l’importance mortalité et l’impact de la mortalité sur le volume relative des économies d’échelle statique par rapport d’activité des établissements. Les auteurs à l’apprentissage au cours du temps, Gaynor observent que les hôpitaux présentant de et al. (2005) mettent en cause l’importance de faibles taux de mortalité attirent les pa- l’effet d’apprentissage pour les procédures de tients hospitalisés pour anévrisme de l’aorte pontage aorto-coronarien. En examinant la abdominale (AAA), fracture du fémur, ulcère mortalité dans les hôpitaux californiens entre 1983 gastrique, ablation de la prostate et cardiaque et 1999, les auteurs suggèrent que le lien volume-qua- by-pass (renvoi sélectif). En revanche, pour les lité est expliqué principalement par les économies infarctus du myocarde, les syndromes de d’échelles statiques. Ils font également une détresse respiratoire, hystérectomie, opéra- estimation du niveau d’oubli (érosion des connais- tions sur l’estomac et sur l’intestin et la PTH, ils sances acquises) en introduisant les valeurs des trouvent que l’effet d’apprentissage est signi catif. volumes annuels d’activité des années antérieures fi Les résultats pour les autres interventions étudiées dans leur régression de mortalité. Leurs estimations ne sont pas concluants (cirrhose, appendicectomie, montrent que les volumes ont uniquement un effet angioplastie cardiaque, cystectomie, réparation immédiat sur les résultats (seuls les volumes des d’hernie, les hémorragies sous-arachnoïdiennes). années présentes sont influents), ce qui suggère Ils en concluent que l’importance relative des que le phénomène d’« oubli » est significatif. phénomènes de renvoi sélectif et d’apprentissage Mais les auteurs reconnaissent qu’ils n’ont pas pu varie selon les diagnostics ou la procédure, en examiner correctement la structure de décalage fonction des aspects cliniques, mais que les deux dans le temps dans la fonction d’apprentissage, explications sont valides. en raison de la multicolinéarité des données.

Farley et Ozminkowski (1992) soutiennent éga- Vivian Ho (2002, n°161) avance par ailleurs que lement que l’impact du volume sur les résultats l’amélioration des résultats peut être liée au temps varie d’une procédure à l’autre et que le renvoi (progrès technologique) et à l’accumulation d’une sélectif, ainsi que l’effet d’apprentissage, connaissance collective : le présupposé est que peuvent être importants. À partir d’une analyse l’on peut également apprendre en observant des donnés longitudinale de 500 hôpitaux les pratiques des autres (learning by watching). américains (données de panel), ils concluent L’auteur utilise des données longitudinales (de qu’une augmentation de volume d’activité réduit 1984 à 1996) pour examiner le lien entre le la mortalité hospitalière pour l’infarctus du myo- nombre d’ACPT (angiographies coronaires carde, la réparation de hernie et le syndrome de percutanées transluminales) réalisées dans les détresse néonatal. En revanche, ils concluent que années précédentes et les résultats actuels en le renvoi sélectif est signi catif pour le pontage termes de coût et de mortalité hospitalière, en fi aorto-coronarien et qu’il n’y a aucun effet pour la contrôlant par l’établissement et par des effets prothèse totale de hanche. temporels fixes. De cette façon, elle distingue l’effet d’apprentissage propre de l’effet des progrès technologiques. Ho observe que l’amélio- À partir d’une analyse du volume et les résultats de ration substantielle des résultats d’ACPT réalisée la réparation d’anévrisme de l’aorte abdominale,

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 38 -

Hannan et al. (1992) pointent que l’effet Dans une étude récente, Gowrisankaran et al. d’apprentissage, et en particulier la spécialisation (2006) tentent de déterminer, en utilisant des des chirurgiens, est un facteur important pour variables instrumentales, si le volume engendre réduire la mortalité. Ils examinent les variations une amélioration des résultats ou si, au contraire, dans les volumes d’activité des chirurgiens et les les hôpitaux qui mettent le mieux en œuvre établissements sur trois ans (1985-1987) pour certains types d’intervention attirent plus de tester le sens de la causalité avec la mortalité patients. Ils testent le rôle de l’apprentissage, de hospitalière. Ils observent que les établissements l’oubli et du renvoi sélectif dans l’analyse du lien qui avaient des volumes d’activité élevés initiale- volume-qualité pour trois procédures chirurgicales : ment attiraient plus de patients dans les années la pancréato-duodénectomie, le pontage aorto- suivantes par rapport à ceux qui avaient un taux coronarien et la réparation d’anévrisme de l’aorte de mortalité élevé, mais cet effet (renvoi sélectif) abdominale. Ils utilisent des données provenant n’était pas significatif pour expliquer la réduction de Floride, entre 1988 et 1999, et de Californie, de mortalité observée. En examinant le volume entre 1993 et 1997. En premier lieu, ils observent un par type de procédure effectué par chirurgien, les effet causal significatif du volume sur la qualité. De auteurs concluent que l’effet d’apprentissage au plus, leurs résultats montrent, pour le pontage et niveau de chirurgiens est déterminant dans cette l’anévrisme abdominal, que l’effet d’apprentissage relation. joue un rôle important dans l’explication des diffé- rences entre les résultats (ajustés sur le risque) Dans une étude atypique au sein de cette des hôpitaux. Les auteurs notent également que littérature, Tay (2003) examine l’impact de la le degré d’oubli dans les organisations diffère de qualité et de la distance sur le choix d’hôpital manière significative selon les procédures, l’oubli des patients Medicare (États-Unis) atteints d’un étant plus lent pour la pancréato-duodénectomie infarctus du myocarde en 1994. La qualité est et plus rapide pour le pontage aorto-coronarien. mesurée par plusieurs indicateurs : les taux de mortalité et de complications, mais aussi le Cette brève revue de littérature présente les niveau de ressources dédiées (nombre d’infirmiers principaux concepts nécessaires pour comprendre par lit), et la capacité de l’hôpital à réaliser deux la littérature empirique examinant le lien entre le interventions cardiologiques de haute technologie : volume de l’activité et la qualité des soins fournis la pose d’un cathéter et la revascularisation. dans les hôpitaux. Le sens de la causalité entre L’auteur observe que la demande hospitalière volume et qualité fait toujours débat dans le secteur (volume) est négativement affectée par la distance hospitalier. Toutefois, l’hypothèse selon laquelle, du patient, et positivement par la capacité techno- toutes choses égales par ailleurs, la qualité de logique et humaine de l’hôpital. Elle calcule, par soins s’améliore avec l’expérience accumulée, a exemple, que l’adoption d’un plateau technique été largement démontrée en économie industrielle. permettant la pose de cathéters augmenterait la Dans le cas des soins hospitaliers, l’importance demande de 65 %, alors que l’accroissement de de l’effet d’apprentissage semble varier, d’une 1 % du nombre d’infirmiers par lit augmenterait la procédure à l’autre et est significatif, notamment demande de seulement 24 %. Ainsi, en utilisant dans le cadre de soins complexes. La littérature une modélisation relativement sophistiquée, Tay montre par ailleurs que le volume d’activité d’un démontre que la demande hospitalière (volume) établissement détermine également son niveau accompagne la qualité. Elle souligne également d’influence dans sa zone d’implantation et a donc que l’importance de la distance dans les choix des répercussions sur l’activité des hôpitaux hospitaliers est souvent sous-estimée. Néanmoins, environnants et sur le niveau de concurrence. Les il convient de noter que dans sa modélisation, Tay dynamiques d’apprentissage (c’est-à-dire la façon considère que la qualité est une variable exogène dont l’apprentissage se développe avec l’expé- (indépendante) de la demande. Cela dit, elle rience et la façon dont il est transféré) constituent ne traite pas le problème d’endogénéité poten- des questions importantes pour les établissements tielle entre la demande et la qualité, les hôpitaux de santé et pour le planificateur qui doit anticiper pouvant décider de leur niveau de qualité en fonc- les risques d’une trop forte concentration. tion de la demande existante.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 Conclusion

- 41 -

Conclusion

Dans ce rapport, nous avons réalisé un examen les patients sont orientés vers les établissements complet de la littérature portant sur la relation de qualité, n’est pas complètement réfutée pour entre le volume d’activité dans les établissements certaines procédures hospitalières. À cet égard, de santé et les résultats des soins au cours des il ne faut pas négliger les effets pervers possibles dix dernières années. Nous avons identifié les sur la qualité des soins d’une trop forte concentra- principales interventions chirurgicales, ainsi que tion de l’activité. d’autres types de soins fournis à l’hôpital et dans d’autres contextes de soins, dont les résultats ont In fine, il est important de relever les critères qui été étudiés en fonction du volume d’activité. Au tempèrent cette relation. En effet, le lien volume- total, nous avons examiné 175 articles. La grande qualité est sensible : majorité des études (84 %) portait sur des inter- ventions chirurgicales, mais on repère, de plus en – à la nature des procédures et interventions plus, des études portant sur une grande variété analysées, et particulièrement à leur niveau de de soins, y compris les soins mentaux et ceux de technicité. Plus l’intervention est spécifique et ville. complexe, plus la corrélation entre volume et qualité est affirmée ; Nous avons présenté les principaux résultats et enseignements des études retenues, en décrivant – aux seuils d’activité retenus dans les études. leurs principales caractéristiques méthodologi- En l’occurrence, pour la plupart des interven- ques. Rappelons ici que les méthodes statistiques tions, il n’existe pas de seuil d’activité una- utilisées se sont nettement améliorées et diversifiées nimement accepté. De plus, certaines étu- ces dernières années. La littérature sur le lien des montrent que la relation volume-qualité entre le volume et la qualité des résultats cliniques devient marginale au-delà d’un seuil relative- des établissements de soins suggère clairement ment bas ; que, pour certaines procédures et interventions, on peut sensiblement améliorer la qualité lorsque – à des évolutions temporelles. L’intensité des le volume d’activité augmente. Ceci semble relations entre volume et qualité peut évoluer être particulièrement valable pour la plupart de au cours du temps : l’influence du volume sur interventions cardio-vasculaires et pour la chirurgie la qualité peut diminuer, voire disparaître au complexe, en particulier dans le traitement des cours du temps pour certaines procédures, cancers. notamment à mesure que la maîtrise de cette procédure s’étend et que celle-ci se banalise. Bien que le sens de la causalité et les mécanismes sous-jacents de cette relation fassent toujours En tout état de cause, bien que le volume d’acti- débat, il semble que l’effet d’apprentissage au vité semble être un critère permettant d’évaluer la niveau individuel (chirurgien) et au niveau de qualité des soins, notamment chirurgicaux, il n’est l’organisation (transfert de connaissance) explique pas le seul à prendre en considération pour mesu- une grande partie de cette corrélation. Toutefois, rer et améliorer la qualité dans les établissements l’hypothèse du « renvoi sélectif », selon laquelle de santé.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008

Bibliographie

- 45 -

Bibliographie

Bibliographie de la revue systématique G.E. (2005). Hospital volume-outcome re- (175 articles) lationships among medical admissions to ICUs. Chest, 128 (3) : 1682-1689.

[1] Blatier J.F.F., Fra J. (2001). Relation entre [10] Ellis S.G., Weintraub W., Holmes D., Shaw volume d’activité et mortalité intra-hospitalière R., Block P.C., King S.B., III (1997). Relation après intervention pour pose de prothèse of operator volume and experience to totale de hanche dans les établissements procedural outcome of percutaneous publics de la région Rhône-Alpes. Revue coronary revascularization at hospitals with d’Epidemiologie et de Sante Publique, 49 high interventional volumes. Circulation, 95 (2) : 173-182. (11) : 2479-2484.

[2] Hartz A.J., Kuhn E.M., Pulido J. (1999). Pres- [11] Gilligan M.A., Neuner J., Zhang X., tige of training programs and experience Sparapani R., Laud P.W., Nattinger A.B. of bypass surgeons as factors in adjusted (2007). Relationship between number of patient mortality rates. Med Care, 37 (1) : breast cancer operations performed and 93-103. 5-year survival after treatment for early- stage breast cancer. Am J Public Health, [3] Hamilton B.H., Ho V. (1998). Does practice 97 (3) : 539-544. make perfect? Examining the relationship between hospital surgical volume and [12] Gillison E.W., Powell J., McConkey C.C., outcomes for hip fracture patients in Spychal R.T. (2002). Surgical workload and Quebec. Med Care, 36 (6) : 892-903. outcome after resection for carcinoma of the oesophagus and cardia. Br.J Surg, 89 [4] Holmboe E.S., Wang Y., Tate J.P., Meehan (3) : 344-348. T.P. (2006). The effects of patient volume on the quality of diabetic care for Medicare [13] Glance L.G., Dick A.W., Osler T.M., beneficiaries. Med Care, 44 (12) : 1073-1077. Mukamel D.B. (2005). The relation between surgeon volume and outcome following off- [5] Hamilton B.H., Hamilton V.H. (1997). pump vs on-pump coronary artery bypass Estimating surgical volume--outcome graft surgery. Chest, 128 (2) : 829-837. relationships applying survival models : accounting for frailty and hospital fixed [14] Habib M.S., Bunce C.V., Fraser S.G. (2005). effects. Health Econ., 6 (4) : 383-395. The role of case mix in the relation of volume and outcome in phacoemulsification. [6] Tsai A.C., Votruba M., Bridges J.F., Cebul Br.J Ophthalmol., 89 (9) : 1143-1146. R.D. (2006). Overcoming bias in estimating the volume-outcome relationship. Health [15] Habib M., Mandal K., Bunce C.V., Fraser Serv.Res, 41 (1) : 252-264. S.G. (2004). The relation of volume with outcome in pharmaco-emulsification sur- [7] Chen E.W., Canto J.G., Parsons L.S., gery. Br.J Ophthalmol., 88 (5) : 643-646. Peterson E.D., Littrell K.A., Every N.R., Gibson C.M., Hochman J.S., Ohman E.M., [16] Hamilton S.M., Johnston W.C., Voaklander Cheeks M., Barron H.V. (2003). Rela- D.C. (2001). Outcomes after the regionali- tion between hospital intra-aortic balloon zation of major surgical procedures in the counterpulsation volume and mortality in Alberta Capital Health Region (Edmonton). acute myocardial infarction complicated by Can.J Surg, 44 (1) : 51-58. cardiogenic shock. Circulation, 108 (8) : 951-957. [17] Hannan E.L., Wu C., Ryan T.J., Bennett E., Culliford A.T., Gold J.P., Hartman A., Isom [8] Druss B.G., Miller C.L., Pincus H.A., Shih O.W., Jones R.H., McNeil B., Rose E.A., S. (2004). The volume-quality relationship Subramanian V.A. (2003). Do hospitals and of mental health care : does practice make surgeons with higher coronary artery by- perfect? Am J Psychiatry, 161 (12) : 2282- pass graft surgery volumes still have lower 2286. risk-adjusted mortality rates? Circulation, 108 (7) : 795-801. [9] Durairaj L., Torner J.C., Chrischilles E.A., Vaughan Sarrazin M.S., Yankey J., Rosenthal

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 46 -

[18] Hannan E.L., Wu C., Walford G., King S.B., [27] Nomura E., Tsukuma H., Ajiki W., Oshima A. III, Holmes D.R., Jr., Ambrose J.A., Sharma (2003). Population-based study of relation- S., Katz S., Clark L.T., Jones R.H. (2005). ship between hospital surgical volume and Volume-outcome relationships for percuta- 5-year survival of stomach cancer patients neous coronary interventions in the stent in Osaka, Japan. Cancer Sci., 94 (11) : era. Circulation, 112 (8) : 1171-1179. 998-1002.

[19] Heller G., Richardson D.K., Schnell R., [28] Panageas K.S., Schrag D., Riedel E., Misselwitz B., Kunzel W., Schmidt S. (2002). Bach P.B., Begg C.B. (2003). The effect of Are we regionalized enough? Early-neonatal clustering of outcomes on the association of deaths in low-risk births by the size of procedure volume and surgical outcomes. delivery units in Hesse, Germany 1990- Ann.Intern.Med, 139 (8) : 658-665. 1999. Int.J Epidemiol, 31 (5) : 1061-1068. [29] Schelbert E.B., Vaughan-Sarrazin M.S., [20] Ho V. (2000). Evolution of the volume- Welke K.F., Rosenthal G.E. (2005). Hospital outcome relation for hospitals performing volume and selection of valve type in older coronary angioplasty. Circulation, 101 (15) : patients undergoing aortic valve replacement 1806-1811. surgery in the United States. Circulation, 111 (17) : 2178-2182. [21] Jollis J.G., Peterson E.D., Nelson C.L., Stafford J.A., DeLong E.R., Muhlbaier L.H., [30] Simunovic M., Rempel E., Theriault M.E., Mark D.B. (1997). Relationship between Coates A., Whelan T., Holowaty E., Langer physician and hospital coronary angioplasty B., Levine M. (2006). Influence of hospi- volume and outcome in elderly patients. tal characteristics on operative death and Circulation, 95 (11) : 2485-2491. survival of patients after major cancer surgery in Ontario. Can.J Surg, 49 (4) : [22] Kotwall C.A., Maxwell J.G., Brinker C.C., 251-258. Koch G.G., Covington D.L. (2002). National estimates of mortality rates for radical [31] Sung V.W., Rogers M.L., Myers D.L., Clark pancreaticoduodenectomy in 25,000 M.A. (2006). Impact of hospital and surgeon patients. Ann.Surg Oncol., 9 (9) : 847-854. volumes on outcomes following pelvic reconstructive surgery in the United States. [23] Metreveli R.E., Sahm K., Denstman F., Am J Obstet.Gynecol., 195 (6) : 1778-1783. Abdel-Misih R., Petrelli N.J. (2005). Hepatic resection at a major community-based [32] Urbach D.R., Baxter N.N. (2004). Does it teaching hospital can result in good matter what a hospital is «high volume» outcome. Ann.Surg Oncol., 12 (2) : 133-137. for? Specificity of hospital volume-outcome associations for surgical procedures : [24] Meyerhardt J.A., Catalano P.J., Schrag analysis of administrative data. BMJ, 328 D., Ayanian J.Z., Haller D.G., Mayer R.J., (7442) : 737-740. MacDonald J.S., Benson A.B., III, Fuchs C.S. (2003). Association of hospital procedure [33] Vakili B.A., Kaplan R., Brown D.L. (2001). volume and outcomes in patients with co- Volume-outcome relation for physicians lon cancer at high risk for recurrence. Ann. and hospitals performing angioplasty for Intern.Med, 139 (8) : 649-657. acute myocardial infarction in New York state. Circulation, 104 (18) : 2171-2176. [25] McCabe J.E., Jibawi A., Javle P. (2005). Defining the minimum hospital case-load [34] Ward M.M. (2005). Association between to achieve optimum outcomes in radical physician volume and in-hospital mortality in cystectomy. BJU Int., 96 (6) : 806-810. patients with systemic lupus erythematosus. Arthritis Rheum., 52 (6) : 1646-1654. [26] Nallamothu B.K., Wang Y., Magid D.J., McNamara R.L., Herrin J., Bradley E.H., [35] Wu C., Hannan E.L., Ryan T.J., Bennett Bates E.R., Pollack C.V., Jr., Krumholz H.M. E., Culliford A.T., Gold J.P., Isom O.W., (2006). Relation between hospital specia- Jones R.H., McNeil B., Rose E.A., lization with primary percutaneous coro- Subramanian V.A. (2004). Is the impact of nary intervention and clinical outcomes in hospital and surgeon volumes on the ST-segment elevation myocardial infarction : in-hospital mortality rate for coronary artery National Registry of Myocardial Infarction- bypass graft surgery limited to patients at 4 analysis. Circulation, 113 (2) : 222-229. high risk? Circulation, 110 (7) : 784-789.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 47 -

[36] Rill V., Brown D.L. (1999). Practice of [43] Al-Khatib S.M., Lucas F.L., Jollis J.G., coronary angioplasty in California in 1995 : Malenka D.J., Wennberg D.E. (2005). The comparison to 1989 and impact of coronary relation between patients’ outcomes and stenting. Circulation, 99 (21) : e12. the volume of cardioverter-defibrillator im- plantation procedures performed by physi- [37] Solomon D.H., Losina E., Baron J.A., Fossel cians treating Medicare beneficiaries. J Am A.H., Guadagnoli E., Lingard E.A., Miner A., Coll. Cardiol., 46 (8) : 1536-1540. Phillips C.B., Katz J.N. (2002). Contribution of hospital characteristics to the volume-out- [44] Moscucci M., Share D., Smith D., O’Don- come relationship: dislocation and infection nell M.J., Riba A., McNamara R., Lalonde following total hip replacement surgery. T., Defranco A.C., Patel K., Kline R.E., Arthritis Rheum., 46 (9) : 2436-2444. D’Haem C., Karve M., Eagle K.A. (2005). Relationship between operator volume and [38] Simunovic M., To T., Theriault M., Langer B. adverse outcome in contemporary percu- (1999). Relation between hospital surgical taneous coronary intervention practice : an volume and outcome for pancreatic resection analysis of a quality-controlled multicenter for neoplasm in a publicly funded health percutaneous coronary intervention clinical care system. CMAJ., 160 (5) : 643-648. database. J Am Coll Cardiol., 46 (4) : 625-632.

[39] Spaulding C., Morice M.C., Lancelin B., El [45] Nobilio L., Fortuna D., Vizioli M., Berti E., H.S., Lepage E., Bataille S., Tresca J.P., Guastaroba P., Taroni F., Grilli R. (2004). Mouranche X., Fosse S., Monchi M., de Impact of regionalisation of cardiac surgery V.N. (2006). Is the volume-outcome relation in Emilia-Romagna, Italy. J. Epidemiol still an issue in the era of PCI with Community Health, 58 (2) : 97-102. systematic stenting? Results of the greater Paris area PCI registry. Eur Heart J, 27 (9) : [46] Barker F.G., Klibanski A., Swearingen 1054-1060. B. (2003). Transsphenoidal surgery for pituitary tumors in the United States, 1996- [40] Wu S.C., Chen C.C., Ng Y.Y., Chu H.F. 2000 : mortality, morbidity, and the effects of (2006). The relationship between surgical hospital and surgeon volume. J. Clin. Endo- site infection and volume of coronary artery crinol. Metab, 88 (10) : 4709-4719. bypass graft surgeries : Taiwan experien- ce. Infect.Control Hosp.Epidemiol, 27 (3) : [47] Hollenbeck B.K., Dunn R.L., Miller D.C., 308-311. Daignault S., Taub D.A., Wei J.T. (2007). Volume-based referral for cancer surgery : [41] McGrath P.D., Wennberg D.E., Malenka informing the debate. J Clin.Oncol., 25 (1) : D.J., Kellett M.A., Jr., Ryan T.J., Jr., O’Meara 91-96. J.R., Bradley W.A., Hearne M.J., Hettleman B., Robb J.F., Shubrooks S., VerLee P., [48] Meyerhardt J.A., Tepper J.E., Niedzwiecki Watkins M.W., Lucas F.L., O’Connor G.T. D., Hollis D.R., Schrag D., Ayanian J.Z., (1998). Operator volume and outcomes in O’Connell M.J., Weeks J.C., Mayer R.J., 12,998 percutaneous coronary interven- Willett C.G., MacDonald J.S., Benson A.B., tions. Northern New England Cardiovascular III, Fuchs C.S. (2004). Impact of hospital Disease Study Group. J Am Coll Cardiol., procedure volume on surgical operation 31 (3) : 570-576. and long-term outcomes in high-risk cura- tively resected rectal cancer : findings from [42] Malenka D.J., McGrath P.D., Wenn- the Intergroup 0114 Study. J Clin.Oncol., berg D.E., Ryan T.J., Jr., Kellett M.A., Jr., 22 (1) : 166-174. Shubrooks S.J., Jr., Bradley W.A., Hettlemen B.D., Robb J.F., Hearne M.J., Silver T.M., [49] Yao S.L., Lu-Yao G. (1999). Population- Watkins M.W., O’Meara J.R., VerLee P.N., based study of relationships between O’Rourke D.J. (1999). The relationship hospital volume of prostatectomies, patient between operator volume and outcomes outcomes, and length of hospital stay. J. after percutaneous coronary interventions Natl.Cancer Inst., 91 (22) : 1950-1956. in high volume hospitals in 1994-1996 : the northern New England experience. Northern [50] Hodgson D.C., Zhang W., Zaslavsky New England Cardiovascular Disease Study A.M., Fuchs C.S., Wright W.E., Ayanian Group. J Am Coll Cardiol., 34 (5) : 1471- J.Z. (2003). Relation of hospital volume to 1480. colostomy rates and survival for patients with rectal cancer. J Natl.Cancer Inst., 95 (10) : 708-716.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 48 -

[51] Perler B.A., Dardik A., Burleyson G.P., [60] Birkmeyer J.D., Siewers A.E., Finlayson Gordon T.A., Williams G.M., comment S., E.V.A., Stukel T.A., BATISTAI., Welch H.G., Perler B.A., Debord J.R., Nath R.L. (1998). Wennberg D.E. (2002). Hospital volume Influence of age and hospital volume on and surgical mortality in the United States. the results of carotid endarterectomy: A The New England Journal of Medicine, 346 statewide analysis of 9918 cases. (15) : 1128-1137. Discussion.Journal of vascular surgery, 27 (1) : 25-33. [61] Canto J.G., Every N.R., Magid D.J., Rogers W.J., Malmgren J.A., Frederick [52] Dimick J.B., Cowan J.A., Jr., Henke P.K., P.D., French W.J., Tiefenbrunn A.J., Mis- Wainess R.M., Posner S., Stanley J.C., ra V.K., Kiefe C.I., Barron H.V. (2000). The Upchurch G.R., Jr. (2003). Hospital volume- volume of primary angioplasty procedures related differences in aorto-bifemoral by- and survival after acute myocardial infarction. pass operative mortality in the United National Registry of Myocardial Infarction 2 States. J Vasc.Surg, 37 (5) : 970-975. Investigators. N Engl J Med, 342 (21) : 1573- 1580. [53] Dueck A.D., Kucey D.S., Johnston K.W., Alter D., Laupacis A. (2004). Survival after [62] Carey J.S., Danielsen B., Gold J.P., Rossiter ruptured abdominal aortic aneurysm : effect S.J. (2005). Procedure rates and outcomes of patient, surgeon, and hospital factors. J of coronary revascularization procedures in Vasc.Surg, 39 (6) : 1253-1260. California and New York. J Thorac.Cardio- vasc.Surg, 129 (6) : 1276-1282. [54] Kahn J.M., Goss C.H., Heagerty P.J., Kramer A.A., O’Brien C.R., Rubenfeld G.D. [63] Vaziri K., Roland J.C., Robinson L., Fakhry (2006). Hospital volume and the outcomes S.M. (2007). Optimizing physician staffing of mechanical ventilation. N Engl J Med, and resource allocation : sine-wave variation 355 (1) : 41-50. in hourly trauma admission volume. J. Trauma, 62 (3) : 610-614. [55] Roddy S.P., O’donnell T.F., Jr., Wilson A.L., Estes J.M., Mackey W.C. (2000). [64] Glance L.G., Osler T.M., Dick A., Mukamel The Balanced Budget Act : Potential D. (2004). The relation between trauma implications for the practice of vascular center outcome and volume in the Natio- surgery. Discussion. Journal of vascular nal Trauma Databank. J Trauma, 56 (3) : surgery, 31 (2) : 227-236. 682-690.

[56] Thiemann D.R., Coresh J., Oetgen W.J., [65] Rabeneck L., Davila J.A., Thompson M., Powe N.R. (1999). The association between El-Serag H.B. (2004). Surgical volume and hospital volume and survival after acute long-term survival following surgery for myocardial infarction in elderly patients. colorectal cancer in the Veterans Affairs N Engl J Med, 340 (21) : 1640-1648. Health-Care System. Am J Gastroenterol., 99 (4) : 668-675. [57] Tucker J. (2002). Patient volume, staffing, and workload in relation to risk-adjusted [66] Kimmel S.E., Sauer W.H., Brensinger outcomes in a random stratified sample C., Hirshfeld J., Haber H.L., Localio A.R. of UK neonatal intensive care units : A (2002). Relationship between coronary prospective evaluation. Lancet, 359 (9301): angioplasty laboratory volume and outcomes 99-107. after hospital discharge. Am Heart J, 143 (5) : 833-840. [58] Begg C.B., Cramer L.D., Hoskins W.J., Brennan M.F. (1998). Impact of hospital [67] Manheim L.M., Sohn M.W., Feinglass J., volume on operative mortality for major Ujiki M., Parker M.A., Pearce W.H. (1998). cancer surgery. JAMA, 280 (20) : 1747-1751. Hospital vascular surgery volume and procedure mortality rates in California, [59] Birkmeyer, J.D., Samuel, R.G., Finlayson, 1982-1994. J Vasc.Surg, 28 (1) : 45-56. M.P.H., Tosteson, A.N.A., Sharp, S.M., Warshaw, A.L.,Fisher, E.S. (2000). Major [68] Magid D.J., Calonge B.N., Rumsfeld J.S., HPB Procedures Must be Undertaken in Canto J.G., Frederick P.D., Every N.R., High Volume Quaternary Centres? HPB Barron H.V. (2000). Relation between Surgery, 11 : 359-361. hospital primary angioplasty volume and mortality for patients with acute MI treated

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 49 -

with primary angioplasty vs thrombolytic [76] Simunovic M., To T., Baxter N., Balshem A., therapy. JAMA, 284 (24) : 3131-3138. Ross E., Cohen Z., McLeod R., Engstrom P., Sigurdson E. (2000). Hospital procedure [69] Mustafa M.U., Cohen M., Zapotulko K., volume and teaching status do not influence Feinberg M., Miller M.F., Aueron F., Wasty treatment and outcome measures of rectal N., Tanwir A., Rogal G. (2005). The lack cancer surgery in a large general population. of a simple relation between physician’s J. Gastrointest.Surg, 4 (3) : 324-330. percutaneous coronary intervention volume and outcomes in the era of coronary [77] Ahn H.S., Yoon S.J., Jo H.Y., Lee H.Y., stenting : A two-centre experience. Inter- Lee J., Seo H.J. (2006). Association national Journal of Clinical Practice {INT. J between unplanned readmission rate and CLIN. PRACT.}, 59 (12) : 1401-1407. volume of breast cancer operation cases. Int. J. Clin. Pract., 60 (1) : 32-35. [70] Van Domburg R.T., Takkenberg J.J., van Herwerden L.A., Venema A.C., Bogers A.J. [78] Díaz-Montes T.P., Zahurak M.L., Giuntoli II, (2002). Short-term and 5-year outcome Gardner G.J., Bristow R.E. (2006), Uterine after primary isolated coronary artery bypass cancer in Maryland : Impact of surgeon graft surgery : results of risk stratification case volume and other prognostic factors in a bilocation center. Eur J Cardiothorac. on short-term mortality. Gynecol Oncol, Surg, 21 (4) : 733-740. 103 : 1043-1047.

[71] Glance L.G., Li Y., Osler T.M., Dick A., [79] Elit L., Bondy S.J., Paszat L., Przybysz R., Mukamel D.B. (2006). Impact of patient Levine M. (2002). Outcomes in surgery for volume on the mortality rate of adult intensive ovarian cancer. Gynecol.Oncol., 87 (3) : care unit patients. Crit Care Med, 34 (7) : 260-267. 1925-1934. [80] Begg C.B., Riedel E.R., Bach P.B., Kattan [72] Tu J.V., Austin P.C., Chan B.T. (2001). M.W., Schrag D., Warren J.L., Scardino P.T. Relationship between annual volume of (2002). Variations in morbidity after radical patients treated by admitting physician and prostatectomy. N Engl J Med, 346 (15) : mortality after acute myocardial infarction. 1138-1144. JAMA, 285 (24) : 3116-3122. [81] Klein L.W., Schaer G.L., Calvin J.E., [73] McGrath P.D., Wennberg D.E., Dickens Palvas B., Allen J., Loew J., Uretz E., Parrillo J.D., Jr., Siewers A.E., Lucas F.L., Malenka J.E. (1997), Does low individual operator D.J., Kellett M.A., Jr., Ryan T.J., Jr. (2000). coronary interventional procedural volume Relation between operator and hospital correlate with worse institutional procedural volume and outcomes following percuta- outcome? J Am Coll Cardiol, 30 : 870-877. neous coronary interventions in the era of the coronary stent. JAMA, 284 (24) : 3139- [82] Nathens AB., Jurkovich GJ., Maier RV., 3144. Grossman DC., MacKenzie EJ., Moore M., Rivara FP. (2001), Relationship between [74] Maynard C., Every N.R., Chapko M.K., trauma center volume and outcomes ; Jama ; Ritchie J.L. (2001). Improved outcomes 285(9):1164-71. associated with stenting in the health- care cost and utilization project. J Interv. [83] Katz J.N., Barrett J., Mahomed N.N., Baron Cardiol., 14 (2) : 159-163. J.A., Wright R.J., Losina E. (2004),Association between hospital and surgeon procedure [75] Makowiec F., Post S., Saeger H.D., volume and the outcomes of total knee Senninger N., Becker H., Betzler M., Buhr replacement. J Bone Joint Surg Am, H.J., Hopt U.T. (2005). Current practice 86-A (9) : 1909-16. patterns in pancreatic surgery : results of a multi-institutional analysis of seven large [84] Cooper, A., Hannan E.L., Bessey P.Q. surgical departments in Germany with 1454 Farrell Szypulski L., Cayten C.G., Mottley pancreatic head resections, 1999 to 2004 L. (2000), An Examination of the Volume- (German Advanced Surgical Treatment Mortality Relationship for New York State study group). J Gastrointest.Surg, 9 (8) : Trauma Centers. Journal of Trauma-Injury 1080-1086. Infection & Critical Care, 48 (1) : 16.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 50 -

[85] Sollano J.A., Gelijns A.C., Moskowitz A.J., operation. Journal The American College of Heitjan D.F., Cullinane S., Chen J.M., Surgeons. Vol. 204, 383-391. Roohan P., Reemtsma K., Shields E.P. (1999), Volume-Outcome Relationships [94] Weller W.E., Hannan E.L. (2006), In Cardiovascular Operations : New York Relationship between provider volume and State, 1990-1995 ; J Thorac Cardiovasc postoperative complications for bariatric Surg.; 117: 419-430. procedures in New York State. Journal The American College of Surgeons. Vol. 202, [86] Jain N., Pietrobon R., Hocker S., Guller 753-761. U., Shankar A., Higgins L.D. (2004), The Relationship Between Surgeon and [95] Choti M.A., Bowman H.M., Pitt H.A., Sosa Hospital Volume and Outcomes for Shoulder J.A., Sitzmann J.V., Cameron J.L., Gordon Arthroplasty. The Journal of Bone and Joint T.A. (1998). Should hepatic resections be Surgery, 86 : 496-505. performed at high-volume referral centers? J. Gastrointest Surg, 2 : 11-20. [87] Freeman J., Nicholl J., Turner J., (2006), Does size matter? The relationship between [96] Thompson R., Kane R.L., Gromala T., volume and outcome in the care of major McLaughlin B., Flood S., Morris N. and Bor- trauma. J Health Serv Res Policy, 11 (2) : bas C. (2002), Complications and short-term 101-105. outcomes associated with total hip arthro- plasty in teaching and community hospitals. [88] J.A. Battistella F.D. (2003), Is The- The Journal of Arthroplasty. Vol. 17, N° 1. re a Relationship between Trauma Center Volume and Mortality? Journal of Trauma- [97] Bartels D.B., Wypij D., Wenzlaff P., Injury Infection & Critical Care, 54 (1) :16-25. Dammann O., Poets C.F. (2006), Hospital volume and neonatal mortality among very [89] Katz J.N., Losina E., Barrett J., Phillips C.B., low birth weight infants. Pediatrics, Vol. 117, Mahomed N.N., Lew R.A., Guadagnoli E., N°6, 2206-2214. Harris W.H., Poss R., Baron J.A. (2001), Association Between Hospital and Surgeon [98] Cox C.E., Salud C.J., Cantor A., Bass Procedure Volume and Outcomes of Total S.S., Peltz E.S., Ebert M.D., Nguyen K., Hip Replacement in the United States Reintgen D.S. (2001), Learning curves for Medicare Population. The Journal of Bone breast cancer sentinel lymph node mapping and Joint Surgery, 83 : 1622-1629. based on surgical volume analysis. J Am Coll Surg, 193 : 593-600. [90] Jain N.B., Pietrobon R., Guller U., Ahluwalia A.S., Higgins L.D. (2005), Influence of [99 ] Liu J.H., Etzioni D.A., O’Connell J.B., provider volume on length of stay, operating Maggard M.A., Ko C.K. (2003), Using volume room time, and discharge status for rotator criteria : do California hospitals measure up? cuff repair. J Shoulder Elbow Surg. Journal of Surgical Research 113, 96-101. 14(4):407-13. [100] Tepas J.J., Patel J.C., DiScala C., Wears [91] Judge A., Chard J., Learmonth I., Dieppe P. R.L. and Veldenz H.C. (1998), Relationship (2006), The effects of surgical volumes and of trauma patient volume to outcome expe- training centre status on outcomes following rience : can a relationship be defined? The total joint replacement : analysis of the Journal of Trauma : Injury, infection, and Hospital Episode Statistics for England. Critical Care. Vol. 44, N°5. Journal of Public Health. Vol. 28, N°2, pp116-124. [101] Richardson J.D., Schmieg E., Boaz Ph., Spain D.A., Wohltmann Ch., Wilson M.A., [92] Hervey S.L., Purves H.R., Guller U., Toth Carrillo E.H., Miller F.B. and Fulton R.L. A.P., Vail T.P., Pietrobon R. (2003), Provider (1998), Impact of trauma attending surgeon Volume of Total Knee Arthroplasties and case volume on outcome : is more better? Patient Outcomes in the HCUP-Nationwide The journal of Trauma : Injury. Infection and Inpatient Sample. Journal of Bone and Joint Critical Care. Vol. 44 N°2. Surgery, 85 (9) : 1775-1783. [102] Bell C.M., Hatch W.V., Cernat G., Urbach [93] Weller W.E., Rosati C., Hannan E.L. (2007), D.R. (2007), Surgeon volumes and selected Relationship between surgeon and hospital patient outcomes in cataract surgery. The volume and readmission after bariatric American Academy of Ophthalmology. Vol. 114,405-410.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 51 -

[103] Hammond J.W., Queale W.S., Kyun Kim T., [112] Needham D. M., Bronskill, S. E., Rothwell, McFarland E.G. (2003), Surgeon Experience D. M., Sibbald, W. J., Pronovost, P. J., and Clinical and Economic Outcomes for Laupacis A.., Stukel, Th.A. (2006), Hospital Shoulder Arthroplasty. The Journal of Bone volume and mortality for mechanical and Joint Surgery, 85 : 2318-2324. ventilation of medical and surgical pa- tients: A population-based analysis using [104] Urschel J.D., Urschel D.M. (2000), The administrative data *; Critical Care Medicine. hospital volume-outcome relationship in 34(9):2349-2354. general thoracic surgery. Is the surgeon the critical determinant? The Journal of Cardio- [113] Callahan M.A., Christos P.J., Gold H.T., vascular Surgery. Vol. 41 N° 1. Mushlin A.I., and Daly J.M. (2003). Influence of Surgical Subspecialty Training on In- [105] Zyzanski S.J., Stange K.C., Langa D., Hospital Mortality for Gastrectomy and Flocke S.A. (1998), Trade-offs in high- Colectomy Patients. Ann Surg, 238 (4) : volume primary care practice. The Journal 629–639. of Family Practice. Vol. 46, N°5. [114] Ho V., Heslin M.J. (2003), Effect of hospi- [106] De Witte T., Tirschwell D.L., Clark M.A., tal volume and experience on in-hospital Tuden D., Derdeyn C.P., Moran Ch.J., mortality for pancreaticoduodenectomy. Dacey R.G. (2003) ; Mortality rates after Ann Surg, 237 (4) : 509-14. subarachnoid hemorrhage : variations according to hospital case volume in 18 [115] Nguyen N.T., Paya M., Stevens C. M., states. J. Neurosurg./volume 99 : 810-817. Mavandadi Sh., Zainabadi K., and Wilson S.E., (2004), The Relationship Between [107] Lee K-T., Chang W-T., Huang M-C., Chiu Hospital Volume and Outcome in Bariatric H-C. (2004), Influence of surgeon volu- Surgery at Academic Medical Centers ; me on clinical and economic outcomes of Annals Surgery. 240(4): 586–594. laparoscopic cholecystectomy. Digestive Surgery. Vol. 21 : 406-412. [116] Sosa J. A., Bowman H.M., Tielsch J.M., Powe N.R., Gordon T.A. and Udelsman [108] Marcin J.P., Romano P.S. (2004), Impact R. (1998), The importance of surgeon of between-hospital volume and within- experience for clinical and economic hospital volume on mortality and readmission outcomes from thyroidectomy. Annals rates for trauma patients in California. Crit Surgery ; 228(3): 320–330. Care Med. Vol. 32, N°7. [117] de Gara C., Hanson J., Hamilton S. (2003), [109] SooHoo N.F., Zingmond D.S., Lieberman A population-based study of tumor-node J.R., Ko C.Y. (2006), Primary total knee relationship, resection margins, and surgeon arthroplasty in California 1991 to 2001. volume on gastric cancer survival. The Does hospital volume affect outcomes? American Journal of Surgery, 186 (1) : 23-27. The journal of Arthroplasty. Vol. 21, N°2. [118] Dimick J.B., Stanley J.C., Axelrod D.A., [110] Gordon T.A., Bowman H.M., Bass E.B., Kazmers A., Henke P.K., Jacobs L.A. Wake- Lillemoe K.D., Yeo C.J., Heitmiller R.F., field T.W., Greenfield L.J., Upchurch G.R., Choti M.A., Burleyson G.P., Hsieh G., (2002), Variation in Death Rate After Abdo- Cameron J.L. (1999), Complex gastrointes- minal Aortic Aneurysmectomy in the United tinal surgery : impact of provider experience States : Impact of Hospital Volume, Gender, on clinical and economic outcomes. J. Am and Age. Ann Surg, 235 (4) : 579–585. Coll Surg, 189 (1) : 46-56 . [119] Ho V., Heslin M.J., Yun H., Howard L., [111] Freixinet J.L., Julià-Serdà G., Rodríguez (2006), Trends in hospital and surgeon P.M., Santana N.B., Rodríguez de Castro volume and operative mortality for cancer F., Fiuza M.D., López-Encuentra A. Bron- surgery. Ann Surg Oncol, 13 (6) : 851-8. chogenic Carcinoma Cooperative Group of the Spanish Society of Pneumology and [120] Tu J.V., Austin P.C. and Johnston K.W. Thoracic Surgery. (2005), Hospital volume : (2001), The influence of surgical specialty trai- operative morbidity, mortality and survival ning on the outcomes of elective abdominal in thoracotomy for lung cancer. A Spanish aortic aneurysm surgery ; Journal of Vasculary multicenter study of 2994 cases. Eur J. Surgery. Cardiothorac Surg, 29 : 20-25.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 52 -

[121] Rogers SO Jr, Wolf RE, Zaslavsky AM, surgery of the ascending aorta/arch : is Wright WE, Ayanian JZ. (2006), Relation of there a relationship with caseload? Eur J surgeon and hospital volume to processes Cardiothorac Surg.; 25 : 676-682. and outcomes of colorectal cancer surgery ; Annals of Surgery. 244(6):1003-11. [131] Harjai K., Berman A., Grines C., Kahn J., Marsalese D., Mehta R., Schreiber T., [122] Brown P.P., Mack M.J., Simon A.W., Boura J., O’Neill W. (2004), Impact of Battaglia S.L., Tarkington L.G., Culler S.D., interventionalist volume, experience, and Becker E.R. (2001). Comparing clinical board certification on coronary angioplasty outcomes in high-volume and low- outcomes in the era of stenting. The American volume off-pump coronary bypass opera- Journal of Cardiology, 94 (4) : 421-426. tion programs. Ann Thorac Surg, 72 : S1009-S1015. [132] Vakili B.A., Brown D.L. (2003), Relation of total annual coronary angioplasty volume [123] Nomura E., Tsukuma H., Ajiki W., Ishikawa of physicians and hospitals on outcomes O. and Oshima A. (2006), Population-based of primary angioplasty for acute myocardial study of the relationship between hospital infarction (Data from the 1995 coronary surgical volume and 10-year survival of angioplasty reporting system of the New breast cancer patients in Osaka, Japan ; York State Department of Health); The Cancer Science Volume 97 Issue 7 Page American Journal of Cardiology; Vol.91. 618-622. [133] Carey J.S., Robertson J.M., Misbach G.A., [124] Kuo E.Y., Chang Y., Wright C.D. (2001), Fisher A.L. (2003), Relationship of hospital Impact of hospital volume on clinical and volume to outcome in cardiac surgery economic outcomes for esophagectomy. programs in California. Am Surg, 69 : 63-8. Ann Thorac Surg, 72 (4) : 1118-24. [134] Gilchrist I.C., Gardner L.H., Muhlestein [125] Dimick J.B., Pronovost P.J., Cowan J.A., J.B., Arnold A.M., Lincoff A.M., Califf R.M., Lipsett P.A. (2003), Surgical volume and Tcheng J.E., Topol E.J. (1999), Effect of quality of care for esophageal resection : do institutional volume and academic status high-volume hospitals have fewer compli- on outcomes of coronary interventions : The cations? Ann Thorac Surg, 75 : 337-341. IMPACT-II experience. Am Heart J, 138 : 976-982. [126] Ui S., Chino M., Isshiki T. (2005), Rates of primary percutaneous coronary intervention [135] Pearce W.H., Parker M. A., Feinglass J., Worldwide ; Cir.J; 69:95-100. Ujiki M., Manheim L. (1999), The impor- tance of surgeon volume and training in [127] Tsuchihashi M., Tsutsui H., Tada H., Shihara outcomes for vascular surgical procedures ; M., Takeshita A., Kono S. (2004), Volume- Journal of vascular surgery. Outcome Relation for Hospitals Performing Angioplasty for Acute Myocardial Infarction- [136] Dardik A., Burleyson G.P., Bowman H., Results From the Nationwide Japanese Gordon T.A., Williams G.M., Webb T.H., Registry- ; Circulation Journal; Vol. 68. Perler B.A. (1998), Surgical repair of ruptured 887-891. abdominal aortic aneurysms in the state of Maryland : Factors influencing outcome [128] Wu S.C., Chien L-N., Ng Y-Y., Chu H-F., among 527 recent cases. J Vasc Surg, 28 : Chen C-.C. (2005), Association of case 413-21. volume with mortality of chinese patients after coronary artery bypass grafting. Taiwan [137] Ho V. (2004), Certificate of need, volume, experience. Circulation journal. Vol. 69. and percutaneous transluminal coronary angioplasty outcomes. American Heart [129] Lyman S, Jones E C, Bach P B, Peterson Journal, 147 (3) : 442-448. M G E, Marx R G. (2005), The association between hospital volume and total shoulder [138] Nallamothu B. K., Saint S., Hofer T. P., arthroplasty outcomes. Clinical Orthopaedics Vijan S., Eagle K. A., Bernstein S. J. (2005), and Related Research, 432 : 132-137. Impact of Patient Risk on the Hospital Volume–Outcome Relationship in Coronary [130] Narayan P., Caputo M., Rogers C.A., Alwair Artery Bypass Grafting ; Arch Intern Med. H., Mahesh B., Angelini G.D., Bryan A.J. 165 :333-337. (2004), Early and mid-term outcomes of

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 53 -

[139] Feasby T.E., Quan H., Ghali W.A. (2002), O.R.C., Obertop H. (2000), Rates of Hospital and Surgeon Determinants of Complications and Death After Pancreatic- Carotid Endarterectomy Outcomes. Arch oduodenectomy : Risk Factors and the Neurol, 59 :1877-1881. Impact of Hospital Volume. Ann Surg, 232 (6) : 786-795. [140] Kansagra S.M., Curtis L.H., Anstrom K.J., Schulman K.A. (2007), Trends in operator [149] Marusch F., Koch A., Schmidt U., Pross M., and hospital procedure volume and outcomes Castinger I. and Lippert H. (2001), Hospi- for percutaneous transluminal coronary tal caseload and the results achieved in angioplasty, 1996 to 2001. Am J Cardiol, 99 patients with rectal cancer. British Journal (3) : 339-43. of Surgery, 88, 1397-1402.

[141] Welke K.F., Barnett M.J., Vaughan Sarrazin [150] Van Lanschot J.J.B., Hulscher J.B.F., M.S., Rosenthal G.E. (2005), Limitations Buskens C.J., Tilanus H.W., ten Kate of hospital volume as a measure of qua- F.J.W., Obertop H. (2001), Hospital volume lity of care for coronary artery bypass graft and hospital mortality for esophagectomy; surgery, The Society of Thoracic Surgeons. American Cancer Society.

[142] Plomondon M. E., Casebeer A.W., Schooley [151] Glasgow E.R., Showstack J.A., Katz P.P., L.M., Wagner B.D., Grunwald G.K., Corvera C.U., Warren R.S., Mulvihill S.J. McDonald G. O., Grover F. L., Shroyer A. L. (1999), The Relationship Between Hospital W. (2006), Exploring the volume-outcome Volume and Outcomes of Hepatic Resection relationship for off-pump coronary artery for Hepatocellular Carcinoma. Arch Surg, bypass graft procedures ; The Society of 134 : 30-35. Thoracic Surgeons. [152] Dimick J.B., Goodney P.P., Orringer M.B., [143] Westvik H.H., Westvik T.S., Maloney S.P., Birkmeyer J.D., (2005), Specialty Training Kudo F.A., Muto A., Leite J.O.M., Gusberg and Mortality After Esophageal Cancer R.J., Cha C. and Dardik A. (2006), Hospi- Resection. Ann Thorac Surg, 80 : 282-286. tal-based factors predict outcome after ca- rotid endarterectomy; Journal of Surgical [153] Van Heek N.T., Kuhlmann K.F., Scholten R.J., Research 134, 74-80. de Castro S.M., Busch O.R., van Gulik T.M., Obertop H., Gouma D.J. (2005), Hospital [144] Glance L.G., Dick A.W., Mukamel D.B., volume and mortality after pancreatic Osler T.M. (2003), Is the hospital volume- resection : a systematic review and an mortality relationship in coronary artery by- evaluation of intervention in the Netherlands; pass surgery the same for low-risk versus Annals Surgery ; 242(6):781-8, discussion high-risk patients? Ann Thorac Surg, 76 : 788-90. 1155-1162. [154] Dimick J.B., Cattaneo S.M., Lipsett P.A., Pro- [145] Pasquale MD, Peitzman AB, Bednarski novost P.J., Heitmiller R.F. (2001), Hospital J., Wasser TE. (2001), Outcome analysis volume is related to clinical and economic of Pennsylvania trauma centers : factors outcomes of esophageal resection predictive of nonsurvival in seriously injured in Maryland. Ann Thorac Surg, 72 : 334-340. patients ; The Journal of Trauma ; 50(3):4. [155] Vitale M.A., Arons R.R., Hyman J.E., [146] Ly D.P., Liao J.G. Burd R.S. (2005), Effect Skaggs D.L., Roye D.P., Vitale M.G. (2005), of Surgeon and Hospital Characteristics on The contribution of hospital volume, payer Outcome After Pyloromyotomy. Arch Surg, status, and other factors on the surgical 140 : 1191-1197. outcomes of scoliosis patients. A review of 3,606 Cases in the States of California ; [147] Wenner J., Zilling Th., Bladström A., J. Pediatr Orthop. Volume 25, Number 3. Alvegard T.A. (2005), The influence of surgical volume on hospital mortality and 5-year [156] Morris D.S., Taub D.A., Wei J.T., Dunn R.L., survival for carcinoma of the oesophagus Wolf J.S., Hollenbeck Jr. and B.K. (2006), and gastric cardia, Anticancer Research Regionalization of percutaneous nephroli- 25 : 419-424. thotomy : evidence for the increasing burden of care on tertiary centers. The Journal of [148] Gouma D.J., van Geenen R.C.I., van Gulik Urology. Vol. 176, 242-246. T.M., de Haan R.J., de Wit L.T., Busch

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 54 -

[157] Birkmeyer J.D., Stukel T.A., Siewers A.E., [166] McKee M.D., Cropp M.D., Hyland A., Goodney P.P., Wennberg D.E., Lucas F.L. Watroba N., McKinley B., Edge S.B. (2002), (2003). Surgeon volume and operative Provider Case Volume and Outcome in the mortality in the United States. N. Engl J. evaluation and Treatment of Patients with Med, 349 (22) : 2117-2127. Mammogram-Detected Breast Carcinoma, American Cancer Society. [158] Smith E. R., Butler W. E., Barker F. G. (2004), In-hospital mortality rates after [167] Forshaw M.J., Gossage J.A., Stephens J., ventriculoperitoneal shunt procedures in Strauss D., Botha A.J., Atkinson S., Mason the United States, 1998 to 2000 : relation R.C. (2006), Centralisation of oesophago- to hospital and surgeon volume of care ; gastric cancer services : can specialist units Journal of neurosurgery. deliver? Ann R College Surg Eng, 88 (6) : 566-570. [159] Morss Dy S., Cromwell D. M., Thuluvath P. J. and Bass E.B. (2003), Hospital experience [168] McGrath D.R., Leong D.C., Gibberd R., and outcomes for esophageal variceal Armstrong B., Spigelman A.D. (2005), bleeding ; International Journal for Quality Surgeon and hospital volume and the in Health Care, 15 :139-146. management of colorectal cancer patients in Australia. ANZ Journal of Surgery, 75 [160] Billingsley K.G., Morris A.M., Dominitz J.A., (10), 901–910. Matthews B., Dobie S., Barlow W., Wright G.E., Baldwin L.M. (2007). Surgeon and [169] Harcourt K.F., Hicks K.L. (2003), Is there Hospital Characteristics as Predictors of a relationship between case volume and Major Adverse Outcomes Following Co- survival in breast cancer? The American lon Cancer Surgery : Understanding the Journal of Surgery, 185 : 407-410. Volume-Outcome Relationship. Arch Surg. 142 : 23-31. [170] Schrag D., Panageas K.S., Riedel E., Cramer L.D., Guillem J.G., Bach P.B., [161] Ho V. (2002), Learning and the evolution Begg C.B. (2002), Hospital and surgeon of medical technologies : the diffusion of procedure volume as predictors of outcome coronary angioplasty. Journal of Health following rectal cancer resection. Annals of Economics 21; 873-885. Surgery. Vol. 236, N°5, 583-592.

[162] Dasgupta R., Kim P. (2005), Relationship [171] Rogowski J.A., Horbar J.D., Staiger D.O., between surgical volume and clinical Kenny M., Carpenter J., Geppert J. (2004), outcome : should pediatric surgeons be Indirect vs direct hospital quality indicators doing pancreaticoduodenectomies? Journal for very low-birth-weight infants. JAMA. Vol. of Pediatric Surgery, 40, (5) : 793-796. 291 n°2.

[163] Konety B., Dhawan V., Allareddy V., Joslyn [172] Weissman J.S., Rothschild J.M., Benda- S., (2005), Impact of hospital and surgeon vid E., Sprivulis P., Fachi F., Cook E.F., volume on in-hospital mortality from radical Evans R.S., Kaganova Y., Bender M., Da- cystectomy : data from the health care vid-Kasdan J.A., Haug P., Lloyd J., Selbo- utilization project. The Journal of Urology, vitz L.G., Murff H.J., Bates D.W. (2007), 173 (5) : 1695-1700. Hospital workload and adverse events. Medical Care. Vol. 45, N°5. [164] Anger J., Rodríguez L., Wang Q., Pashos C., Litwin M. (2007), The Role of [173] Richardson D.B. (2006), Increase in patient Provider Volume on Outcomes After Sling mortality at 10 days associated with emer- Surgery for Stress Urinary Incontinence. The gency department overcrowding. MJA, Vol. Journal of Urology, 177 (4) : 1457-1462. 184, N°5.

[165] Tracy S.K., Sullivan E., Dahlen H., Black [174] Sprivulis P.C., Da Silva J.-A., Jacobs I.G., D., Wang Y.A., Tracy M.B. (2006) Does size Frazer A.RL. and Jelinek G.A. (2006), The matter? A population-based study of birth association between hospital overcrowding in lower volume maternity hospitals for low and mortality among patients admitted risk women; BJOG International Journal of via western Australian emergency depart- Obstetrics and Gynaecology. An . ments. MJA, Vol. 184, N°5.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 55 -

[175] Hannan E.L., Racz M., Kavey RE., Farley D.E., Ozminkowski R.J. (1992); Volume- Quaegebeur J.M., Williams R. (1998), Outcome relationships and inhospital mortality : Pediatric cardiac surgery : the effect of the effect of changes in volume over time. Medical hospital and surgeon volume on in-hospital Care, Vol. 30, N°1. mortality. Pediatrics, Vol. 101, N°6; 963-969. Gaynor M. (2004); Competition and quality in health care markets. What do we know? What don’t we know? Economie publique n°15- Bibliographie complémentaire 2, 3-40.

Argote L., Beckman S.L. and Epple D. (1990); The Gaynor M., Seider H., Vogt W.B. (2005); The persistence and transfer of learning in industrial volume-outcome effect, scale , and settings. Management Science. Vol. 36, N°2. learning by doing. The American Economic Review, vol. 95, no.2, pp. 243-47. Banker R.D., Chang H., Cunningham R. (2003); The public accounting industry production function. Gaynor M. (2006); What do we know about Journal of Accounting and Economics 35; 255-281. competition and quality in health care markets? Centre for Market and Public Organisation, CMPO, Barro J.R. (2005); The effects of cardiac specialty Working Paper N° 06/151. hospitals on the cost and quality of medical care. NBER Working Paper N° 11707. Geubbels E., Wille J C., Nagelkerke N J.D., Vandenbroucke-Grauls C., Grobbee D. E., De Boer Besanko D., Doraszelski U., Kryukov Y., AS. (2005); Hospital-Related Determinants for Satterthwaite M. (2007); Learning-by-doing, Surgical-Site Infection Following Hip Arthroplasty, organizational forgetting and industry dynamics. Infection control and Hospital Epidemiology, Center for Economic Policy Research, Discussion Vol. 26, N°5, pp. 435-441. Paper N° 6160. Goodney P.P., Stukel T.A., Lucas F.L., and others Birkmeyer JD., Siewers AE., Finlayson E V.A., (2003); Hospital volume, length of stay, and read- Stukel T A., Lucas F. L., Batista I., Welch H. mission rates in high-risk surgery, (AHRQ grant G., Wennberg D E. (2002); Hospital Volume HS10141), Annals of Surgery 238(2), pp. 161-167. and Surgical Mortality in the United States, The New England Journal of Medicine, Volume 346 : Gowrisankaran G., Ho V., Town R. (2006); 1128-1137 N° 15. Causality, learning and forgetting in surgery, working paper (on line: http://www.u.arizona. Darr E.D., Argote L., Epple D. (1995); The edu/~gowrisan/pdf_papers/learn_forget_surgery. acquisition, transfer, and depreciation of knowledge pdf). in service organizations : productivity in franchises. Management Science, Vol. 41, N° 11. Hannan EL., Kilburn H., O’Donnell J F., Bernard H. R. Shields E P. Lindsey M L., Yazici Dranove D. (1998); “Economies of scale in non- A. (1992); A Longitudinal Analysis of the Relation- revenu producing cost centers: implications for ship between In-Hospital Mortality in New-York hospital merger”, Journal of Health Economics, State and the Volume of Abdominal Aortic Aneu- 17(1), 69-83. rysm Surgeries Performed, Health Services Re- search. Dudley R A., Johansen K L., Brand R., Rennie D J., Milstein A. (2000); Selective referral to high- Hillner BE., Smith TJ., Desch C.E. (2000), volume hospitals : estimating potentially avoidable Hospital and physician volume or specialization deaths, Jama : the journal of the American Medical and outcomes in cancer treatment : importance in Association, Vol. 283, N° 9, pp. 1159-66. quality of cancer care, J. Clin Oncol, 18(11): 2327-40.

Elixhauser A., Steiner C., Fraser I. (2003); Volume Hirsch W. (1952); “Manufacturing progress func- Thresholds And Hospital Characteristics In The tion”, Review of Economics and Statistics, vol. 34, United States, Health Affairs, Volume 22, Number 2. no.2, 143-155.

Ethan A, Lee C., Chassin M. (2002); Is volume Hirschmann W. (1956); “Profit from the learning related to outcome in health care? A systematic curve”, Harvard Business Review, vol.42, no.1, review and methodological critique of the literature, pp.125-139. Ann Internal Medicine, 137 : 511-520.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 56 -

Ho V., Heslin M J, Yun H., and Howard L. (2006); Sheldon TA. (2004); The volume-quality relation- Trends in Hospital and Surgeon Volume and ship : insufficient evidence for use as a quality Operative Mortality for Cancer Surgery, Annals of indicator, Quali. Saf. Health Care, 13, 325-326. Surgical Oncology. Slim K, Chipponi F. (2002); Relation entre le Hodgson C.D., Zhang W., Zaslavsky A.M. and volume d’activité et résultats du chirurgien : mythe others. (2003); Relation of hospital volume to ou réalité?, Annales de Chirurgie, 127 : 502-511. colostomy rates and survival for patients with rectal cancer, (AHRQ grant HS09869), Journal of Sowden A, Grilli R, Rice N. (1997); The relationship the National Cancer Institute 95(10), p. 708-716. between hospital volume and quality of health outcomes, CRD Report 8 (Part 1), University of Hugues RG., Hunt SS., Luft HS. (1987); Effects of York. surgeon volume and hospital volume on quality of care in hospitals, Med Care, 25(6): 489-503. Stobaugh R. et Townsend P. (1975); « Price fo- recasting and strategic planning: the case of pe- Huber G. (1991); “Organizational learning: the trochemicals », Journal of Marketing Research, contributing processes and the literatures”, Orga- vol.12, pp.19-29. nization science, vol 2, n°1, 88-115. Tay A. (2003); “Assessing competition in hospital ICIS (2005); Analysis of current research related care markets: the importance of accounting for to the impact of low volume procedures and care quality differentiation”, RAND Journal of Econo- on outcomes of care, Canada, Institute for Clinical mics, vol.34, no.4, pp. 786-814. Evaluative Sciences. Thiemann D.R., Coresh J., Oetgen W. J., Powe Kimberly J. (1981); “Managerial innovation” , in N.R. (1999); The Association between Hospi- Nystrom P.C et Starbuck W. ( Eds), Handbook of tal Volume and Survival after Acute Myocardial organizational design, vol. 1, London : Oxford Uni- Infarction in Elderly Patients, The New England versity Press, 85-104. Journal of Medicine, N° 21.

Lieberman M.B. (1984); The learning curve and Urbach D., Baxter N. (2004); Does it matter what pricing in the chemical processing industries, a hospital is „high volume“ for? Specificity of Rand Journal of Economics, Vol. 15, N° 2. hospital volume-outcome associations for surgical procedures : analysis of administrative data, BMJ, Leroux F (1980); Introduction à l’économie 328; 737-740. d’entreprise, Boucherville : 1980. Vettori G. (2000); Les Economies d’Echelle : Du Luft H.S., Hunt S.S., Maerki S.C. (1987); The concept à l’application, le secteur bancaire suis- volume-outcome relationship : practice-makes- se, Cahier de recherche, Université de Genève, perfect or selective-referral patterns? HSR : Health 2000. Services Research 22:2. Wright T.P. (1936); “Factors affecting cost of air- Luft H., Bunker J., Enthoven A. (1979); “Should planes”, Journal of Aeronotical Science, vol.3, operations be regionalized? The empirical rela- no.4, pp.122-128. tionship between surgical volume and mortality, New England Journal of Medicine, 301(25), 1364- Zimmerman H. ( 2002); A report on the Volume- 69. Quality Relationship in Hospital Tertiary Care Services in Rhode Island : A report on Compliance with the Hanaway Act.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 Liste des tableaux et graphiques

- 59 -

Liste des tableaux

Tableau 1 Répartition des études par pays ...... 22

Tableau 2 Nature des données utilisées dans la littérature ...... 23

Tableau 3 Nature des soins et des interventions étudiés dans la littérature ...... 23

Tableau 4 Nature des interventions liées au cancer étudiées dans la littérature ...... 24

Tableau 5 Nature des interventions cardio-vasculaires étudiées dans la littérature ...... 25

Tableau 6 Nature des interventions orthopédiques étudiées dans la littérature ...... 26

Tableau 7 Nature des autres interventions chirurgicales étudiées dans la littérature ...... 26

Tableau 8 Indicateurs de résultats utilisés selon les interventions étudiées ...... 28

Liste des graphiques

Graphique 1 Répartition des études par année de publication ...... 21

Graphique 2 Les formes envisageables de la relation « volume d’activité - mortalité »...... 36

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008

Annexes

- 63 -

Annexe 1 - Stratégie de recherche bibliographique

Les filtres utilisés pour limiter la recherche sur Research 3 : différentes bases sont : Centralization or Centralized or Centrali- – Publication dans les dix dernières années sation or Centralised and Hospitals and [Quality or Outcome(s) or performance] – Uniquement les articles contenant les mots- clés dans les titres et résumés À partir de ces mots-clés, nous avons procédé à des combinaisons en rajoutant d’autres mots- – Langues utilisées : anglais, français clés dont : « relationship », « impact », ou « relation ». – Exclusion des revues et des méta- analyses dans « type de publication » Le tableau de la page suivante présente les différentes combinaisons des mots-clés, le – Inclusion des ouvrages, rapports, thèses, nombre d’articles identifiés sur toutes les bases congrès avec ces mots-clés, ainsi que ceux choisis pour la revue, après élimination des articles hors sujets. Research 1 : Ainsi, nous avons parcouru les titres et/ou Volume and Outcome(s) or Quality or abstracts de près de 1 000 articles. Après élimination Mortality or Survival or Readmission or des doublons, nous avons recensé au total 226 Infection or Length of stay références uniques entrant dans le cadre de l’étude sur l’incidence du volume sur la qualité. 29 articles Research 2 : ont été exclus de la revue car ils ne correspon- daient pas à l’étude empirique ou qu’ils n’avaient Physician or Surgeon and Volume and pas la relation « volume-qualité » comme pro- [Quality or Outcomes or Mortality or Survival blématique principale. 22 autres ont été exclus or Readmission] dans la mesure ou nous n’avons pas réussi à les obtenir dans les délais de notre étude ou à un coût raisonnable.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 64 -

Mots – Clés Étude Select Result

«Volume » [Title/Abstract] AND (« outcomes » [Title/Abstract OR (« quality » [Title/Abstract] OR « mortality » [Title/Abstract] OR « survival » [Title/Abstract] OR « readmission » [Title/Abstract] OR « infection » [Title/Abstract] OR « length of stay » [Title/Abstract])) AND (« hospitals » [Title/Abstract] OR « hospital » [Title/Abstract]) NOT H1 202 123 Review [Publication Type] NOT Meta-Analysis [Publication Type] AND « relationship » [Title/Abstract]

Field : Title/Abstract, Limits : added to PubMed in the last 10 years, English, French, published in the last 10 years, Humans

Research 1

« Volume » [Title/Abstract] AND (« outcomes » [Title/Abstract] OR (« quality » [Title/Abstract] OR « mortality » [Title/Abstract] OR « survival » [Title/Abstract] OR « readmission » [Title/Abstract] OR « infection » [Title/Abstract] OR « length of stay » [Title/Abstract])) AND (« hospitals » [Title/Abstract] OR « hospital » [Title/Abstract]) NOT H2 66 38 Review [Publication Type] NOT Meta-Analysis [Publication Type] NOT « relationship » [Title/Abstract] AND « relation » [Title/Abstract]

Field : Title/Abstract, Limits : added to PubMed in the last 10 years, English, French, published in the last 10 years, Humans.

« Volume » [Title/Abstract] AND (« outcomes » [Title/Abstract] OR (« quality » [Title/Abstract] OR « mortality » [Title/Abstract] OR « survival » [Title/Abstract] OR « readmission » [Title/Abstract] OR « infection » [Title/Abstract] OR « length of stay » [Title/Abstract])) AND (« hospitals » [Title/Abstract] OR « hospital » [Title/Abstract]) Research 1 H3 200 35 NOT Review [Publication Type] NOT Meta-Analysis [Publication Type] NOT « relationship » [Title/Abstract] AND « impact »[Title/Abstract]

Field : Title/Abstract, Limits : added to PubMed in the last 10 years, English, French, published in the last 10 years, Humans.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 65 -

Mots – Clés Étude Select Result

« Volume » [Title/Abstract] AND (« outcomes » [Title/Abstract OR (« quality » [Title/Abstract] OR « mortality » [Title/Abstract] OR « survival » [Title/Abstract] OR « readmission » [Title/Abstract] OR « infection » [Title/Abstract] OR « length of stay » [Title/Abstract])) AND (« physician » [Title/Abstract] OR « surgeon » [Title/Abstract] OR D1 154 84 « surgical » [Title/Abstract]) NOT « review » [Publication Type] NOT « meta analysis » [Publication Type] AND « relationship » [Title/Abstract]

Field : Title/Abstract, Limits : added to PubMed in the last 10 years, English, French, published in the last 10 years, Humans.

Research 2 « Volume » [Title/Abstract] AND (« outcomes » [Title/Abstract] OR (« quality » [Title/Abstract] OR « mortality » [Title/Abstract] OR « survival » [Title/Abstract] OR « readmission » [Title/Abstract OR « infection » [Title/Abstract] OR « length of stay » [Title/Abstract])) AND (« physician » [Title/Abstract] OR « surgeon » [Title/Abstract] OR D2 « surgical » [Title/Abstract]) NOT « review » [Publication Type] NOT 55 27 « meta analysis » [Publication Type] NOT « relationship » [Title/Abstract] AND Relation [Title/Abstract]

Field : Title/Abstract, Limits : added to PubMed in the last 10 years, English, French, published in the last 10 years, Humans.

« Volume » [Title/Abstract] AND (« outcomes » [Title/Abstract] OR (« quality » [Title/Abstract] OR « mortality » [Title/Abstract] OR « survival » [Title/Abstract] OR « readmission » [Title/Abstract] OR « infection » [Title/Abstract] OR « length of stay » [Title/Abstract])) AND (« physician » [Title/Abstract] OR « surgeon » [Title/Abstract] OR Research 2 D3 « surgical » [Title/Abstract]) NOT « review » [Publication Type] NOT 160 23 « meta analysis » [Publication Type] NOT « relationship » [Title/Abstract] AND « impact » [Title/Abstract]

Field : Title/Abstract, Limits : added to PubMed in the last 10 years, English, French, published in the last 10 years, Humans.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 66 -

Mots – Clés Étude Select Result

(« Centralisation »[Title/Abstract] OR « centralization »[Title/Abstract] OR « centralised »[Title/Abstract] OR centralized[Title/Abstract]) AND ((«hospitals »[Title/Abstract] OR « hospital »[Title/Abstract]) AND « quality »[Title/Abstract] OR « outcomes »[Title/Abstract] OR Research « performance »[Title/Abstract]) NOT « Review »[Publication Type] C1 10 2 3 NOT « Meta-Analysis »[Publication Type] AND « relationship »[Title/ Abstract]

Field : Title/Abstract, Limits : added to PubMed in the last 10 years, English, French, published in the last 10 years, Humans.

(« Centralisation »[Title/Abstract] OR « centralization »[Title/ Abstract] OR « centralised »[Title/Abstract] OR « centralized »[Title/ Abstract]) AND (« hospitals »[Title/Abstract] AND (« quality »[Title/ Abstract] OR « outcomes »[Title/Abstract] OR « performance »[Title/ Research Abstract])) NOT « review »[Publication Type] NOT « meta C2 2 1 3 analysis »[Publication Type] NOT

« Relationship »[Title/Abstract] AND « relation »[Title/Abstract]

Field : Title/Abstract, Limits : added to PubMed in the last 10 years, English, French, published in the last 10 years, Humans.

Cindoc, Embase, - Recherché avec les mots clés précédents 363 45 Pascal

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 67 -

Annexe 2 - Grille d’analyse

Méthode Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Méthode d’étude Observation statistique Relation entre volume Chirurgien < 50 PTH ; =>50 PTH ; + 100 PTH. Modèle linéaire - CalculRégression odds logistique. Age, sexe, motif de Mise en place Mortalité intra- d’activité et mortalité intra- 1 1997 France/ n = 5,521 PTH ; + Hôpital ratios. l’intervention chirrugicale ; d’une prothèse hospitalière. hospitalière apparaît faible si Rhône- 50 (Nombre total nombre de PTH par de la hanche ajustement sur âge et le motif Alpes Etablissement. de PTH/An). établissement. (PTH). de l’intervention. L’association entre le niveau d’activité de l’établissement et la mortalité paraît vraisemblable seulement dans tranche d’âge 82 à 89.

2 1989- USA 275 Surgeons Logistic linear regression, Physician age, prestige and Mortality ratios decreased with 1992 Surgeon Surgeon with > 200 patients. Tested using correlations Physician experience of surgeons, increasead volume and years Cardio/coronary rates. weighted by the number of Chi-square tests. performance, number of years pratice. of experience of surgeons, age, artery bypass patients that the physicians 83,547 Patients surgeon’s and academic rank ; Surgical treated. 200 operations observed performance not associated during 3 years predicted patient with graduation and prestigious in New-York mortality. school, residency. and 30 operations Hospital rates Quartile of surgical volume : Hvol 3 1990- Canada, nper = 7,483year in Hip fracture Postsurgery Pennsylvannia. 1993 Quebec length of stay Number of comorbidities ; Age,Signi ficant negative relationship and impatient Determinants of postsurgerySex, Married, Income/10 000 between; surgical volume and =<32; Longitudinal data (OLS), inhospital h length of stay assessed both post surgery length of mortalty. 33 = < Hvol<= 49 ; 50 =< Hvol<= 72 data from the MEDECHO h h using ordinary least stay and in-hospital mortality Hvol >= 73. database contains h squares regression and among hip fracture, persisted information from hospital the explanators of inpatient with time. discharge abstract.. mortality using logistic regression with hospital specificfixed effects, to control for systematic differences in outcomes that persist with time.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 68 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôléesPatient with Résultats highest volume étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Number of patient in a pratice ; d’étude Using claims-based CurrentHierarchical Linear Models physician quintile were frequency of visits, physician Physician Patient volume refers to the numbersProcedural Terminology (HGLMs) ; Initial bivariate significantly more likely to get experience, patient comorbidity, Diabetic of elderQuality of care, rates. of Medicare diabetic patients (CPT). analyses. these care than patient in the 4 2001 USA/ 65 years or patient demographics (age, patients. as assessed by divided into 5 groups of Quintiles : lowest physician quintile. Connecticut older ; the claims-based Group I = 1-4 ; Group II = 5-10 ; Group gender, ethnicity) ; chronic 26,260 processes of III = 11-18 ; Group IV = 19 - 31 and illness. Medicare care hemoglobin Group V = 32 - 166 ; Physicians in beneficiaries A lc testing, lipid volume group IV and V care for 75 % of with diabetes. pro le testing ; the patient. Hip fracture. Lenghtfi of stay, Hospital Number of operation by hospital 68 hospitals and retinal eye 5 1991- Canada, post surgery in- h in the time periodprior : with 53 examinations. 1993 Quebec hospital mortality. HVOL surgeries per year ; 6% Higher volume associate with Longitudinal data. Proportional hazards Year of admission, median of patient an increased conditional of male income in postal code of admitted with a specification, controlling live discharge. Signi cant residence and number and type fi hip fracture. hospital - Fixed effect. relationship between surgical = tot of number of surgeries of comorbities at the time of ht volume ant length of stay. performed in hospitalh in the 12 months admission. Charlson index. prior to the date (t) of the current surgery, Average volume in 3 groups : The effect of volume on low volume : < 34 surgeries = > 25 % ; Comorbid conditions, vital A linear probability regression mortality was smaller in Average : 34-71 surgeries = > 50%; Compares administrative signs, clinical status, and ; magnitude when clinical data 30 days in- Hospital rates. Bed size accross ; The hospitals in the model Cardio/ High : > 71 surgerie = > 25 %. versus use of clinical data. demographic characteristics is used. A stategy of volume 6 1991- USA/Ohio 21,555 hospital mortality. sample was 276 (range 84–906), the (LPM) was fit to the congestive heart Controls for potential (age, Gender, race) ; laboratorybased referral for CHF may 1997 Medicare average CHF case volume was 459 Longitidunal data to assess failure (CHF). endogeneity in hospital test results for Medicare- not really improve outcomes for patients aged (77–1,203), and the average resident- the effects of hospital volume volume with instrumental insured patients aged 65 yearselderly patients (different from 65 years and to-bed ratio was 0.13 (0–0.69). on 30-day mortality ; with and older. older in all variables. Checks the impactinstrumental variables. surgery). in-hospital Hospital Intra- of patient distance to hospital. 12,730non federal Cardio/ 7 1994- USA/South mortality Aortic Ballon hospitals. (AMI) Acute of San patients at 750 1998 Myocardial Francisco -hospitals. Infarction. California 150 fewer deaths per 1 000 patients treated at the high Demographics (age, sex, Hospitals divided into tertiles IABP Cohort study. NRMI-2, a voluntary, IABP hospitals. High hospital race), medical history, clinical Volume : 64,4% Lowest ; 54,1% observational database. The IABP volume for patients charasteristics (Diabetes, method of diagnosis of AMI. with acute myocardial Counterpulsation. intermediate ; and 50,6 % Highest Hypertension, Angina, Stroke, Analised with a series of infarction was associated with volume (P for trend < 0.001). Chest pain on admission, ECG logistic regression. lower mortality. Among the findings) ; procedures, myocardial infarction patients Additional analyses were with cardiogenic shock who performed by forcing variables underwent IABP placement, such as number of AMI mortality rate case signi cantly cases per hospital per year, fi lower at high - IABP volume cardiogenic hospitals compared with low- shock cases per year, no IABP volume hospitals. of staffed beds, and shock patients receiving IABP).

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 69 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) d’étude Health employer Data and Logistic regression. Patients charateristics Plans in the lowest quartile Mental health Information Set (HEDIS) (age, sex), payment method of outpatient and inpatient 8 1999- USA/Eight 384 plans Quartile of Total annual Three measures of volume of mental care. 2000 Descriptive data. (Medicare, Medicaid), region.mental health volume had : 2000 geographicspatients ; performance ambulatory health servicesfi : rst outpatient measure an 8,5 (95% CI [confidence regions, covering on each mental visits, inpatientexamined the total annual number of interval]=4,97 -14,37) to US. HMO 73 millions health quality discharges plan members using any ambulatory 21,09 (95% CI=11,32-39,28) enrollees measure. HEDISservices, mental health services. Two others : times increase in odds if notionwide. mental health and inpatient inpatient volume of care (the total poor 7- and 30 day follow-up performance days - Hospitalnumber of members with an inpatient after discharge from inpatient measures : level. mental health hospitalization and the psychiatric hospitalization. 7 and 30 days total number of inpatient mental health Low volume plans had a outpatient days). Low volume of ambulatory 3,5 ( 95% CI= 2,15 - 5,67) follow-up after mental health use (<1, 090 meantal to 5,42 (CI= 3,21 - 9,15) discharge ; health patient/year) ; Low volume of times increase in odds of Acute (enrollees inpatient mental health discharges poor performance on the took an (<65 Inpatient hospitalization/year) ; acute,continuous and provider antidepresant Low volume of inpatient mental health measures of antidepressant continuously for at least days (<36 inpatient days/year). treatment. Large and 3 months) ; consistent association between Continuation mental health volume and (enrollees took an performance. antidepressant continuously for at least 6 months)

Practitioner contacts (enrolles took an antidepressant Association between ICU 9 1991 - during at least Age, gender, admission volume and risk-adjusted 1997 three follow-up Retrospective cohort study. The X² test or analysis ofseverity of illness (measured mortality were significant for visits). variance. Models which areby APACH), admitting patient with GI diagnoses In-hospital Hospital Tertiles of hospitals volume : high, an extension of the usual diagnosis and source, dates and for sicker patients with Patients admission. medium, low based on cut-offs that US, 29 hospitals mortality. of ICU and hospital admissionrespiratory diagnoses. admitted to ICU yielded roughly equivalent number of Cox regression model. Using Northest Respiratory n and discharge, vital status However, associations were with Respiratory patient in each volume category. In cox proportionnal hazards Ohio Single= 16,949 at discharge, discharge not signi cant for patient diagnoses, the pulmonary disease cohort, low- models for compare in- fi metropolitanneurologic destination, ICU and Hospitalwith neurologic diagnoses. neurologic volume hospitals had < 500 admissions, hospitals mortality between area, USA n = 13,805; LOS, presence of seven The lack of a consistent diagnoses and medium-volume between 500 and tertiles of hospitals volume for GI dideases speci c comorbid conditions. volume-outcome relationship GI disorders. 1 000 admissions, and high-volume ICU. Severity of illness was fi n = 12,881. may re ect unmeasured had >1 000. In the GI and neurology measured using the APACHE fl patient complexity in higher cohorts, low volume hospitals admitted III methodology. volume hospitals, relative < 400 patients, medium-volume standardization of care across admitted 400 to 700, and high-volume ICs, or lack of ef cacy of some admitted > 700 patients during the study fi accepted ICU process of care. period.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 70 -

Risk-adjusted measures of Pays et/ Baselines characteristics (Age,both death and the composite Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir adverse outcome were N° ou région MéthodeThe relations of physician sex),Variables Presentation contrôlées with acute Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) volume and years of MI (onset 24 hours), Canadianinversely related to the number d’étude Risk-adjusted measures experiences to raw and Cardiovascular Society anginaof cases each operator Physician High and low volume operators ; 38 of both death and the adjusted outcome were class, cardiogenic shock, left performed annually but bore 10 1993 to USA 38 PhysiciansCardio/ ; In-hospital rates. phycisians with >= 30 cases per yearcomposite ; adverse outcomeevaluated with least- ventricular ejection fraction, no relation to total years of 1994 5 high volume Percutaneous Death ; models the average physician performed a were inversely related squares linear regression modified ACC/AHA lesion experience. Both adverse centers; (1 000Coronary predictive of mean +/-SD of 163 +/- 24 cases per to the number of cases techniques. models used morphology classification outcomes were more closely procedures perRevascularization. death and the year ; physician operators were also each operator performed Risk adjustement, Logistic score, number of diseased related to the logarithm of year); composite divided into quintiles by caseload, andannually. Each maintains regression Models. vessels, prior bypass surgery,caseload (for death, r =.37, 12,985 Patients outcome of the relationship between caseload credentialing standards for prior restenosis, and unstableP =.01; for death, Q-wave treated. death, Q-wave quintile and risk-adjusted results in individual physicians and angina were recorded. infarction, or bypass surgery, infarction, or low risk (ACC/AHA A or B1) and high-an interventional database r =.58, P<.001) than to linear caseload. emergency risk (ACC/AHA B2 or C) lesions was characterized by prospective A relationship between bypass surgery. evaluated. entry of selected clinical Patient characteristics (age andhospital volume and all cause and angiographic data, Cox proportional hazards race) and tumor characteristicsa mortality, that is there were routine postprocedure regression models for For each patient with (size, grade, nodal moderate reductions in both Hospital rates. Hospital volume was categorized intoanalysis 3 of ECGs analysis survival data/Cohort study. breast cancer, information involvement, and hormone all cause mortality and breast Breast cancer 5-year survival groups (low : 0–19 cases per year ; and coding of complications obtained on demographic receptor status) ; comorbidity cancer cancer-specific mortality operations. rates - All causes medium : by trained personnel, and 11,225 characteristics, extent of index for each patient with among women treated in 11 1994 USA and breast 20–39 cases per year ; high : 40 or moreinternal audits and checks Medicare disease, and initial therapy. Charlson Index (comprises 15hospitals with annual volumes -1996 -New York cancer specific cases per year) with approximately for data completeness and patients ; 457 SEER information has noncancer conditions, each ofof 40 or more operations State and - Two outcomes equivalent consistency. hospitals ; 11 been linked to Medicare which is weighted according performed on Medicare breast California analysed : time numbers of patients in each. population claims for people aged cancer patients. until death by to its impact on mortality) ; based tumor 65 years and older. Used any cause and capita income and educational registries Medicare Provider Analysis time until death level ; the size of metropolitan covering 14% and Review and outpatient by breast cancer. standard area. of the US Standard Analytic File data pop. Medicare to determine the hospital claims for 94% at which the operation was of people aged performed. 65 years and older.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 71 -

Pays et/ In this unselected population- Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir based series there was N° ou région Méthode All Variablesresections ;contrôlées Oesophagus, Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Cardia, Gender, Stage, little evidence of a trend of d’étude The study was a retrospectivePearson and Mantel- American Society of improving 30-day mortality rate Surgical Surgeon workload : 41 Infrequent case-note review. Haenszel X² test for linear Anesthesiologist (ASA) with increasing workload, or 1,125 patients. Cardio- 30-day mortality UK region annual operators < 4 resections per year n = association to test differencesgrade, smoking habit, Age, between workload and long 12 1992 rate ; - West oesophageal worklaod. 146 (13%); 18 Intermediate operators in 30-mortality. Kaplan-MeierDisease site, Histological term survival. The 30-day -1996 5-year survival Midlands cancer. between 4-11 resections per year n survival rates were estimatedtype, Emergency resection ormortality rate was significantly rate and Long region of =488 ( 43,4%); 5 Frequent operators and survival curves were Elective resection, (operation,increased in patients having term survival. In- England 12 or more resections per year n =491 compared by means of palliative, curative), whether emergency resection (P =0,005), hospital mortality. (43,6%); Low volume unit (< 20 per log rank methods. Logistic preoperative chemotherapy in patient aged over 70 years year) ; High volume unit (> 20 per regression was used to was administred. No(P =0,001). association The between mortality the rate year) . identify predictors of 30- numberwas higher of CABG and long-term procedures Model performance was day mortality and thereby performedsurvival worseoff-pump in the by 16% of assessed using the c-statistic. to estimate adjusted odds apatients surgeon whose and in-hospital full hospital Using random-effects logisticDemographics (ie, age, gender, A retrospective cohort studyratios. Cox’s proportional mortalitynotes were (p not 0.93) available after for regression modeling. and body surface area [BSA]) ; analyzing the association hazards model used to controllingreview. for hospital CABG In-hospital Hospital and Hospital volume based on total number Separate multivariate Logisticmeasures of disease severity 36,930 CABG, Coronary between surgeon volume andidentify predictors of survival, surgery volume and patient- 13 1998- US - New Operative Surgeon of off-pump and on-pump CABG regression models were (ie, clinical status, ejection patients ; Artery Bypass in-hospital mortality rate for and to estimate adjusted risk level risk factors. Also, no 1999 York State mortality rates. procedures performed. Surgeons were created for on-pump and off-fraction, history of previous 181 surgeons, Graft. off-pump and on-pump CABGof death. association between the off- divided into volume quartiles with equal pump CABG. myocardial infarction, history 33 hospitals. surgery. pump CABG surgery mortality numbers of surgeons in each quartile. These models included onlyof congestive heart failure, rate and the total number of For on-pump surgery, surgeons with patient-level data. Random-previous open heart surgery, both off-pump and on-pump very low case performed 52 procedures effects models were then and coronary anatomy) ; and CABG surgery cases (p 0.78). over the 2-year period, with low case constructed in order to modelcomorbidities (ie, COPD, However, in the on-pump 52 to 155, with medium case 156 to the effect of surgeon volumediabetes, renal failure, CABG surgery cohort, 273, and high volumes performed > on outcome, after adjusting peripheral vascular disease, 273 procedures. For off-pump surgery, for the effect of hospital and hypertension). surgeons performing a high surgeon case volume : Very low (< 5 volume and patient case volume of CABG procedures procedures) ; Low (5 -10 procedures) ; mix, using the macro PROC had significantly lower risk- Medium (11-31 procedures) ; High >31. GLIMMIX (SAS). adjusted mortality rates among their patients compared to those performing a low and medium volume of CABG procedures (p < 0.006). The absence of an association for off-pump surgery is surprising given the more technical nature of off-pump CABG surgery.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 72 -

Pays et/ This study re-emphasizes Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir the importance of case mix N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) adjustment in comparative d’étude The X² test. (1) General co-morbidity—forassessment of healthcare This study was designed to example, chest or spine quality. These results may Complications Surgeon Low volume surgeons defined as diseases that could prevent explain in part the trend UK 667 cases of Pharmaco- assess the influence of case 14 1996 - rates. surgeons performing fewer than 400 optimal positioning of the patient. emulsification - rates. mix on the volume-outcome previously demonstrated of 2001 pharmaco- cataracts per year) had a higher surgery. relation in (2) Significant ocular history—lower complication rates for emulsification ; average complication rate than the for example, previous attack pharmacoemulsification higher volume surgeons. 6 surgeons. high volume ones dened as surgeons fi surgery. The case of angle closure glaucoma or performing more than 400 cataracts complexity was assessed vitreoretinal surgery, problems per year. with the rst eye. using a potential difficulty fi score (PDS) devised from (3) Ocular co-morbidity— preoperative data predictive for example, corneal opacity, unstable lens. of potential surgical difficulty. The PDS was validated by Pseudoexfoliation has a retrospective analysis of a previously been sample of 100 cases. thought to be an independent risk factor even in the absence of obvious weak zonules or a small pupil.4 5 Similarly post-angle closure eyes were thought to be a potential risk even without short axial lengths or shallow anterior chambers. (4) Cataract density as recorded at the last preoperative visit before the operation. 15 1996 - UK at 2001 Sunderland Eye Correlation Analisys None. When the data were pooled Infirmary Using theatre logbooks and(Spearman’s RHO). for all the surgeons there was Surgeon rates The surgeons into ‘‘high’’ and ‘‘low’’ cross checking with the evidence that complication volume. High hospital database, the total rate decreased over time 6 consultant Cataract, Complication volume surgeons were designated asnumber of (Spearman’s RHO = 20.319, surgeons surgery using inrates those who performed pharmacoemulsification p = 0.058). If the data were (ophtalmologist) ;pharmaco- (a) Posterior more than 400 cases per year while lowprocedures performed pooled from all the years and 16,975 cases ; emulsification. capsule volume was taken as per surgeon per year was all the surgeons then there over 6 years. rupture with or fewer than 400 cases per year - calculated . Significant was strong evidence of a without anterior The volume of cataract surgery for eachintraoperative complications decrease in complication rate vitrectomy surgeon occurred. with an increase in the number (b) Loss of cases (Spearman’s RHO was calculated and stratified as of nuclear moderate (51–200), high = 20.63, p, 0.01). There are fragments into (201–399) or very high (400 cases per however some caveats in that the vitreous year). the issue of case mix was not (c) Sulcus addressed and that the results placed posterior are from a single unit and may chamber lens not necessarily be true for other (d) Anterior units. chamber lens implantation.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 73 - Demographic data( Age,Sex), Charlson’s comorbidity index, Univariate comparisons The impact of regionalization on A computer search of hospital Pays et/ Case volume of patient characteristics Inhospital death outcomes - (pre/post - regionalization).discharge abstracts coded Période Echantillon Intervention Indicateur hospitalIndicateur - Seuil utilisé pour définir were conducted using the N° ou région and 30-day for the Canadian InstituteMéthode for Variables contrôlées Résultats pre-/post- χ², t, and Mann–Whitney U étudiée de l‘étude étudiée readmissionde résultat volume haut/bas volume (n) Health Information. d’étude regionalization tests as appropriate. The rates ; length of The reliability and utility of CABG, status. positive skew of length of stay. administrative health care 2 hospitals - aortic valve stay (LOS) data, the natural 16 1993- Canada- data have been established. Post-regionalization, case 9,250 patients logarithm of LOS was used 1997 Alberta replacement, volume increased by 15%, (,.727 in all analyses. Variables capital repair of and 43.6% of patients used procedures) significant at the p < 0.05 number of urgent and emergent Edmonton abdominal some form of community- - 50% men, level were considered for cases. aneurysm, based health care services. 66.3 years age entry into a multiple linear femoropopliteal The median length of hospital of the post regression model predicting bypass, stay decreased from 8.0 days regionalization LOS. Adjusted regression mitral valve pre-regionalization to 7.0 group. coefficients relating post- replacement, days (p < 0.001). Overall and regionalization status to the colon resection for speci c procedures the logarithm length of stay for fi for cancer, death rate was unchanged each procedure. Adjusted colon resection (3.1% pre-regionalization, Odds Ratios relating post- for diverticular 2.4% postregionalization, p = regionalization status to disease, 0.06). The readmission rates 30-day readmission for each pancreatectomy were similar (8.0% versus procedure. for cancer, 7.0%). The consolidation of radical 16 major surgical procedures nephrectomy, had minimal impact on death radical and readmission rates even prostatectomy, though patients in the post- craniotomy regionalization group were for tumour, slightly older and had greater craniotomy comorbidity. The decline in for subdural the LOS occurred nationally hematoma, over the same period, and hip fracture a corresponding increase in procedures, total the use of community-based hip arthroplasty, services. total knee arthroplasty, thoracotomy for tumour.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 74 -

CABG surgery performed by Pays et/ surgeons with annual volumes Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir Age, gender, discharge status,of 125 in hospitals with annual N° ou région MéthodeThese multivariable analysespatients’ Variables clinical contrôlées risk factors Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) volumes of 600 was 0.84%, were conducted by using 3 and complications ; ventricularcompared with a prevalence of d’étude This study examines both methods—logistic regressionfunction (ejection fraction) ; a 1.35% for patients undergoing All annual hospital volume thresholdshospital and surgeon volume- Annual analysis, generalized recent myocardial infarction, CABG surgery performed by 17 1997 - New-York- N° of Hosp? CABG, CoronaryIn-hospital between 200 - 800 ; Surgeon volume mortality relations for CABG surgeons estimating equations (GEE),left main artery disease, surgeons with annual volumes 1999 US N° of Surgeon?artery bypass mortality rates ; thresholds between 50 -200 ; To test surgery through the use of and hospitals and hierarchical logistic compromised hemodynamic of <125 in hospitals with Number of graft. risk-adjusted the interaction effects of hospital anda population-based clinical volume rates. regression analysis. The laststate (shock, hemodynamic annual volumes of < 600 procedures mortality rates. surgeon volume, annual surgeon anddata set. two methods were used to instability, cardiopulmonary (P< 0.001). That both hospital 57,150 patients. hospital volume were each separatedThe bivariate relation account for clustering of theresuscitation), a large group ofand surgeon volume are still into 2 groups, and the resulting 4 between in-hospital CABG observations within hospitals.comorbidities (carotid disease,associated with lower risk- hospital volume/surgeon volume groupssurgery mortality rates renal failure, diabetes, previousadjusted mortality rates. The were then compared by calculating theand each of the patient Higher-volume operators/ stroke, aortoiliac disease, differential in mortality rates risk-adjusted mortality rate for each risk factors contained in hospitals experience lower risk- congestive heart failure, between hospitals was highest group. the registry was examined Demographics (Age, Sex), adjusted PCI outcome rates. previous PCIs, open heart for lower hospital volume by means of 2 tests after Infection fraction, patient The study suggest that volume- Three hospital-volume thresholds Administrative and clinical The patient risk factors, wereoperations, stents, hepatic thresholds. Hospital and creating categories for the preprocedural risk factors, based criteria be considered 3 outcomes : (400, 500, and 600 procedures per data. compared by use of X² tests,failure). Cardio/ Surgeon continuous variables of age complications of care, for referrals to higher-volume 18 1998 - US - 34 hospitals, - in-hospital year) . logistic regression analysis. Percutaneous Volume. and ejection fraction. discharge disposition, and hospitals until better data 2000 New York 263 operators ; mortality ; The regression coefficient of Coronary provider (hospital and become available for assessing n =107,713 - same-day this binary variable was then Interventions operator) ; cardiac quality. Assessment of relative patients. CABG surgery ; exponentiated to obtain the (PCI). function, coronary vessels quality of care could be based - same-stay adjusted odds ratio for the diseased, vessels diseased on risk-adjusted outcomes, not CABG surgery. adverse outcome occurring in the lower-volume group and attempted, previous openon much cruder criteria, such relative to adverse outcomeheart surgery, and previous as provider volume. However, occurring in the higher- myocardial infarctions ; type Cpopulation-based clinical data volume group .Adjusted oddslesion ; comorbidities (carotidare rare, for risk adjusting ratios. disease, renal failure, diabetes,PCI outcomes. previous stroke, aortoiliac disease, congestive heart failure, previous PCIs, open heart operations, and stents).

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 75 -

Pays et/ The results indicate that early- Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôléesneonatal Résultatsdeath in low-risk births Type of birth clinic : closed, étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Administrative Data from Several logistic regression in very small delivery units d’étude attending hospitals, the perinatal birth register ofmodels adjusting (500 births) is substantially Born units/ Hospital volume was categorized government hospitals, perinatal Hesse. for additional risk factors. increased when compared Neonatology andBirthweight- hopital according to the number of births 19 1990- Germany/ 582,655 centres. with low-risk births in large perinatology. specific mortality per year into very small (500 births/ Hesse infants born, volume. Size of birth clinic : (low-risk 1999 rates during year), small (501–1,000 births/year), delivery units (1,500 births). 67 differents birth); labour intermediate (1,001–1.500 births/year) Even small (501–1,000 births) items. Risk factors by type of birth and intermediate delivery units or within the first and large (1,500 births/year). Low-risk clinic (95% CI): Death (1,001–1,500 births) showed 7 days of life births were assumed for normal weight from congenital anomaly; a significantly increased early- (early-neonatal babies (using the traditional cut-off of Congenital anomaly ; Low death). 2,500 g birthweight), excluding those neonatal death rate compared birthweight (<2,500g) ; Pretermto large delivery units (1,500 infants with a documented congenital (< 37completed weeks); anomaly as a cause of death. All other births). This analysis presents Caesarean section . an urgent public policy issue births were classified as non-low risk. All births : Size of birth clinic of whether such elevated (births/year); Birthweight- risk in smaller delivery units specific subgroups. is acceptable or if further consolidation of birthing units should be considered to reduce early-neonatal mortality. Reorganization of obstetric care should be discussed to reduce early-neonatal death rates.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 76 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir In a logistic regression, PTCA N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Descriptive statistics and Patient characteristics procedures significantly d’étude Patient data were obtained logistic regression model. by hospital volume and predicted in-hospital mortality Hospitals The relation between volume and from the OSHPD discharge period : Age, sex ; AMI, and and bypass surgery rates in all In-hospital Cardio/PTCA, volume. outcomes was analyzed over 3 time data, Version A, from 1984 to multivessel PTCA ; Length of 3 time periods. 20 1984- US 353,488 bypass surgery percutaneous periods : 1984 to 1987, 1988 to 1992,1996. The OSHPD database stay ; comorbidity index with However, coef cient estimates 1996 California patients ; (CABG) death fi transluminal and 1993 to 1996. Tables comparing provides standardized Charlson comorbidity indicate that improvements 3 periods. rates and in- coronary outcomes were generated for patientsinformation from hospital index (diabetes, renal over time in outcomes for hospital mortality angioplasty. treated in hospitals with low volume discharge abstracts for all insufficiency, and peripheral hospitals performing, 200 rate. (200 procedures per year), intermediatepatients admitted to California vascular disease). procedures were comparable volume (200 to 400 procedures per hospitals. to the predicted benefits of year), and high volume increasing volume above 400 (400 procedures per year). procedures within time periods. Over time, the disparity in outcomes between low- and high-volume hospitals has narrowed, and outcomes have improved significantly for all hospitals. Given these improvements, lower minimum volume standards may be advisable in less populated areas, where the alternative is no angioplasty at all. Although the volume-outcome relation persists over time, in-hospital mortality and CABG rates have decreased for all hospital sizes. Moreover, the relative performance advantage of high- versus low-volume hospitals has decreased overtime. These results have important implications for decisions regarding minimum volume standards and regionalization of innovative technologies. The relative benefit of more favorable outcomes at larger facilities must be weighted against the potential decline in access resulting from minimum volume standards or regionalization.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 77 -

Pays et/ After adjusting for patient Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôléescaracteristics, Résultats low volume étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Logistic regression models. Patient’s age, sex, race, physicians were associated d’étude Data come from Part discharge status including with higher rates of bypass Outcomes were generated for low- Hospital and A (hospital) and part B death, up to 10 discharge surgery (P<.001) and low Cardio/CoronaryIn-hospital volume (<25 annual procedures), 21 1992 US 6,115 Physician (physician, outpatient) claims diagnoses and 6 procedures volume hospitals were Angioplasty. bypass surgery medium-volume (25 to 50 annual physicians Angioplasty containing demographic and identi ed by International associated with higher rates of (CABG) death procedures), and high-volume (>50 fi 97,478 Volume. limited clinical information on Classi cation of Diseases (ICD)bypass surgery and death rates rates and in- annual procedures) physicians after fi Medicare all inpatient hospitalizations codes, and attending physicianof bypass surgery (P.001). This hospital mortality stratifying. that were billed to Medicare, study suggests that adherence patients at 984 rate. according to Unique Physician hospitals. including the patient’s Health Identification Number. to minimum volume standards Insurance Claims (HIC) Clinical information from by physicians and hospitals number. physician and outpatient bills willto lead to better outcomes Medicare : service or procedurefor elderly patients undergoing provided according to CPT coronary angioplasty. Improved codes, AMI, number of vesselsoutcomes up to treshold of 75 undergoing angioplasty, Medicare cases per physician and hospital characteristics and 200 cases per hospital. including hospital volume and Nationwide Inpatient Sample Hospital Hospital volumes differed significantly medical school affiliation and Pancreaticodu- In-hospital database from 1988 to 1995, volume among the three hospital types comorbidity according to the 22 1988- US, Patients older odenectomy. mortality containing 24,926 hospital status.(P .0001), Charlson index. 19 states than 65 years (procedure patients undergoing 1995 both in an overall way and pairwise in US of age, specific) pancreatectomy for (P .0001). 720 hospitals periampullary cancer. One 24,926 of the databases of the procedures Healthcare Cost and (25,000 Utilization Project (HCUP), Regression results show that Patients). a standardized, multistate Spearman rank correlation Type of hospital (rural, urban sex (male), age, emergency health data system. analysis with hospital as thenonteaching, urban teaching),admission and hospital volume unit of analysis. Comparisonspatient sex, age (50, 50 – 59,were significantly predictive of of hospital volume for the 60 – 69, 70 – 79, and 80 hospital mortality. In-hospital three hospital types in an years), year of admission (1988mortality in the low-volume overall manner with the – 1989, 1990 – 1992, 1993 – hospital setting is prohibitive, Kruskal-Wallis test, and 1995), and admission status and review of each institution’s pairwise comparisons among(elective/missing or urgent/ mortality rates must occur hospital types with a Wilcoxonemergency. before these procedures are rank sum test. performed in those To evaluate patient and institutions. In addition, patients hospital factors that may over the age of 60 years, influence mortality, a multiple male patients, and those with logistic regression model used, an urgent admission are at a accounting for the correlation signi cant risk of in-hospital structure of patients from the fi same hospital having similar death, and consideration outcomes using methods should be given toward transfer of generalized estimating to an experienced institution. equations.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 78 -

Major liver resection can be Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir performed safely with low rates N° ou région Méthode Variables contrôléesof morbidity Résultats and operative étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Age, sex, type of hepatic resections (hepatic lobectomiesmortality with careful selection d’étude Very small sample. Half theObjective of the study was to evaluate the experience (n = 15), of patients at a low-volume Procedures performed were hepatic patients (23) were operated Complications Hospital and of a low-volume community-trisegmentectomies (n = 5), community-based teaching 23 1992- US 14 surgeons, Hepatic lobectomies (n = 15 ; right, n = 11 ; left,by 1 sugeon, the rest by 13 rates, 30-day surgeons based teaching hospital segmentectomies (n = 16), andhospital. 2002 23 operations, Resection n = 4), trisegmentectomies (n = 5 ; right,different ones. postoperative volume. with a surgical residency wedge resections (n = 10)); 46 patients. (Liver). n = 3 ; left, n = 2), segmentectomies (n Mortality. training program. This was Indications for hepatic = 16 ; left lateral, n = 12 ; right posterior, a retrospective review of theresection (liver metastases n = 4), and wedge resections (n = 10). patients medical records. from colorectal cancer); Low volume varied between Outcomes were limited to comorbidities (coronary artery <1 to <22 cases per year, with most in those in the perioperative disease and hypertension, the range of 1 to 5 cases per year ; low, period. Diabetes, Chronic obstructive medium -, and high-volume hospitals pulmonary disease, Congestive for pancreatic resection were defined as heart failure) ; Coronary <3, 3 to 13, and >13 per year. artery disease was defined by the clinical signs of angina pectoris, a history of myocardial infarction, angiographic, evidence of ischemia on prior noninvasive studies,or peripheral vascular disease.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 79 -

According to prospectively Pays et/ Baselines characteristics (Age,recorded data from a Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir sex, race), performance large clinical trial, patients N° ou région Méthode Variables contrôléeswhose colon Résultats cancer was étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) status, stage of disease, and d’étude Nested cohort study within aAmong hospitals that were peritoneal implants, tumor resected at low-volume hospitals experienced a higher Hospital Defined tertiles of hospital procedurerandomized clinical trial. jointly represented in our differentiation was included Mortality : volume (low [46 cases per year, based cohort study and the NIS in the models ; number of risk for long-term mortality ; 24 1988- US- 3,161 patients Resection of procedures, on Medicare volume], medium however, this increased 11 states 1,078 hospitalsColon cancer. 5-years overall Medicare case database (255 hospitals), thelymph nodes ; Baseline ECOG, 1992 mortality was not attributable to survival rates ; volume [47 to 84 cases per year], and high Spearman rank correlation rates of bowel obstruction Cancer [85 cases per year]) on the basis of coefficient for annual or perforation at the time of differences in colon cancer recurrence : Medicare procedure volume of the colorectal cancer surgery surgery, tumor location withinrecurrences. In this trial of 5-year hospital where study participants had volume, as measured by thethe bowel, and chemotherapychemotherapy patients with recurrence-free surgery. Medicare claims database assignment. high-risk stage II and stage III survival rates. and the NIS database. colon cancer, overall The distribution of baseline survival after curative surgery characteristics across was significantly improved hospital procedure volume among patients who underwent tertiles was evaluated by surgery at a hospital with a using Chi-square tests for high volume of colon cancer categorical variables and surgeries. 16% increase in Wilcoxon rank-sum tests overall mortality observed for continuous variables. among patients Overall and recurrence- resected at low-volume free survival rates were centers, but there was no examined by using the significant Analysis of HES data con rms Kaplan–Meier method and relation between hospital fi an inverse relationship between differences among tertiles The characteristics of patientsprocedure volume and the risk This produced a scatter hospital caseload and mortality All cystectomies for urologicalwere assessed by using treated : age (66.3 v 65.7 for cancer recurrence or plot with a parabolic for radical cystectomy. Applying cancer in the log-rank test. The entireyears) and gender (68.3% cancer-specific mortality. Hospital’s This threshold would appear to be distribution. Using logarithmic 95% con dence intervals, Radical 30-day in- England over 5 years were cohort was analyzed by usingmale vs 66.4% male) and the fi 25 Over 5 England 6,317 patients annual case reached at a case volume of 5 – 15 transformation, there was a the minimum caseload cystectomy hospital mortality analysed from Cox proportional hazards significant difference in MR years : 210 Centres, volume. cystectomies/year. At a caseload significant inverse correlation required to achieve optimum urological. rate (MR) ; Hospital Episode Statistics regression. above and below the threshold 50 operations/ threshold of 6 cystectomies/year. 8 – (−0.733, P < 0.01) between 1998- Hospital Stay. is confirmed (4.58% vs 8.11%,outcomes was 11 procedures/ year. 10 cases/year as possible threshold (HES) data. 2003 the variables of operative P<0.05). The duration of year. Increasing the caseload values. volume and MR. hospital stay between the beyond this minimum did not High-volume hospitals҆ (11 case/year) groups (21.9 vs 21.7 days) produce a significant reduction to those in low volume not a satisfactory variable for in MR. Thisfi nding would hospitals (< 11 case/year). measuring outcome. appear to support guidelines The case volume threshold to be suggesting that complex as high as 16 case/year includes procedures should take place more centres with optimum mortality in high-volume centres. outcomes.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 80 - Greater PPCI specialization was associated with a lower Patient and hospital relative risk of in-hospital characteristics : Demographicsmortality in patients treated Pays et/ Multivariable hierarchical Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir (age, sex, race), Medical with PPCI (adjusted relative N° ou région Data from the NationalMéthode logistic-regression modelshistory, Variables comorbidities, contrôlées prior risk comparing Résultats the highest and étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) examined the independent myocardial infarction or d’étude Patient and Hospitals caracterize into 4 quartiles Registry of Myocardial lowest quartiles, 0.64 ; P 0.006) relation between PPCI revascularization, chest pain, In-hospital hospital level ;on the basis of the relative proportionInfarction-4. of The National but not in those treated with specialization and the systolic blood pressure, heart mortality, PPCI cases atreperfusion-treated Registry of Myocardial fibrinolytic therapy. Compared 26 2000- US 37,233 patientsCardio/ likelihood of in-hospital Treatment times patients treated with PPCI : 34.0%, 34.0Infarction (NRMI) is an rate, Killip classification, with patients at hospitals in 2002 463 hospitals. ST-segment hospitals. to 62.5%, 62.5 to 88.5%, and 88.5%.ongoing ; observational mortality and delayed performance of a prehospital the lowest quartile of PPCI elevation In STEMI (time PPCI volume (30, 30 to 70, 70) and study of acute myocardial reperfusion, after adjusting ECG, ECG findings, time specialization, adjusted door- myocardial to reperfusion) estimates for hospital volume based oninfarction patients admitted forto patient and hospital from symptom onset to to-balloon times in the highest infarction - door to balloon the number of STEMI cases per yearparticipating hospitals in thecharacteristics, including hospital arrival, time and day quartile were significantly (STEMI), treated times or (50, 50 to 150, 150). United States. We included PPCI volume. of presentation (weekdays shorter (99.6 versus 118.3 with primary - door to needle patients in the NRMI-4 cohortMultivariable hierarchical between 8 AM and 6 PM minutes ; P 0.001), and the percutaneous times. admitted between Jan. 2000linear-regression models or other), hospital location likelihood of door-to-balloon coronary and Dec. 2002. evaluated the independent and teaching status, and the times exceeding 90 minutes intervention relation between PPCI presence of on-site cardiac was significantly lower (relative (PPCI) and specialization and door- surgery. risk, 0.78 ; P 0.001). Adjusting brinolytic fi to-balloon times and door-to- for PPCI specialization therapy. needle times. diminished the association between PPCI volume and clinical outcomes. Greater specialization with PPCI is associated with lower in- hospital mortality and shorter door-to-balloon times in STEMI patients treated with PPCI.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 81 - This study suggests that the positive relationship between hospital volume and 5-year Pays et/ survival for stomach cancer Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir * The number of hospitals and N° ou région Méthode Variables contrôléespatients observed Résultats in Osaka étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) number and range of Observed 5-year survival was Prefecture during 1975–79 had d’étude stomach cancer surgical Using data in the OCR’s calculated by the Kaplan- decreased in the two decades procedures per hospital, by Stomach cancer. 5-years survicalHospital The study was divided into four periodsdata ; processing for cancer Meier method. Survival of the study, and disappeared 27 1975- Japan- 28,608 patients statistics, primary facilities study period and category of rates. surgical 1975–79, 1980–84, 1985–89 and 1990– differences were analyzed in 1990–94except for the Osaka aged 35-79 hospital volume; 1994 volume. 94, according to year of diagnosis. for treatment of each cancerby Cox’s proportional years old ; * mean age of analyzed very-low-volume hospitals. All hospitals in Osaka which reportedwere at determined and codedhazards model adjusting study divided patients at diagnosis Stomach cancer patients least one case with surgical treatmentaccording to the following for age at diagnosis and into 4 periods. according to the hospital treated in these hospitals for stomach cancer were divided intocategories. sex. The statistical package volume and the study period;remained at significantly 4 categories of hospital volume with software STATA was used * the observed 5-year survivalhigher risk of death than in almost equal sizes : high, medium, low, for statistical analysis. of the cases whose cancer the high-volume hospitals. In and very low, according to the total Statistical significance and the case of the distant group, was confined to the original number of surgical procedures in each 95% con dence intervals fi organ (localized), together there was no clear relationship study period. (CI) of hazard rate ratio were with their age- and sex- between hospital volume and obtained and judged by using adjusted hazard rate ratios 5-year survival during the study a 2-sided test. compared by hospital periods. volume; * the results for the cases whose cancer had spread to the regional lymph nodes; * the 5-year survival for the cases whose cancer had spread to immediately adjacent tissue. Differences in survival rates between the high-, medium- and low- volume groups were not seen in the period 1975–79; * the results for the cases whose cancer had metastasized to distant organs.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 82 -

Substantialclustering was Pays et/ Clinical and demographic observed in the analyses Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode variablesVariables were contrôlées used to adjustinvolving Résultatsmorbidity and points. étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Volume–outcome trends werefor differences in patient caseThe 2 statistical techniques d’étude analyzed by a conventional mix among individual surgeons.produced noticeably different Reanalysis of 3 previously method (logistic regression)These included age ; sex ; results in some analyses. Volume was modeled as a continuous 28 1992- US- Medicare * Colon Cancer,For colon cancer : Surgeon published volume–outcomeand corrected for clustering.ethnicity ; disease stage at The presence of clustering * 2-Year variable in all analyses. Odds ratios are 1996 Carolina/ beneficiaries * Prostate rates/ studies. All 3 data sets Two widely used statistical diagnosis ; disease grade ; represents variations in reported per 100- unit decline in volume Texas 65 years of age; Cancer, mortality rates, Hospital were obtained by using the methods for analyzing income level ; and the presenceoutcomes among providers increments for the colon and prostate Patients : 3 * Rectal Cancer.* Ostomy. volume. linkage of the Surveillance, clustered data, a random- of particular features such with similar volumes. Thus, cancer analyses and per 10-unit decline cohorts ranged : For Prostate Epidemiology, and End effects model (in which the as emergent presentation, in volume–outcome studies, for the rectal cancer analysis. * 24,166 cancer : Results registries with impact of each surgeon obstruction, or perforation. the degree of clustering colectomies * Postoperative Medicare claims from the is modeled explicitly) and of outcomes should be performed by complications, Centers for Medicare & generalized estimating characterized because it may 2,682 * Late urinary Medicaid Services. equations GEE (in which the provide insight into variations in surgeons, complications. Evaluated the effect underlying patient-specific quality of care. * 10,737 For Rectal Cancer : of surgeon volume on analysis is adjusted to Since the impact of clustering prostatectomies outcomes in 3 settings : accommodate the effect of typically attenuates the strength performed by adjustment solely for case clustering on the statistical of the statistical signi cance, 999 * 2-Year mortality. fi mix ; adjustment for case signi cance of the analyses), volume-outcome trends are surgeons, *Abdominoperineal fi * 2,603 rectal resection. mix and hospital volume ; were used and compared, frequently modest. Large resections and adjustment for case and the degree of clustering sample sizes are needed performed by mix, hospital volume, and was presented graphically. to establish many of these 1,141 surgeons. clustering. trends in a convincing way. However, volume–outcome relationships in many studies are overwhelmingly consistent in demonstrating a positive association between hospital volume and outcomes.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 83 - Hospital volume was a strong predictor of bioprosthetic Patients’ state of residence ; valve use in older patients Pays et/ undergoing AVR. The lower Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir primary and secondary N° ou région Méthode diagnosesVariables and contrôlées procedures. use of bioprosthetic Résultats valves étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Bivariate associations in low-volume hospitals is at d’étude age (in 5 groupes : 70 to 74, The risk-adjusted associationbetween valve type and 75 to 79, 80 to 84, 85 to 89, odds with recent guidelines Hospital Volume was represented by 9 indicatorbetween volume and the recommending bioprosthetic Aortic Valve Quality : demographic variables and 90 years, with reference of 29 1999- US 80,470 older volume. variables representing patients use of bioprosthetic valves valves in patients aged 65 Replacement. Valve type primary and secondary 65 to 69 years), Race : black 50 States patients aged undergoing valve replacements in 2001 selection (use was determined by adding diagnosis and procedure or other nonwhite ; year of years. Thesefi ndings further >= 65 years hospitals in the 2nd through 10th of bioprosthetic indicators for hospital codes that represented AVR surgery ; gender ; anginasupport the use of volume as a 1,045 hospitals. deciles of increasing hospital volume. valves). volume to the multivariable potential patient risk factorspectoris ; and several comorbidmarker of hospital quality. Annual valve surgery volume for the 10 risk adjustment model that were determined with the useconditions (peripheral vascular deciles were 20, 30, 41, 51, 65, 82, 103, included patient-specific riskof the Wilcoxon test or the X²disease, nonmetastatic cancer, 135, 181, and 573. factors. statistic. Significant variablescoagulopathy, peptic ulcer, (P 0.05) in bivariate analysesand primary diagnosis of were entered into a stepwiseAMI). Diabetes was defined logistic regression. The by ICD-9-CM codes ; Surgical bivariate association betweenpriority : emergent and urgent hospital volume deciles andadmissions, relative to elective the use of bioprosthetic admissions. Admission source : valves was determined by thetransferred from another 2 tests for trend. Bivariate facility or admitted through associations between emergency. volume deciles and patient Increased hospital procedure characteristics that were Patient demographics (age, volume correlated with Results were robust to Logistic regression and improved longterm survival Hospital Procedure volume cut-off points/3 years identified as multivariable sex, postal code or place of In-hospital death changing volume groups. proportional hazards models for patients who underwent Cancer surgery/ procedure Hospital volume Characteristic : predictors of valve type wereresidence), major diagnoses rate ; long-term were used to consider the resections of breast, lung breast, colon, volume. Low ;Low–medium ; Medium–high; High also examined by ANOVA and procedures, and outcomes survival rate. clustering of data at the and liver. No significant HR 30 1990- Canada - Number of lung, esophagus Breast : ҅102; 103̄158; 159̄264; for continuous variables andsuch as discharge status (dead patients and and liver hospital level and to assessor alive) ; hospital teaching for operative death for patients 2000 Ontario ҆265. the 2 tests for linear trend for hospital per lobectomy. operative death and long- status (yes or no), various who underwent colon, liver, Colon : ҅61; 62̄90; 91̄137;҆138 categorical variables. procedures : term survival. patient characteristics includinglung and esophageal cancer Lung : ҅32; 33̄85; 86̄130;҆131 Breast age, sex, comorbidity score, resection. Hospital teaching Esophagus ҅: 7; 8–19; 20–43;҆ 44 (n = 14,346 ; place of residence (rural v. status had no impact on patient N = 152) Liver : ҅23; ҆24. urban) and socioeconomic outcomes. Colon status (highincome, medium- (n = 8,398 ; income or lowincome level). N = 151) Lung (n = 2,698 ; N = 67) Esophagus (n=629; N = 68) Liver (n = 362 ; N = 41).

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 84 -

Pays et/ Women who had procedures Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode PatientVariables demographic contrôlées at low volume Résultats hospitals had étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Univariate and bivariate The effect of hospital volume information (age, race), and almost a 3-fold higher risk of d’étude analyses were performed death and were more likely to Three volume groups were de ned byon outcomes using data from hospital characteristics : year, Hospital and fi using Chi-square and simple have a nonroutine discharge, In-hospital mean annual hospital volume : into 3 the Nationwide Inpatient median area income, Payer 31 1998- US 310,759 Urogynecologic surgeon linear regression analysis. compared to those at high- mortality ; groups of equal size (tertiles). Low- Sample (NIS). status (Medicare, Private, 2003 women; surgery (pelvic procedures volume hospitals. Women who perioperative volume hospitals (LVH) performed onRetrospective cohort study of Medicaid, Other), number of 2,986 reconstruction). volumes. had procedures by low-volume complications average <92 cases per year, medium-women. procedures per admission, hospitals. surgeons were also more likely and nonroutine volume hospitals (MVH) performed 92 hospital teaching status, to suffer complications and discharge. to 185 cases, and high-volume hospitals Hospital in urban setting, have non-routine discharges (HVH) performed >185 cases per year. location, and region ; compared to those with high- Surgeon volumes : low-volume (LVS) comorbidity index based volume surgeons. performed on average <8 cases per on the Dartmouth-Manitoba algorithm. 32 1994- Canada, N° patient year, medium (MVS) performed 8 to 18 31,632/ 1999 Ontario- cases, and high (HVS) performed >18 With the exception of colorectal N° Hosp Canada cases per year. Patient level characteristics resection, 30 day mortality Oesopha : 613 ; Odds ratio and Linear modelsof age, sex, and comorbidity is inversaly related not only 47 Col. res : Administrative data. for binary outcomes to using Charlson score with to the hospital volume of the 18,898 ; Calculated the average hospital volumeElectronic records to identifyestimate associations Median (interquartile range), same procedure, but also to 134 Pancrea : 30 days Hospital of each procedure on the basis of thehospital separations in between hospital volume No of patients, No of hospitals,the hospital volume of most 686 ; Oesophagectomy,operative volume number of identical procedures done Ontario between 1994 to and 30 day mortality. To Average annual hospital other procedures. We found 49 Lung res : Colorectal mortality. procedures. at the hospital over the ve year study1999. Linked to a database accountof for the effects of volume (Median, range), Meanlittle evidence of an association 5,156 ; resection, fi period. Divided into two equal groupsvital statistics; ve separateclustering at the hospital level(SD) age in years. between volume and outcome 54 Rep of AA : Pancreatico- fi of volume categories (high volume cohorts of patients (one for in the adjusted analyses, for colorectal resection. 6,279 ; 57. duodenectomy, hospitals and low volume hospitals) ateach surgical procedure) we estimated the model For some complex surgical Lung resection, the median average annual hospital and did ve volume-outcomeparameters and standard procedures, operative mortality Repair of fi volume, such that patients having a analyses for each of the veerrors by using generalised is lower not only in hospitals abdominal fi surgical procedure. procedure based cohorts. estimating equations. that do a high volume of the aortic aneurysm same procedure but also in (AAA). hospitals that do a high volume of different procedures.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 85 - First, an inverse relation exists between physician primary Pays et/ Demographic (age and sex), angioplasty experience and Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° of patients, medical historyin-hospital mortality. N° ou région MéthodeUsing stepwise multiple Variables contrôlées Résultats (smoking, the prevalence Second, there is a strong trend étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) logistic regression models. d’étude Data set represents the of diabetes, hypertension, toward a relation between Categorical variables were Physicians and hospitals were experience across a state peripheral vascular disease hospital primary angioplasty In-hospital Physicians compared by X 2 analysis, categorized into terciles based on the(New York) and includes PVD, stroke) ; Cardiac volume and mortality such that 33 1995 US, New 151 Physicians,PTCA performedmortality rates. and hospitals and continuous variables volume of primary angioplasty patients treated in rural, history (prior MI, prior cardiac18 lives per 1,000 patients -York 1,342 Primary as primary procedures were compared by Student’s procedures performed in 1995. urban, teaching, nonteaching, heart surgery, prior PTCA, treated would be saved by US angioplasty reperfusion volumes. t test. Significance of trends The small number of patients in the not-for-prorst fit and for-profit history of congestive heart treatment in a high-volume procedures, therapy for acute fi for increasing hospital and tercile of both physicians and hospitals,hospitals. The responsible failure) ; preprocedural hospital. Finally, an interaction 32 Hospitals, myocardial physician volume categories the rst and second terciles in each physician, completes a shock or hemodynamic between hospital and physician procedures infarction fi was assessed with the questionnaire containing instability (the requirement forprimary angioplasty volume within 23 hours(AMI) (primary category were combined and labeled Cochran-Armitage test. information on patient/ pharmacological or mechanicaland in-hospital mortality angioplasty). as low volume, with the highest tercile All probability values are of onset of AMI, operation caracteristics after support to maintain blood exists such that those AMI defined as high volume, resulting in 2-tailed. each intervention. pressure or cardiac output), patients treated in high-volume 22,000 patients. the following 4 groups : low-volume Unadjusted and Adjusted and time to treatment (< 6 or hospitals҆ by high-volume physicians, high-volume physicians, Relative Risk of In-Hospital 6 hours). physicians have a 49% lower low-volume hospitals, and high-volume Death. hospitals. in-hospital mortality rate than those treated by low-volume physicians in low-volume hospitals.

34 2000 US, - 2001 New-York -2002 and

Findings indicate that higher Patient age, sex, ethnicity, disease-specific physician medical insurance status volume is associated with lower The unit of analysis was theComparisons of patient characteristics across (Private or Public, Medicare, risks of in-hospital mortality in Physician volume was categorized intopatient. All adjusted models Risk of in- Physician categories no insurance, or unknown), patients with SLE. 15,509 patientsSystemic lupus 3 groups : included all covariates hospital mortality.experience of physician volume were presence or absence of The association was most aged 18 years erythematosus 1 hospitalization of a patient with SLEsimultaneously. and hospital made using analysis of nephritis, SLE Comorbidity evident among patients with or older (SLE) per year, 1–3 hospitalizations volume. variance or rank sum tests forIndex, and type of admission nephritis, the 40,668 **a rare and per year, or 3 hospitalizations per year. Pennsylvania continuous variables, and chi-(emergency versus urgent) ; subgroup with the highest hospitalizations.uncommon These cut points were based on whole US square tests for categorical hospital volume and bed mortality risks. potentially fatal numbers and accommodated the variables. The association ofsize, rural or urban location, immune disease. positively skewed distribution of the physician volume with in- teaching status, and year of average physician volumes, with 46%, hospital mortality. hospitalization. 42%, and 12% of patients in these categories.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 86 -

For both low-risk and moderate-to-high-risk patients, Pays et/ Patients age, gender, higher provider volume is Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode VentricularVariables function contrôlées ; Previousassociated Résultats with lower risk of étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) The interaction effects myocardial Infarction (betweendeath. d’étude A logistic model was of hospital volume and < 6 hours to ҆8 days) ; Left 30-day in- Hospital and developed. Many risk factorssurgeon volume was the main coronary artery disease, 57,150 patientsCABG, Coronary 35 1997- US, hospital mortalitysurgeon Annual hospital volume thresholds forin this model were also reference group in the logistichemodynamic state (unstable, 192 surgeons Artery Bypass 1999 New-York rate. isolated CABG surgery (between 200signi ficant predictors in the regression models, and shock, cardio-resuscitation), 33 hospitals. Graft . US annual volumeand 600 cases, in units of 100 cases).EuroSCORE cardiac surgeryadjusted ORs were calculatedComorbities (Stroke, for the other 3 groups. In Aortoiliac disease, Malignant cases. Four surgeon/hospital volume groupsmodel and the Society of were created by separating each typeThoracic Surgeons (STS) addition to conventional ventricular arrhythmia, Chronic of volume into 2 groups. The group withCABG model11 for 30-day logistic regression analyses,obstructive pulmonary disease, low surgeon volume (125 cases) andmortality. This model was generalized estimating Extensively calcified ascending low hospital volume (600 cases). used to predict each patient’sequations and hierarchical aorta, Diabetes, Hepatic failure, A predicted probability of death of probability of death, which logistic regression were Renal failure, Renal failure 2% was chosen as a cutoff point to represents surgical risk. also used to account for therequiring dialysis), Previous clustering of patients within open heart operations, Number separate patients into low-risk (2%) An inverse relationship and moderate-to-high-risk groups. surgeons and hospitals in Patientsof categorical age, sex,risk factors LOS, For statistical analysis, the between hospital PTCA and Thehospital annual volume hospital thresholds PTCA volumebetween is 200Used of the Office of the analyses that examineddischargesquared. status (including Annual continuity (Yates’) correction stent volume and adverse Cardio/Coronary In-hospital dichotomizedand 600, for both at 2 risklevels, groups, 200 andpatients 400Statewide Health Planning the joint impact of hospital death), and hospital charges 1989 : 24,883 hospital for X² analysis, Fisher’s patient outcomes (less risk of 36 1989 et US, Angioplasty. mortality after PTCAsin higher-volume per year. hospitals consistentlyand Development (OSHPD)and surgeon volume on for each admission ; diabetics, PTCAs in 110 PTCA volume exact tests, and 2-tailed t dying or needing to undergo 1995 California - PTCA in 1989 had a lower risk of death than those indata set. Administrative outcomes. Coronary intervention (patients hospitals and stent tests were performed as emergency CABG). Resource US and in 1995. lower-volume hospitals, as evidenceddatabase. undergoing single vessel PTCA 1995 : 37,118 placement. appropriate. utilization was also significantly by adjusted ORs<1. with thrombolysis), principal PTCAs in 146 reduced in these high volume diagnosis (AMI). hopsitals. hospitals. 37 1995 or 1996

Hospital-level factors were not Patient age, sex, race, primaryindependent arthritis diagnosis (Rheumatoid,predictors of the association Used a Chi-square test Aseptic necrosis, Degenerativebetween hospital Data were derived for trend to assess the arthritis, Paget’s disease), volume and orthopedic adverse 6 mutually exclusive hospital–surgeon- US, 5,211 Hip replacement. Perioperative Volume from several sources, relationship between the comorbid medical conditions,events. The volume of volume groups, pairing each of the 3 Colorado, Medicare morbidity and of THRs including Medicare Part hospital-level factors, as and eligibility for Medicaid : low-THRs performed byndividual i hospital-volume strata (1–25, 26–100, patients mortality rates performed A and Part B claims, the well as patient-level and income status). A comorbiditysurgeons is the most important Pennsylvania, and over 100 THRs annually) with each 167 hospitals. and predictors at hospital, American Board of Medical surgeon-level factors, and index was calculated from determinant of orthopedic Ohio- of the 2 surgeon-volume strata (1–10 of postoperative and surgeon Specialties, a hospital the rate of adverse events. Medicare Part A claims basedcomplications and US and over 10 THRs annually). complications. volume. survey regarding institution-All potential predictor on Charlson score ; Year of should be considered in efforts 90-day rate of specific characteristics, variables associated with theTHR. teaching hospital. to improve THR outcomes. orthopedic and the American Hospital rate of adverse orthopedic adverse events, Association 1995 Annual events were introduced to defined as Survey. multivariate logistic models to dislocation of the assess whether the inverse relationship between hospital volume and adverse events prosthetic hip and was confounded by hospital- deep wound level factors and/or surgeon infection of the hip. variables.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 87 -

The study supports the Pays et/ hypothesis that there is a Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir Hospital and patient N° ou région Méthode characteristicsVariables contrôlées (age, sex), positive relation Résultats between The X² test or analysis étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) A population-based patient’s discharge status surgical volume and patient d’étude of variance was used to Hospital The distribution of surgical volumes forretrospective analysis. (dead or alive), length of outcome after pancreatic determine the statistical Pancreatic Postoperative Volume of each hospital and defined total hospital stay from date of resection for neoplasm ; Canada, 842 signi cance of the 38 1988- resection death procedures. procedure volume as low fi admission, comorbid conditionsthat is, patients treated in Ontario - procedures, differences. 1989 and for neoplasm. complication : (fewer than 22 procedures over the by Charlson Score, admissioninstitutions with a higher Canada 68 Hospitals. Logistic regression was used 1994- in-hospital case period or fewer than 3 status (elective or urgent/ volume of procedures have to assess relations between 1995 fatality rate and procedures/year), medium (22–42 emergent), hospital affiliationbetter outcomes, in terms of surgical volume and mortality readmission data procedures over the period (teaching or nonteaching postoperative death and mean rate, and linear regression mean length of or 3–6 procedures/year) or high (more institution) and hospital size length of stay, than those was used to assess relations stay in hospital. than 42 procedures (number of beds). treated in institutions with a over the period or more than 6 between surgical volume lower volume of procedures. procedures/year). and length of stay. The linear regression models considered only patients who were discharged alive.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 88 - First, in the modern age of stenting, there is no relationship between Patients age, sex, pre- hospital PTCA volume and Pays et/ procedural clinical status in-hospital mortalityin non- Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir Univariate analysis carried (silent, stable or unstable emergency procedures (in N° ou région Clinical status at dischargeMéthode Variables contrôlées Résultats out to identify the risk factorsangina, pre-operative low-risk patients). In contrast, étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) (dead or alive) was also d’étude Hospital PTCAVolume thresholds used in the analysis associated with in-hospital assessment for cardiac in emergency procedures In-hospital recorded in another database volume. were those recommended by the mortality. The associations valvular surgery, AMI, (PMSI). A cross check performed in patients with Cardio/primary mortality French Society of Cardiology : between categorical variablescardiogenic shock, OHCA, 39 2001- France 44 Hospitals AMI, cardiogenic shock, or percutaneous and overall was performed using this were assessed using the x2and other reasons), reasons 2002 (IdF) 37,848 < 250 procedures/year : insufficient to successfully resuscitated transluminal in-hospital database to validate the test and differences betweenfor performing the procedure Paris Area patients. ensure patient safety) OHCA, a clear inverse complications deaths recorded in the continuous variables were in case of silent or stable angioplasty 250–400 procedures/year : relationship exists. Second, rates. CARDIO-ARIF database. compared by Student’s ischaemia, ST- or T-wave (PTCA) after considered insufficient, but could work. the overall complication rate is t-test. Multivariable backwardchanges, and/or creatine emergency 400–600 procedures/year : higher in low-volume centres in logistic-regression models kinase (CK) or troponine procedures for Adequate activity volume both planned and emergency were used to identify elevation if unstable angina, AMI, cardiogenic > 600 procedures per year : Sufficient procedures. oleranceT of shock, or OHCA determinants of in-hospital myocardial infarction, or activity. Level of patient low-volume thresholds for due to acute mortality. Calibration OHCA) procedural info. (no safety is deemed optimal. angioplasty centres with the coranary artery was evaluated using the of lesions dilated, amount of A threshold of 400 procedures per year purpose of providing primary occlusion. Hosmer–Lemeshow test andstents implanted), pre- and used to define low (<400) and high PTCA in AMI should not discrimination was based post-procedural complications (equal to or >400) volume centres. be recommended, even in on the area under the ROC (AMI, repeat PTCA, emergency underserved areas. curve. CABG, stroke, vascular complications requiring surgery, There was a negative renal failure requiring dialysis, Patients’ sex and age, LOS association between surgical blood transfusion, death). Used X² test to analyze thein intensive care, underlying volume and the incidence Status at discharge (dead or Used claim data from the relationships between the diseases (eg, diabetes, chronicof surgical site infection National Health Insurance characteristics of the patientsalive), emergency procedures, Surgeon The surgical volume was divided into obstructive pulmonary disease,following coronary artery program in Taiwan lacked and hospitals and the (AMI of less than 24 h onset, Incidence volume and quartiles. heart failure, renal failure, andbypass graft surgery. The 40 2000- Taïwan 3,227 patients,CABG, cardiogenic shock, or OHCA). of Surgical Hospital volume : clinical information on the incidence of SSI. Then usedrenal insufficiency), hospital 2001 285 Ssis. Coronary Artery Hospital hospital volume has a greater First quartile (҅132) ; severity of infection, such a random-intercept model tocharacteristics (eg, ownership Bypass Graft site infection volume. effect than does surgeon Second quartile (133-182) ; as the depth of infection investigate the relationship and accreditation status). Surgeries. (SSI) during volume on reducing the (ie, super cial, deep, or between the incidence of SSI hospitalisation Third quartile (183-299) ; fi incidence of surgical site organ-space SSIs) and the and the surgical volume. and post Fourth quartile҆ ( 300). infection. discharge (within Surgeon volume : perioperative 30 days). First quartile (҅ 36) ; level of blood sugar. Second quartile (37-77) ; Third quartile (78-99) ; Fourth quartile҆ ( 100).

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 89 -

There is a significant relation between operator volume and Pays et/ outcomes in PCIs. High volume Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir Age, gender, presence of operators are more successful N° ou région Méthode Variables contrôlées Résultats diabetes, history of peripheraland encounter fewer adverse étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Pearson chi-square tests d’étude vascular disease, history of outcomes. Study demonstrates were used to assess the Study by Northern New chronic obstructive pulmonarytwo relations between operator Surgeon Operators were categorized into terciles univariate association Rates of CABG England Cardiovascular disease, previous PCIs, volume and outcomes for PCIs : 41 1990- US, New 12,988 Cardio/ volume. based on annualized volume between potential predictors (emergency/non study group. Univariate previous CABG and previous1) Higher operator volume 1993 Hampshire,Procedures Percutaneous of procedures. (Low, Middle, High). and the outcome variables. emergency), new and multivariate regression myocardial infarction is associated with a greater Maine, and5 Hospitals Coronary For each outcome, logistic MI, angiographic analyses were used to (MI). Indication for the degree of angiographic and Interventions. regression models run Massachusetts31 Cardiologists. success, clinical control for case-mix. used procedure (Asymptomatic, clinical success ; and 2) US (controlling case-mix) with success, in- our prospective, multicenter, Stable angina, Unstable lower operator volume is entry and exit criteria set at p hospital mortality. clinical data base to address angina, Postinfarction angina,associated with an increased ҅0.05. Rates for successful the question of whether Postinfarction anatomy, Primarylikelihood of postprocedural and adverse outcomes operator volume is related to therapy for MI, Cardiogenic CABG. However, a significant were then calculated using angiographic success, clinical shock) ; the priority of the difference in in-hospital the beta estimates from the success and adverse events procedure (emergency, urgentmortality across terciles was logistic regression model, and after PCIs. No control of and nonurgent) ; Angiographicnot present. direct standardization was hospital characteristics. data : presence of 50% left used to calculate adjusted main coronary artery stenosis rates. (LMCA) ; LAD, left circumflex artery, right coronary artery or in a bypass graft ; the number of lesions attempted ; the interventional devices used ; and the location and complexity of the treated lesion or lesions. Lesion complexity (A, B, C),

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 90 - Using current data, there is no significant relationship between operator volumes averaging҆ 68 per year and outcomes at high volume hospitals. Pays et/ 1) site and primary operator; Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour dé nir fi 2) demographic data: patient N° ou région MéthodeLogistic regression was Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) age, gender, height, weight ; Operators were categorized into tercilesStudy by Northern New used to assess whether d’étude Surgeon 3) medical history : previous based on their annualized volume of England Cardiovascular changes in case-mix differed volume. coronary artery bypass graft Cardio/ In-hospital procedures (Low, Middle, High). study group. Updates across terciles of operator 42 1994- US, 15,080 PCIs surgery (CABG), PCI or Percutaneous outcomes their previous results and volume, except for the ordinal 1996 Northern performed myocardial infarction (MI), Coronary included death, finds no significant relation variables of indication, New during 14,498 family history of premature Interventions. emergency for this period studied. priority and lesion type England hospitalizations coronary artery disease, the coronary No control of hospital where a Mantel-Haenszel and 47 cardiologists presence of congestive heart artery bypass characteristics. chi-square statistic was used. 6 Hospitals. failure, hypertension, treated graft surgery Multivariate models were Massachussets diabetes, current smoking, (eCABG), used to adjust for differences US hypercholesterolemia, chronic non-emergency in case-mix and severity obstructive pulmonary disease, CABG of illness across terciles peripheral vascular disease, (non-eCABG), of operator volume when cerebrovascular disease, renal myocardial comparing outcomes. infarction (MI), failure, baseline creatinine ; death and 4) primary indication for PCI : clinical success. stable angina, unstableangina, postinfarction angina, postinfarction anatomy, primary therapy for acute MI, cardiogenic shock ; 5) priority at PCI : emergent, urgent, nonurgent ; 6) therapy before during and after the procedure : intravenous heparin, intravenous nitroglycerin, thrombolytic therapy, insertion of an intraaortic balloon pump ; 7) cardiac anatomy and function : percent stenosis of left main coronary artery, number of other diseased native coronary vessels, dominance, number of bypass grafts, ejection fraction, left ventricular end diastolic pressure ; 8) PCI procedure information.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 An association between a higher volume of ICD implants and a lower rate of mechanical Patient characteristics adjustedcomplications and infections. for included age, gender, race,This association suggests that - 91 - urgency of admission (outpatientICD implantation should not Categorical variables are procedure, elective, urgent, be performed by physicians summarized as frequencies. emergency), admission for without regard to their procedural Using the 20% Part B MedicareDifferences in patient acute myocardial infarction, volume. Pays et/ files claims, identified all characteristics by physician Physician volume categories by assigning and Charlson comorbidity scoreConclude that patients who Période Echantillon Intervention Indicateur PhysicianIndicateur ICD Seuil utilisé pour définir claims with an appropriate volume category were N° ou région one-quarter of the patients to each Méthode (includesVariables cancer, contrôlées liver disease,have their Résultats ICD implanted by Volume. current procedural terminologyassessed using the chis-quare étudiée de l‘étude étudiée de résultat volume quartile.haut/bas Annual Medicare volume Volume(n) to renal failure, diabetes mellitus,high-volume physicians have d’étude 30-day mortality ; (CPT code) for placement test. Differences in outcomes 1672 PhysiciansCardio/ Outcome : myocardial infarction, peripherallower rates of mechanical 90-day mortality ; of a complete ICD for any across physician volume ICDs, implantable Open bars = 1 to 10 ; ruled bars = 11 to vascular disease, dementia, complications and ICD infection 30-day indication, a valid physiciancategories were assessed 43 1999- US, 9,854 Patients.cardioverter- 18 ; dotted bars = 19 to 28 ; chronic obstructive pulmonarythan patients who have their mechanical identifier, and a procedure dateusing the chi-square test 2001 Durham, defibrillator (ICD). solid bars = 29. disease, rheumatic heart disease,ICD implanted by low-volume complications ; between 1999-2001.These of trend. Multiple logistic North ICD Annualized Volume : 1-6 (n = 2,487) ; and acquired immuno-deficiencyphysicians. 90-day claims were collapsed into aregression models were Carolina ; mechanical 7-11 (n = 2,559) ; syndrome). We also controlledThis findings suggest that ICD Lebanon, set of unique patient identifiuseders to assess the effect of complications ; 12-17 (n = 2,336) ; 18 (n = 2,472). for the volume of pacemaker implantation should be directed New that became our study cohort.physician volume category on 30-day ICD implantations. toward high-volume physicians. Hampshire ; outcomes while adjusting for infection ; patient characteristics. and 90-day ICD Portland, infection. Maine- US

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 92 -

Analysis of variance Pays et/ was used for differences Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir Clinical, procedural, and N° ou région outcome data were Méthodebetween means, and Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Pearson chi-square test d’étude Operator Operators were grouped by quintile (Q)prospectively collected in volume. according to the number of procedures14 hospitals in Michigan was used for differences Clinical and point Cardio/ performed as follows : the first quintilein the calendar year 2002. in frequencies. Trends 44 2002 US, 165 operators ; Adverse Baseline demographics Although the relationship percutaneous included operators performing 1 to 33Data were collected with a across quintile of procedure 18,504 Outcome: (age and gender, smoking), between operator volume Michigan coronary PCI/year, the second quintile includedstandardized data collectionvolume were evaluated consecutive Contrast comorbidities (historical : and in-hospital mortality is - US - intervention operators performing 34 to 89 PCI/year,form and standardized with the Cochran-Armitage PCIs ; hypertension, diabetes mellitus, nephropathy, the third quintile included operators definitions. Data are trend test for significance. no longer significant, the (PCI). extracardiac vascular disease, 14 hospitals. Vascular performing 90 to 139 PCI/year, the expressed as mean standardIndependent predictors of relationship between volume congestive heart failure, renal complications, fourth quintile included operators deviation or as percentage.in-hospital mortality and and any adverse outcome is failure requiring dialysis, gastro- Transfusion, performing 140 to 206 procedures/year, MACE were determined by still present. Technological intestinal bleeding, chronic, Death, stepwise multivariate logistic advancements have not yet and the fifth quintile included operators procedure obstructive pulmonary MI, performing 207 to 582 procedures/ regression analysis. Model completely offset the influence disease, atrial fibrillation, history Stroke or TIA, of procedural volume on year. The ACC and AHA currently discrimination was assessedof cardiac arrest, prior history of Emergency pro ciency of PCIs. recommend a minimum of 75 PCI/year with the c-statistic, and percutaneous intervention, prior fi CABG, per operator. Therefore, a secondary goodness of fit was assessedhistory of coronary artery bypassAfter adjustment for All CABG, analysis was performed with the 75 with the Hosmer-Lemeshowsurgery. Clinical presentation comorbiditiesat and other Repeat PCI, PCI/year as a cutoff to differentiate “low statistics. Three groups of time of PCI : creatinine 1.5 mg/dl,confounders, low procedure MACE. volume operators” from “high volume models were fitted. In the firstMI within 7 days, MI within 24volume h, continues to operators.” group, hospitals were cardiac arrest, prior thrombolysis,be a predictor of worse considered as fixed effects ;cardiogenic shock, ventricularoutcomes (MACE), and this in the second group, a tachycardia or fibrillation in therelationship appears to be random effect was included,setting of acute MI, emergencyrelatively independent of the assuming normal hospital- angioplasty, rescue angioplastypatients’risk. In addition, the effect distributions ; and in after failed thrombolysis, volume threshold might be the third group, generalizedunstable angina (requiring higher than the threshold of 75 estimating equations were intravenous nitroglycerin and procedures/year suggested by heparin treatment), Procedural fitted to control for clustering the ACC/AHA guidelines. characteristics : number of and variation by hospital. diseased vessels (70% stenosis), left ventricular ejection fraction 50%, ACC type C stenosis, visible thrombus on the initial coronary angiogram, prior history of MI, left main stenosis (70%), anemia (hemoglobin 10g/dl), and the presence of moderate or heavy calcification.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 93 - Study provides additional evidence on the benefit of regionalisation of cardiac Demogrophic characteristics:surgery interventions. Better Pays et/ patient age, sex; Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour dé nir clinical outcomes could be fi Comorbidity (measured with N° ou région MéthodeUnivariate analyses were Variables contrôléesachieved Résultatsif clinical practice étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) the Charlson comorbidity was regulated in such a way d’étude carried out to assess index12) conditions : Chronic Case mix adjusted before as to assure that each surgical Surgical Before/after study. Compares differences in patients’ caseobstructive lung disease, 16,512 patientsCABG, Cardiac. Mortality rates: the outcomes before after 2000 (1998–1999) - after (2000– mix across the ve study centre had a volume of cases 45 1998- Italy, center volume fi Congestive heart failure, aged҆18 years In-Hospital, 30 (regionalisation). 2002) comparison. Patientsyears, using the X² test for sufficient to maintain the 1999 Emilia- of cases. Cerebrovascular disease, 6 Centres. Day. Patients’ were identified in the regionaltrend or analysis of variance required clinical competence. 2000- Romagna Peripheral vascular disease, timely referral hospital discharge databasefor categorical and In-hospital mortality reduced by 2002 Renal failure in previous to hub centers, (SDO) provides information 22%, 30 day about 18%.imely T continuous variables, admissions ; average waiting on patient. The impact of referrals for sugery increased respectively. Yearly mortality Type of surgical intervention times. the regionalisation policy by 21% and mean waiting and referral rates were - Admission (elective was assessed by regression times reduced by 7.5 average adjusted through a logistic versus urgent), degree model including a dummy days. regression model including of risk (assessed on the variables which represent covariates. Year specific presence of key clinical the period (2000–2002) right adjusted mortality and conditions : diagnosis of after regularisation (2000). case mix adjusted referral acute myocardial infarction, The policy effect is expressed rates were calculated by cardiac catheterisation on the in terms of odds ratios (OR) multiplying O/E by the overallsame day of the intervention, with 95% confidence intervals mortality rate observed overpercutaneous transluminal (95%CI). the whole study period. A coronary angioplasty multivariate linear regression(PTCA) on the same day of used in assessing changes intervention, Intra-aortic balloon over the study period in pump), type of surgery. patients’ waiting times, adjusting for the same set of covariates.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 94 -

Higher-volume hospitals and surgeons provide superior Pays et/ short-term outcomes after Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir Patients demographics : age, N° ou région Méthode Variables contrôléestranssphenoidal Résultats pituitary tumor sex, race, median household étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) surgery with shorter lengths of d’étude income for ZIP code of stay and a trend toward lower Statistical methods residence, primary payer TranssphenoidalAdverse For these years, the NIS charges. Complications were 5,497 included the Fisher’s exact (Medicare, Medicaid, private 46 1996- US surgery for outcomes : Annual hospitalSingle cutpoint with resultant contains discharge data on coded in 26.8% of patients. operations and Wilcoxon rank tests, insurance, self-pay, no charge, 2000 non federal pituitary tumor. inhospital and surgeon information loss (8 – 10), 100% of discharges from a Patients of higher-volume 538 hospitals caseload of Spearman rank correlation, other), type of admission Hospital mortality, Low and High volume quartiles for stratified random sample of hospitals and surgeons had 825 surgeons transsphenoidal and loglinear least-squares,(emergency, urgent, elective outcome transsphenoidal pituitary tumor surgerynon federal hospitals lower mortality rates, better 18 - 28 States. surgery, as well ordinary logistic, and or routine) and admission at hospital caseload (one to four admissions perin 19 – 28 states. hospital discharge disposition, as to other proportional-odds ordinal source (emergency room discharge, year and 25 or more admissions per and fewer complications. patient and logistic regression. To correctward, transfer from another complications provider year, respectively) Patients of high-volume for possible clustering of hospital acute care hospital, of surgery, lengthcharacteristics. Hospital volume by quartile : 1-4 ; 5-9 ; providers were younger, similar outcomes within transfer from long-term care, of stay (LOS), 10-24 ; 25 + (admissions/year) healthier, and white, more often hospitals, which could causeand routine), Year of treatment, and hospital Surgeon volume by quartile : 1 ; 2-3 ; had private insurance, and falsely inflated estimates of Procedure : Biopsy, Partial charges. 4-7 ; 8 + (admissions/year). resided in wealthier the statistical significance hypophysectomy, areas. of regression coefficients, Total hypophysectomy, a sandwich variance- Cushing’s disease. Hospital covariance matrix region (Northeast, Midwest, was estimated from the dataSouth, or West), location (rural using methods due to Huberor urban), teaching status, and White, with adjustment and bed size (small, medium, for clustering by hospital or large). surgeon. LOS and hospital charges were analyzed as logarithmic transforms using least-quares regression corrected for clustering as described above.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 95 -

For patients undergoing major Pays et/ Covariates : Patients cancer surgery, the bene ts of Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir fi N° ou région Méthode characteristicsVariables contrôlées (number, age, volume-based Résultats referral depend étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Data from the Healthcare The bivariate relationship race (white, African American,on the interplay between d’étude Within each year, hospital volume was Annual Cost and Utilization Project between each of the Hispanic, other, or missing), procedure utilization, the measured and sorted into deciles hospital Nationwide Inpatient Sampleoutcomes and the categoricsex, admission acuity (electivemagnitude of effect, and the 47 1993- US No. of patientsCancer : surgeriesOperative separately for each procedure. These volume. were abstracted. covariates was assessed v nonelective admission - outcome chosen. High volume 2003 Prostatectomy :(prostatectomy, mortality was deciles were ranked from low (bottom using the Rao-Scott 2 urgent/emergent admission), hospitals had lower operative 141,052 cystectomy, defined as an decile) to high (top decile). test, whereas the bivariate type of insurance (private, mortality rates. Generally, Cystectomy : esophagectomy,intraoperative relationship between the Medicare, or other), and mortality declined over time for 19,319 pancreatectomy,death or death Esophagectomy : outcomes and the continuouscomorbidities. all of the cancer procedures. pneumonectomy, 4,020 during the course covariates was tested using The regionalization of all Pancreatectomyand : liver of hospitalization a simple logistic regression procedures (the bottom 9,153 resection). after cancer model and the Wald 2 90th percentile) to high- Pneumonectomy : surgery. test. For each procedure, volume hospitals would 90,088 Prolonged unconditional logistic save many more lives and Liver Resection : LOS for each regression models were used avoid many more prolonged 3,630. procedure to estimate the adjusted hospitalizations than reported defined as the attributable risk as a result of in this study. We chose the group of patients hospital volume, comparing bottom decile of hospital whose LOS was low-volume with high-volume volume as a threshold for greater than the hospitals. volume-based referral across 90th percentile all procedures for consistency within each year. and practicality; although it may be feasible to regionalize all esophagectomies, it would be impractical to do the same for prostatectomy. Furthermore, the relationship between some procedures and outcomes may be primarily mediated at the level of the surgeon rather than at the level of the hospital.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 Hospital surgical volume had no significant effect on - 96 - Baseline Charateristics : rectal cancer recurrence Number of patients and or survival when patients hospitals, annual case volume,completed standard adjuvant Pays et/ The distribution of baseline age, sex, race (White, African-therapy. Sphincter preserving Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir characteristics across American and Other), bowel surgery was more commonly N° ou région Prospectively recordedMéthode datahospital procedure volume obstructionVariables at contrôléespresentation, performed Résultats at high-volume étudiée de l‘étude étudiée de résultat Hospitalvolume Hospitalhaut/bas procedural volume volume (n) defined by d’étude of Cohort study and the NIStertiles was evaluated usingperformance status, number centers. Nonetheless, patients procedure tertiles of low database (345 hospitals), theMantel-Haenszel X² tests of positive lymph nodes, and who underwent surgery at Rectal cancer Rates of volume. [0 to 8.3 Medicare cases/yr], mediumSpearman rank correlation for 1,330 patients for categoric variables and extent of disease through low-volume centers were 48 1990- US, (primary). sphincter- [8.4 to 16.7 Medicare cases/yr], annual rectal cancer surgery ANOVA for continuous the bowel wall, distance from significantly more likely to 1992 11 states Primary rectal reserving and high [17 to 92 Medicare cases/yr]) volume as measured by thevariables. Overall survival anal verge (on the basis of require a permanent colostomy cancer surgery operations. on the basis of the Medicare Medicare claims database (OS), disease-free survival operative, sigmoidoscopy, or by virtue of undergoing an was performed Cancer procedure volume of the hospital. and the NIS database was (DFS), recurrence-free colonoscopy report), InvasionAPR. at 646 United Recurrence 0.87. survival (RFS ; local or to perirectal fat or adjacent States (Mortality) and Used data from a large distant), and local recurrence-organs, Performance status hospitals. Survival rates : randomized trial of free survival (LRFS) rates at study entry, Grade of * 5-year overall patients with stage II and were examined using the differentiation (Well, grade 1 ; survival ; III rectal cancer to examine methods of Kaplan and Moderately, grade 2 ; Poorly, *5-year disease- the in uence of hospital fl Meier, and differences amonggrade ; Other) ; Influence of free survival ; procedure volume on tertiles were assessed by *5-year Adjuvant Therapy. sphincter-reserving surgerythe log-rank test. The entire recurrence-free and long-term outcome. cohort was analyzed using survival ; Cox proportional hazards * 5-year local regression, with a priori recurrence-free inclusion in the model of survival. Patients characteristics. Differences in rates of APR by hospital volume tertile were analyzed using logistic regression, with adjustments.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 97 -

Pays et/ Within-hospital longitudinal Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir analyses revealed that N° ou région Méthode CovariatesVariables included contrôlées age Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Odds ratios (as estimates of(65–69 years, 70–74 years, hospitals with a relative d’étude Cross-sectional analyses relative risks) were derived or <75 years), race, surgeon increase in prostatectomy descriptive statistics from the logistic regression speciality (urology or other), volume had a 57% greater reduction in the length of a 49 1991 - US 101,604 Radical Length of a Volume of Hospitals were classified into regarding rates of and were used to assess theprostatectomy volume (low, 1994 prostatectomiesprostatectomy. patient’s hospitalhospital load offour categories of prostatectomy complications, readmission,effects of hospital volume medium-low, medium-high, patient’s stay compared with and mortality in each hospital those with a relative decrease aged 65 years stay, Patients prostatectomies.volumes based on the number of on short-term outcomes. Toor high), teaching status of old. outcome : prostatectomies performed : volume category were age-assess volume relatedness,the hospital, comorbidity (yesin volume (P = .005). Changes in prostatectomy volume did surgical 1) Low volume҅ ( 25th percentile or adjusted by use of the entireused linear regressions to or no) of patients affected study cohort as the standard notaffect the frequency of complications, ҅38 prostatectomy surgeries in the conduct trend tests. (Hypertension, Cardiac, readmission rate, study period ; number of hospitals in thispopulation. Pulmonary, Diabetes, Renal complications, mortality, and and mortality category = 2,013); disease); and year of surgeryreadmission. Thesefindings rate in a 30-day 2) medium-low-volume (26th–50th (1991, 1992, 1993, or 1994). suggest that an increase in a period following percentile or 39–74 surgeries in the given hospital’s prostatectomy surgery. study period ; number of hospitals in this volume may decrease the category = 4,463); length of stay without an 3) medium-high volume (51st–75th adverse impact on patient percentile or 75–140 surgeries in the outcomes. study period ; number of hospitals in this category = 4,257) ; 4) high-volume҆ (76th percentile or ҆141 surgeries in the study period ; number of hospitals in this category = 116), Surgery volume : Low (҅38) Medium-low (39–74) Medium-high (75–140) High (҆141).

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 98 -

In unadjusted analyses Demographic (age, sex, race) Pays et/ across decreasing quartiles and clinical characteristics Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir of hospital volume, we N° ou région MéthodeAssociations of hospital (tumorVariables stage, contrôléessocioeconomic Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) observed statistically significant volume with permanent status quartile, hospital volume, d’étude This study assessed the increases in colostomy rates Quartiles of hospital volume : colostomy and comorbidity index, Tumor Number of association of hospital and in 30-day postoperative Low to High <7; 7 – 13 ;14 – 20; >20. and 30-day mortality location (Rectal Rectosigmoid). 7,257 patients Rectal Cancer. Three end pointsoperations mortality and a decrease in 50 1994- US, volume with colostomy rateswere assessed with the 367 Hospitals. were analyzed : performed 2-year survival. The adjusted 1997 California - and survival for patients withMantel–Haenszel trend permanent at a hospital risks of permanent colostomy, US rectal cancer in a X² test and multivariable colostomy, (i.e. hospital 30-day mortality, and 2-year large representative cohort logistic regression. Overall 30-day volume). mortality were greater for identified from the Californiasurvival was examined with postoperative patients at hospitals in the Cancer Registry. Data the Kaplan–Meier method mortality, and lowest volume quartile than Sources and Study Cohort. and a multivariable Cox overall mortality for patients at hospitals in patients diagnosed from withproportional hazards model. rate in the 2 the highest volume quartile. stage I – III rectal cancer whoMultivariable analyses years after Strati cation by tumor stage underwent surgical resection.adjusted for demographic and fi surgery. and comorbidity index did not Registry data were linked toclinical variables and patient appreciably affect the results. hospital discharge abstractsclustering within hospitals. All and ZIP-code-level data from Adjusted colostomy rates tests of statistical significance varied statistically signi cantly the 1990 U.S. Census. were two-sided. fi (P<.001) among individual hospitals independent of volume. Rectal cancer patients who underwent surgery at high- volume hospitals were less likely to have a permanent colostomy and had better survival rates than those treated in low-volume hospitals. Identifying processes of Age, Complexity Score.care There that contributewas an inverse to these differences may improve Data were analyzed with correlation between hospital patients’ outcomes The Maryland Health the Kruskal-Wallis or Mann- caseload and operative Low-volume ҅( 10 cases/year ; n = 12), hospital mortalityin all hospitals. rate, and a Postoperative Services Cost Review Whitney tests Post-hoc 9,918 elective Carotid surgical moderate-volume signi cantly higher complication 51 1990- US, death/strokes Commission database comparisons were performed fi CEAs endarterectomy volume. (11 to 49 cases/year ; n = 24), and rates in low-volume compared 1995 Maryland - and neurologic was reviewed to identify with Fisher’s protected 48 hospitals. (CEA). high-volume with moderate and high volume US complications all patients who underwent least significant difference, (҆50 cases/year ; n = 12) institutions. hospitals. Furthermore hospital rates (Hospital elective CEA as the primaryBonferroni, and Scheffé’s charges were signi cantly charges and procedure in all acute care post hoc tests, using fi lower in high-volume hospitals LOS). hospitals in the state over theanalysis of variance as an past 6 years. All perioperativeapproximation of the Kruskal- reflecting shorter LOS. deaths and strokes identifiedWallis test as a result of the in the database. large sample size.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 99 -

Pays et/ AFB operative mortality was Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode PatientVariables characteristics contrôlées : significantly Résultats lower at high- étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) volume hospitals (more than d’étude Age (Without/With COPD 25/year) in this nationally Unadjusted and case overall, sex, race), nature of admission (Elective admission,representative database (40% Data from a nationally mix–adjusted analyses 52 1997 US 3,073 patientsAorto-bifemoral * In-hospital Hospital Hospital volume was Urgent admission, Emergent decreased risk of hospital representative hospital were performed. 483 hospitals. bypass (AFB). mortality, Volume. a dichotomous variable with low volume admission), comorbidity Indexmortality). Without other discharge database : Univariate comparisons of * Prolonged LOS. (25/y) and high volume (>25/y). diseases, information regarding quality Nationwide Inpatient hospital volume, patient Insurance (Medicare, Private,of surgical care for patients Sample (NIS). The NIS is a characteristics, and outcome Medicaid) ; Median income in need of AFB, referral to a 20% random sample of all variables with the 2 tests, by ZIP code ; COPD, chronic high-volume center can result hospital discharges in the Wilcoxin rank-sum test, simple pulmonary disease. in superior outcome. United States, stratified by logistic and linear regression, geographic region, urban orwhere appropriate. The Charlson comorbidity score rural location, teaching status, For elective AAA repair, and was used with ICD- 9-CM ownership, and bed size. even more so for ruptured AAA codes from an index Patients FactorsAge : (per 5 repair, high-volume surgeons hospitalization to account for y) ; Charlson Comorbidity Indexwith ; subspecialty training comorbidity. Multiple logistic Gender ; Income quintile (per conferred a significant survival Electiveregression aneurysm of in-hospital repairs increase of 1) ; Distance from benefit for patients. Although andmortality repairs and of prolongedruptured patient’s home to hospital (km)this ; would seem to argue In this population-based aneurysmsLOS was used were to considered test their Annual surgeon volume was calculated Surgeon factorsAnnual : in favor of regionalization, Surgeon retrospective cohort separately.association Correlation with hospital by enumerating the number of cases decisions should await a more AAA, abdominalSurvival after Volume Rates study, surgeon billing and betweenvolume after continuous adjusting variables for surgeon volume of ruptured 2,601 patients a given surgeon performed each year. complete understanding of the 53 1992- Canada, aortic aneurysm.repair of rupturedand Hospital administrative data were waspotentially examined. confounding Patient, patientAAA repairs (per 5 cases) ; with ruptured Cut points were selected to de ne relationship between transfer 2001 Ontario - abdominal aorticfactors. fi used to identify all patients surgeon,case mix andvariables. hospital factorsAnnual surgeon volume of time, delay in treatment, and Canada. AAA repair ; aneurysm (AAA) high-volume and low-volume surgeons.who had undergone AAA independently associated elective AAA repairs (per 10 13,701 patients These cut points were chosen a priori outcome. and survival repair. Demographic with survival were identiedfi cases) ; Procedures performed, with elective as the median. after elective information was collected with a proportional hazards by surgeon specialty :General AAA repair. AAA repair. for each patient, as well as multivariate backward surgeon ; Cardiothoracic And Crude and numerous variables related selectionto process. Proportionalsurgeon ; Vascular surgeon. Adjusted 30-day the surgeons and hospitals.hazards survival analysis wasHospital factorsAfter-hours : mortality rate. Data were obtained from used because it has superiorrepair (weeknight, weekend, four sources : the CIHI data power to logistic regression holiday) ; Procedure performed base, the OHIP data base analysis. Hazard ratios for in teaching hospital ; (physician billings), census variables were compared withPopulation of city where data, and the Ontario a two-tailed z test between operation performed ; Annual Physician Human Resourceselective AAA repair and hospital volume of ruptured AAA Data Center (OPHRDC) dataruptured AAA repair. All volumerepairs ; Annual hospital volume base. measures were analyzed in of elective AAA repairs ; Year our multivariate operation performed. survival analysis model as continuous variables.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 100 -

Pays et/ Adjusted for the severity of Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir illness and other differences N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Used data from the in the case mix. Hospital Categorizing volume into quartiles : d’étude Acute Physiology and Bivariate analysis comparing Volume. Quartile 1 (87 – 150 patients/yr) Chronic Health Evaluation demographic characteristics Mechanical Survival rates ; Quartile 2 (151 – 275 patients/yr) (APACHE) clinical outcomesacross the quartiles of 54 2002- US 20,241 ventilation. mortality rates Quartile 3 (276 – 400 patients/yr) Hospitals Characteristic : Mechanical ventilation of database. a cohort study hospital volume was 2003 nonsurgical and the length Quartile 4 (401 – 617 patients/yr). Teaching status {Academic, patients in a hospital with patients ; 37 of the stay at with the use of the APACHEperformed with analysis Community with house staff, a high case volume is acute care hospitals clinical database, which of variance for continuous Community} associated with reduced hospitals ; Length of prospectively collects variables and the chi-squareRegion {Southeast, Northeast,mortality. Further research Midwest, West} 122,498 mechanical information on patients in 104test for categorical variables. is needed to determine the No. of hospital beds {Median ; patients ventilation ; ICUs at 45 hospitals in the Multivariate modeling was mechanism of the relationship Interquartile range} Length of stay United States as part of its performed with logistic admitted No. of APACHE ICUs/hospitalbetween volume and outcome predictive capabilities. The regression for mortality and to 104 ICUs at in ICU ; Length {Median, Interquartile range} among patients with a critical linear regression for length of 45 hospitals. of hospital stay APACHE database (Cerner) Non-APACHE ICU beds illness. Data demonstrate an hospital stay, after adjusting (days) ; Mortality supplies data on risk-adjusted Patients CharacteristicAge, : association between outcomes to participating for potential confounders. in ICU ; Sex, APACHE III score, Racean or increase in hospital For the assessment of In-hospital hospitals for benchmarking ethnic group Transferred fromvolume and a lower risk mortality. purposes. volume as a continuous outside hospital, Discharged adjusted mortality among Local, trained clinical variable, used the fractionalfrom ICU to outside hospital ; patients requiring mechanical coordinators supervise polynomial method, an Type of ICU {Multidisciplinaryventilation. ; data collection. iterative estimation processMedical ; Surgical ; Coronary that determines the best- care ; Neurologic ; Cardiothoracic fitting polynomial-regressionsurgery}; Diagnostic category function. {Pneumonia ; Congestive heart f

ailure ; Neurologic disease ; Sepsis Chronic; obstructive pulmonary disease ; Post – cardiac arrest ; Drugoverdose ; Gastrointestinal bleeding ; Oncologic disease ; Acute coronary syndrome ; Arrhythmia ; Other respiratory disease ; Other medical disease ; Other gastrointestinal disease ; Other cardiac disease} Intensivist physician staffing {Low-intensity care model ; High-intensity care model}.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 101 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) d’étude Carotid endarterectomy has Patients in HVAMCs have CEA volume. Five high volume AMCs (HVAMCs) Carotid In-hospital stroke a further advantage in that Patient age, sex, hospital the best outcomes after CEA. US, Statistical analysis was 55 1995- 10,211 Patientsendarterectomy and mortality were compared with all other the indications and outcomes costs, demographic, and Despite the achievement of performed with analysis of 1996- Massachussets-10,211 CEAs (CEA). rates. nonacademic hospitals, which were for this operation have been charge reports, LOS, and significant efficiencies, AMCs variance to compare the 1997 US 60.3 hospitals further subdivided by annual volume extensively scrutinized in the disposition. Procedure volumehave a small cushion to reduce means of all the cost and 3,388 into high volume non-AMCs҆50 ( United States and abroad (Indication (eg, asymptomaticfurther either LOS or resources LOS data, and χ 2 tests procedures cases), medium volume non-AMCs (24-with randomized prospective versus symptomatic)). The to maintain a competitive cost was used for comparison of in FY95, 3,568 49 cases), and low volume non-AMCsstudies. Health Share total number of CEAs by yearposition and to compensate for incidence (signi cance in FY96, and (12-23 cases). Technology maintains a fi and grouping. The number thefixed expenses of academic assumed for P ҅.05). 3,255 in FY97. database. Data on individual of medical centers in each medicine. The Balanced patients, including charges, of the categories. Minor or Budget Act raises an equity were obtained from the nonteaching hospitals. concern for AMCs because it Massachusetts Division of differentially affects the centers Health Care Finance and with the best outcomes. The Policy, and hospital cost and financial implication of this may charge data were provided be a direct incentive for by the Health Care Financing Administration. The cost

procedures to be done in centers data with less optimal outcomes. were determined with the ratio of cost to charges (RCC) approach, which requires two steps.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 There was a continuous inverse dose–response Clinical Diabetes mellitus, - 102 - relation between hospital Prior congestive heart failure, volume and the risk of death. Prior myocardial infarction, The availability of technology Prior angioplasty, Prior Used proportional- for angioplasty and bypass bypass surgery, Mean heart Pays et/ hazards methods to surgery was not independently Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour dé nir rate, Mean arterial pressure fi Performed a retrospective adjust. Multivariate model, associated with overall N° ou région Méthode anteriorVariables infarction, contrôlées Killip class, Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) cohort study, using data adapted from models for mortality. Patients with acute demographic (age, sex, race), d’étude Hospital Hospital volume as an ordinal variablefrom the Cooperative the assessment of 30-day myocardial infarction who are and health-system–related 30-Day Survival volume, according to quartile, the hazard ratioCardiovascular Project mortality after myocardial admitted directly to hospitals Cardio/Acute variables, including the 56 1994- US 98,898 rates ; Physicians. for 30-day survival among patients in(CCP), which was conductedinfarction. The independent that have more experience Myocardial availability of invasive 1995 Medicare mortality more the lowest versus the highest hospital-by the Health Care Financingvariables were collinear treating myocardial infarction, Infarction in procedures, the specialty of the patients 65 than/within volume quartile was 1.17 (95 percentAdministration (HCFA). This(Spearman correlations, as reflected by their case Eldery Patients. attending physician, and the years of age or 30-day and 1 yr confidence interval, 1.09 to 1.26 ; cohort was uniquely suited <0.5). Used Kaplan–Meier volume, are more likely to area of residence of the older ; after admission; P<0.001); for the second quartile the for the analysis of the effectscurves to compare survive than are patients patient (rural, urban, or Hospital long term ratio was 1.07 (95 percent confidenceof aspects of the health unadjusted survival. To admitted to low-volume metropolitan). = 95,185 ; mortality. interval, 1.01 to 1.13 ; P = 0.02); and care delivery system : the determine whether there hospitals. Physician’s for the third quartile the ratio was 1.05nationwide sample comprisedwas a selection bias, we N = 81,728. (95 percent confidence interval, 0.99 nearlyto 100 percent of elderlyused logistic regression to 1.10 ; P =0.11). patients with myocardial calculate a predicted one- Hospital volume : infarction who had fee-for- year mortality rate for each * <1.4 patients/wk(n =24,829); service insurance coverage,quartile of hospital volume * 1.4–2.5 patients/wk(n =24,743); and the study had extensiveand for each physician * 2.6–4.4 patients/wk (n =24,648); clinical data, blinded data specialty. We used data on * >4.4 patients/wk (n =24,678). abstraction, and reliable long-covariates for each patient term follow-up. from the proportional hazards model while neutralizing the effects of hospital-related variables by standardizing them to the lowest-mortality groups. Generalized estimating equation, and the Cox proportional-hazards survival analysis. Unadjusted Kaplan–Meier survival curves for quartiles of hospital volume.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 103 -

Characteristics of eligible Multiple logistic regression Pays et/ infants : Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour dé nir models of the three fi High consultant availability; N° ou région Méthodeprimary outcomes were Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) To identify the groups of High nurse provision; Number d’étude developed to adjust for risk Data about unit activity and staf ng units required for the factorial of infants; Number of units; fi before the study hypotheses Volume provision allowed strati cation into design, a national census Birthweight; <1500 g; Primary fi were investigated. 57 1998- UK 186 UK NICUsNeonatal of patients, 12 types of NICU with a three (high, was done. Phase I census Gestation (weeks); AttributableStudy showed that high-volume outcomes were : 1999 13,515 infants intensive-care nursing medium, and low volume) by two (highdata identified 186 UK NICUs NICUs in the UK cared for hospital mortality, 54 NICUs. units (NICUs). provision, and low neonatal consultant provision)(units intending to provide sicker infants, were busier, and mortality and neonatal by two (high and low nursing provision)sustained neonatal intensive had higher crude mortality and or cerebral consultant factorial design. Patient volume care). The census was done morbidity than medium-volume damage, and provision. was de ned as number of very low without nationally agreed and low-volume units. Risk nosocomial fi birthweight infants (<1500 g) admittedand verifiable definitions adjusted mortality or cerebral bacteraemia. per year : high-volume NICUs of neonatal intensive care, damage were unrelated to nurse or cot establishment, mortality; Cerebral abnormality; admitted more than 57 per year ; patient volume or staffing or unit size. Procedures medium-volume units admitted 35–57 Nosocomial bacteraemia. provision. and statistical analysis per year ; and units with low patient Controversy still exists about volumes admitted less than 35 per year. the best period—in the case of neonatal intensive care from birth or up to 12 h after admission—to obtain data for use in risk-adjustment methods.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 104 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume Hospitalshaut/bas and surgeons volume were (n) classifiedStudy Cohort Selected: d’étude Hospital into four volume categories on the basisall patients 65 years old or volume and of quartiles of the patient cohort. older who were listed in the 11,522 patientsProstatectomy. Outcomes In men undergoing 58 1992- US, New - surgeon SEER data base as having The following patient-relatedPatients Charateristics 159 surgeons after radical prostatectomy, the rates of York volume (the received a diagnosis of covariates were used for risk(age, race No. of patients, 1996 72 hospitals. prostatectomy. postoperative and late urinary US number of prostate cancer between theadjustment. The extent of No. of patients per hospital, * Postoperative coexisting illnesses according complications are significantly procedures index period. the cohort by Romano–Charlson comorbidity death (Surgery- excluding patients who wereto the Romano modification of reduced if the procedure is performed the Charlson index. Calculatedindex), Stage III or IV cancer. related death 30 not treated in a SEER state, performed in a high-volume at individual hospital volume in two ways : and 60 days), hospital and by a surgeon who hospitals and were not enrolled in both Partwith all the patients and with * Postoperative performs a high number of by individual A and Part B of Medicare, oronly those 65 years of age or complications, older. such procedures. surgeons, were not listed in Medicare * Late urinary The Spearman rank-correlation respectively). records as having undergone complications, prostatectomy within six coefficient. Analyses of volume and were adjusted for within- months after the diagnosis.hospital (or within surgeon) * Long-term correlations in outcome incontinence. with use of the generalized- estimating- equations modification of logistic regression. Used a correlation-adjusted logistic- regression model to estimate the expected rate of events for each surgeon by adjusting for the surgeon’s patients’ ages and coexisting illnesses and by adjusting for hospital volume and surgeon volume, modeled as continuous variables. Used the standardized deviations (z scores) of the observed frequencies from the expected frequencies for each surgeon to calculate the correlation between the outcomes and to characterize the degree to which individual surgeons could be identied fi as being associated with exceptionally high or low risk of adverse outcomes.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 105 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir In the absence of other N° ou région MéthodeMultiple logistic regression.Variables Patients charateristics contrôlées : information Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) age, sex, race, interaction, d’étude The Medicare Provider about the quality of surgery at Analysis and Review year of procedure, the the hospitals near them, Hospital volume, expressed as the relative urgency of the index Mortality (in- Hospital (MEDPAR) files. Thesefiles Medicare patients undergoing 59 1994- US 14 types Cardiovascular average number of procedures admission (elective, urgent, hospital or withinvolume (total contain hospital-discharge selected cardiovascular 1999 of procedures procedures and per year, was first evaluated as a or emergency), the presence 30 days). number of abstracts for the acute or cancer procedures can 43 percent (foreight types of continuous variable. Created categorical procedures care hospitalizations of all of coexisting conditions, and significantly reduce their nephrectomy) major cancer variables, definingfive categories of performed per Medicare recipients to create mean income from Social risk of operative death by resections. hospital volume : very low, low, medium, to 75 percent year). cohorts for the analysis of Security. Charlson score » 3.selecting a high-volume high, and very high. For each (for carotid outcomes, applied several hospital. procedure, the hospitals were restrictions in order to endarterectomy). ranked in order of increasing total increase the homogeneity of hospital volume, and then five volume the study samples and thus groups were defined by the selection of minimize the potential for whole-number cutoff points for annual confounding by case mix. volume that most closely sorted the patients into five groups of equal size (quintiles).

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 106 - For many procedures, the observed associations between Patients characteristics : hospital volume and operative Pays et/ {age group (in ve-year Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour dé nir Used multiple logistic- fi mortality are largely mediated fi regression analyses with intervals), sex, race (black or N° ou région Using information fromMéthode the Variables contrôléesby surgeon Résultats volume. Patients étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) adjustment for characteristicsnonblack), year of procedure national Medicare claims data can often improve their chances d’étude of the patients. adjusted (1998 or 1999), procedure was Volume was evaluated as a continuousbase. The Medicare Provider of survival substantially, even for the effect of clustering performed electively or not, and Operative Surgeon (log-transformed) variable in the Analysis and Review at high-volume hospitals, by 60 1998- US 474,108 Cardiovascular of patients within surgeons the mean income from Social mortality. volume and assessment of statistical significance.(MEDPAR) and inpatient selecting surgeons who perform 1999 patients procedures and clustering of surgeons Security in the ZIP Code of the hospital Cutoff points into three evenly sized les, which contain hospital- the operations frequently. 8 procedures. and 4 cancer fi within hospitals by using patient’s residence. Charlson volume groups with low, medium, and high discharge stracts for the The patients treated by high- resections. binary mixed-effects modelsscores}, (each in volume. In sensitivity analyses, we fee-for-service, acute volume surgeons had lower incorporating the two levelsHospitals characteristics : terms of total recategorized hospital volume as care hospitalizations of all operative mortality rates than of nesting. the type of ownership procedures a binary variable according to the Medicare recipients, were those treated by low-volume (not-for-profit, for-profit, or performed percriteria established by the Leapfrog surgeons. Surgeon volume used to create our main government), location year). Group for four procedures : coronary- accounted for a relatively large data sets for analysis ; the (urban or nonurban), and artery bypass grafting (450 or more proportion of the apparent Medicare denominator file teaching status. procedures per year vs. fewer than was used to determine the effect of hospital volume, to a 450), repair of abdominal aortic vital status of the patients. degree that varied according aneurysm (50 or more per year vs. to the procedure. For some fewer than 50), esophagectomy (13 or procedures, the association more per year vs. fewer than 13), and between hospital volume and outcome disappeared almost pancreatic resection (11 or more per Among hospitals in the entirely after surgeon volume year vs. fewer than 11). United States that have full had been taken into account. The study variables included interventional capabilities, a The relative importance of characteristics of the hospitalshigher volume of angioplasty A linear association among (urbanor rural location, surgeon volume and hospital The National Registry of procedures is associated with the four quartiles was the number of beds, and volume varies according to the Quartiles of Primary angioplasty or Myocardial Infarction (NRMI) a lower mortality rate among Volume assessed with the use of thewhether or not the hospital procedure. In-hospital Thrombolytic therapy. is a voluntary registry of patients undergoing primary Cardio, Acute of primary Mantel–Haenszel chi-squarewas a teaching center) and 61 1994- US 772,586 mortality The volume was calculated as the totalcross-sectional data on angioplasty, but there is no myocardial angioplasty or test for categorical variables,demographic and clinical 1998 patients (Survical after number of patients who underwent patients hospitalized with association between volume infarction thrombolytic analysis of variance for characteristics of the patients, 1,470 AMI). primary angioplasty at each hospital confirmed myocardial and mortality for thrombolytic Primary therapy. continuous variables, and medical history Angina, participating divided by the total number of days forinfarction. Entire NRMI therapy. angioplasty or the nonparametric median Myocardial infarction, Heart hospitals. which the hospital reported data to thecohort. Thrombolytic (Brown–Mood) test for failure, PTCA, CABG, Stroke }, NRMI. Eligible hospitals were ranked therapy. comparisons of median Cardiac risk factors { Diabetes according to volume, and the 25th, values. Forward multiple mellitus, Hypertension, Current 50th, and 75th percentiles were used as logistic-regression models smoking, Family history of cutoff points to define quartiles. were developed to identify coronary, artery disease High predictors of mortality. The serum cholesterol}. C statistic to determine the predictive value of each fully adjusted model.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 107 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Sex,Variables Diabetes, contrôlées Emergency, Risk adjusted Résultats excess coronary étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) The continuity-adjusted X² artery bypass grafting mortality d’étude High and Low volume OSHPD PDD is an test was used to determine Shock CVD, COPD, PVD, In-hospital Procedural in California is related to the CA - CABG ҆( 300/year ; < 300/year) administrative database Renal Failure, Dialysis, Prior Cardio, volume differences in rates between large number of low-volume 6 physicians to mortality and aggregated from 2-page CABG, CHF or Low EF. 62 1999- US, Coronary (CABG and CA - PCI ҆( 600/year ; < 600/year). California and New York. programs. No volume effect the 9 member Risk factors patient discharge abstract 2001 California revascularization PCI). Traumawas noted admissions for risk adjusted conform Cardiac rates. produced from the clinical and New- (CABG and PCI toPCI. a sine-wave Excess PCI pattern intervention with a Advisory record of every patient Database contains York procedures). 3,5-foldmortality increase might be in related admissions to Board, discharged from an acute orThe statistical Package for information including patients Voluntary database, analysis betweencase selection morning or timingand evening of 138 hospitals The total number of admissions duringsubacute-care facility. the Social Sciences versiondemographics, facility pro le, Optimizing of the National Trauma Data fi hours.intervention. Therefore, additional in California, Trauma care. Outcomes a 24-hours period. The lowest numberBank (NTDB). 12 was used for all statisticalscene, emergency departement 421 997 Physician resources needed in the hours US 39 hospitals in of admissions (7,504) occurred between analyses. Frequencies and (ED) was categorized as dead 63 1999- parameters Staffing/ around 7:00 PM and later. patientsNew-York. 5:00 am and 6:00 am, percentages were used on arrival, died in ED, operating 2002 territories, included length Trauma Educational and administrative 268 trauma The Highest number of admissions to describe the pattern of room (OR), Intensive care and the of hospital stay, Admission activities are best scheduled centers in 36 (26,373) occurred between 6:00 pm and admissions. Comparison of units (ICU), telemetry, oor, District of length of ICU Volume. fl during low volume morning states. 7:00 pm. Columbia stay, total days means and odds ratio were ED observation, and burn unit.hours. Night time admissions on ventilatory used to compare subgroupshospital, inter hospital transfer,are higher risk and thus support, and in- on salient variables. procedure type, geographic more likely to need senior- hospital survival. location and day and hour level expertise and consume of admission and other hospital resources. Trauma temporal information, type of centers should use these injury severity (blunt trauma, findings to improve outcomes penetrating trauma, or burns). by developing optimal staffing patterns and matching resource allocation to need as a function of time.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 108 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôléesThefindings Résultats of this study do étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Patients demographics (age, not support the position that d’étude sex), preexisting comorbidities,higher trauma center volumes Logistic regression modeling Retrospective analysis was prehospital care, hospital are associated with improved Trauma care. In-hospital Trauma centerTotal Number of Patients With Major was used to analyze the 64 1994 - US 67 Hospitals based on data obtained from demographics characteristicssurvival. The implication of survival. volume. Trauma by hospital Volume Quartiles relation between survival 1999 29 States the National Trauma Cohort {(trauma center designation : this study is that the hospital (for blunt trauma and penetrating rates and hospital volume 181,371 Databank for 1999. blunt trauma, penetrating volume criteria established trauma) for patient sustaining either Patients aged trauma) ; Hospital size, by the American College of { < 140 ; 140 -261; 262-462 ; >462 }. severe blunt or severe 18 years of age teaching status (University, Surgeons may need to be penetrating trauma. The or older. Community, Nonteaching)}, reexamined. No association models were evaluated by severity scoring components.could be demonstrated assesment of discrimination between trauma volume and (calculated of the area under outcome for blunt trauma, A the ROC receiver operating separate multivariate analysis characteristic) and calibration of patients with penetrating (assessed using the Hosmer- trauma also could not Lemeshow goodness-of-fit- demonstrate a signi cant statistic. fi volume mortality association.

65 1990- US 172 VA Medical 2000 Centers ; 34,888 Greater hospital surgical individuals Patients demographics (age, volume is an independant with a new sex, year of diagnosis) ; Race/ Overall - 5 years cumulative predictor of prologed long-term diagnosis of Cohort study ethnicity, Marital status, Deyo Hospital The median value (50th percentile), as survival was calculated survival following surgery for CRC during Colorectal Long term The log-log survival plots Score, Metastatic disease, surgical the cutt-off value (25 procedures per from Kaplan-Meir estimates, both colon and rectal cancer Cancer. survival. used to examine the CRC site (colon, Rectum), volume. year). Surgical volume was assessed as adjusted risk of death was in the VA health-care system. proportionnal hazards Chemotherapy, Radiotherapy. a categorical variable (< 25 procedures, estimated using Cox The volume-long term mortality assumption, which was met ҆25 procedures per year). proportional hazards models. relationship is greater for rectal in all models. X² test were conducted to than for colon cancer patients, compare demographic and perhaps reflecting the fact that clinical characteristic of surgery for rectal cancer is FY and 22,633 patients who had surgical more technically demanding. surgical resection for CRC compared resection. to patients who didn’t have surgical resection.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 109 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôléesStudy con Résultatsfirmed the Volume/ étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) d’étude Patients and procedure Complication relationship for characteristics : age, sex, inhospital CABG, it did not 1-month, and 6 month Study was a population- race, length of stay during reveal an association between Cardio, PCI Annual volume was categorized as : events was estimated by use 66 1944- US, 25,222 patients * Revascularization Annual PCI based cohort study that index (PTCR > 7) ; Multivesselvolume and postdischarge Philadelphia- (percutaneous * < 400, of multivariable analyses 1995 43 Laboratories. (follow-up (Laboratory) analyzed data from procedure by PCI code events. These results suggest US coronary * 400-599, adjusting for patient and hospitalization forvolume. Pennsylvania Health Care (PCI > 6), ASG, MI (None, that inhospital complications interventions). PCI or CABG) * 600-899, procedural characteristics. Cost Containment Council early MI PCI, Post MI PCI), will remain the standard for * MI as a primary * ҆900 Robust variance estimates (PHC4). Insurance (Medical assistance,assessing laboratory volume or secondary were computed for all survival procedures per year. Medicare, Commercial/other),and that selective use of diagnosis code, analyses and generalized Device (Stent). higher-volume laboratories * Death. estimating equation logistic may not improve long term regression were used for all outcomes. Higher-volume inhospital events. Control for laboratories did not exhibit confounding was provided by lower rates of revascularization use of multivariate Cox and or adverse cardiac events at logistic regression with all 1or 6 months after hospital covariates in the database. discharge, despite their lower risk of inhospital CABG.

67 1982- US, 100,963 LEAB ; 1994 California 106,493 CEA ; 35,130 unruptured In-hospital mortality rates for AAA ; Using multiple logistic Characteristics : Age category,CEA, LEAB and unruptured 7,327 ruptured The data in this study were gender, year of surgery, The number of surgeries categorized regression to control for risk AA have been significantly AAA primary Cardio, vascularIn-Hospital : Hospital analyzed for risk factors whether admission was into : factors and time trends. decreasing over time. Mortality procedures. * LEAB (lower mortality rates. procedure collected in the California througth the emergency * Low volume (under < 20 surgeries for is inversely related to hospital extremity volume. administrative data bases, room, and number of hospital the surgical procedure during the year) volume and directely related namely age, sex… surgeries for that procedure arterial pass) ; * Moderate volume (20 to 49 surgeries) to patient age and emergency during the year. Selected * CEA (carotid * High volume (50 or more for AAA status. Mortality trends over diagnostic Comorbidities endarterectomy), and 50 to 99 and 100 or more for time for ruptured AAA remains (diabetes). * AAA LEAB and CEA). unchanged ; however, mortality (Abdominal is less in high-volume hospitals. aortic aneurism Coronary angioplasty (PTCA) repair) ; has not an impact on rates for * Lower LEAB. extremity angioplasty, * CABG.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 110 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôléesPatients with Résultats AMI treated étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Baseline Charateristics : at hospitals with high or d’étude Developed additional Retrospective cohort study. Demographic (age, sex, intermediate volumes of regression models to test primary agioplasty had Patient with Hospital classified into Primary The study population divided race (white, black, and 68 1994- US 446 acute care * In-hospital Hospital for an interaction between lower mortality with primary AMI treated angioplasty volume quartiles based oninto 2 cohorts based on other), Medical History 1999 hospitals mortality primary hospital primary angioplasty angioplasty than thrombolysis, with Primary the annual number of procedures eachwhether the patients received (Diabetes, Hypertension, 62,299 patients * nonfatal strokeangioplasty volume and reperfusion whereas patients with AMI angioplasty VS had performed. intravenous thrombolysis or Hypercholesterolemia, Current with AMI. procedural modality. treated at hospitals with low thrombolytic (both * 112 classified as Low volume primary agioplasty. smoker, Prior angina, prior Volume. angioplasty volumes had therapy. hemorrhagic (҅16 procedures), CHF, prior MI, piror stroke, and * 223 as intermediate volume prior angioplasty, Prior CABGsimilar mortality outcomes thrombo- (17-48 procedures), surgery), clinical presentation,with primary agioplasty or thrombolisys. embolic), other * 111 as high volume medical therapy, and hospitalLow and intermediate volume major bleeding ( ҆49 procedures). variables. operators can have good following Patients demographics The patient was the outcomes in a high-volume reperfusion Study performed a post hoc (age, sex), coronary risk sampling unit. Standard hospital environment. The ACC/ therapy, and Operator Operator volume according to the analysis of prospectively factors including diabetes, multiple logistic regression AHA guidelines do suggest that Complication subsequent volume average number of PCI procedures gathered PCI data, in the hypertension, previous history 69 2000- US 51 Physicians,Cardio, PCI was implemented, where intermediate volume operators rates performed during the 3-year period, current era of ubiquitous stent of ischemic heart disease, 6,510 PCIs. (percutaneous the physician volume was can continue to do PCI in a 2001- revascularization each operator was categorized into one elevated serum cholesterol, coronary * death, use, at two teriary cadiac high volume > 600 procedures 2002 of three groups ; * Low (< 50 cases per entered along with other smoking, family history of interventions). * myocardialduring the index care centres. per year. Study suggest yearly year), * Intermediate (50-74 per year) ; factors. coronary artery disease, serum infarction, hospitalization. operator volume, although or high volume҆ ( 75 case per year). creatinine, as well as hospital * coronary very important, should not be length of stay, and cost of care. perforation, the only factor used in setting Number of physician, total * emergent standards for interventional volume of cases from 2000 to coronary artery cardiology certification or as the 2002, Mean complexity score bypass surgery full measure of quality. and pericardial per operator, ICVL direct cost, tamponade. ALOS, Major complications.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 111 -

Pays et/ Careful decision making, Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour dé nir selection of low-risk patients for fi Pre-selected variables : age, N° ou région Méthode Variables contrôléesa low volume Résultats and limited facility étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) gender, indication {Stable d’étude location resulted in excellent Cohort data. Used Risk angina, Unstable angina, in-hospital survival with very Retrospective study assessment. Cumulative PostMI angina}; History (of MI, CABG. Short-term Surgical Fully equiped university location (FE) low complications rates. 70 1995 - US ? 832 Patients; investigates the short and survival curves were Coranary angioplast, Obesity}; hospitals volume. and 1996 52 redo CABG mid-term outcome of non- constructed using the Kaplan-Risk factors {Diabetes, Complication The low volume limited facility location operations; emergent primary isolated Meier method. Among patientHyperlipidemia therapy, and 1 & 5-year (LVLF) of departement. 2 Hosptials. cCABG surgery in relation tosubgroups the log-rank Hypertension, COPD, Renal in-hospital risk stratification in FE and test was used to compare impairment}; Angiographic outcome : LVLF. The applicability of survival curves. A logistic factors {Number of diseased * mortality the Parsonnet score and multivariable regression, vessel, 1or 2 or 3 vessels, Left the EuroScore models wasstrati fied by the two hospitals,main stem, Ejection fraction *Acute myocardial tested retrospectively to ourwas used to adjust for the (normal, reduced or moderate, infarction study population. baseline differences to poor)}. * Cerebro determine the impact on in- vascular hospital mortality and hospital accident complications. To determine * Prolonged independant predictors ventilation of 1- and 5- year survival * Reanimation multivariate analysis was carried out using stepwise * Rethoracothomy 71 2001 - US 70,757 Cox regression. for bleeding High ICU patients volumes are 2003 Patients, associated with lower mortality * Wound 92 ICUs. rates in High-risk critically ill complication Hospital characteristics adults. Regionalisation of adult * Intra aortic { Geography (Urban, Suburban, Retrospective cohort study. Hierarchical logistic critical care may result in better ballon pump. Rural), Affiliation with medical regression modeling was school, Affiliation with primaryoutcomes for high-risk patients. ICU patients ICU volume traeated as a categorical Cardio, ICU typeIn-Hospital : used to examine the volume-critical care fellowship, Trauma * Cardiothoracic volume. variables. ICUs divided into volume Mortality rate. outcome association and center designation}, Bed ICU only quartiles (based on annual volume) with it allows the simultaneous Number, Hospital organisation, * Coronary care/ equal number of ICUs in each quartiles : assessement of patient- trauma center, Medical school Medical cardiac Low volume (<632 ICU admissions/ level risk factors and af liation} ; care only year) ; medium volume (632-827), High fi ICU characteristics while ICU characteristics{ Annual volume (828 - 1233), and very High controlling for the correlation * Medical ICU only volume (>1233). High risk volume patient volume, median, Annual of outcome among patients High-risk volume, median, * Medical ICU/ was defined as the annual volume treated within the same ICU coronary care, In-Hospital mortality rate, of patients with SAPS II score >41 (clustering). The C-statistic combined median ; ICU Type, Number of corresponding to the 75th percentile of was used to evaluate model * Medical ICU/ ICU beds median} ICU model, the patient risk distribution. High-risk discrimination, and the Coronary care/ Day coverage, Night coverage, volume groups were defined as follows : Hosmer-Lemeshow statistic specialized low volume (>133 ICU admission/ CCm fellowship/rotation, bed * Medical/surgical was used to examine model year), medium volume (134 - 216), number, Nursing policy ; combined calibration. high volume (217- 295), and very high Patients characteristics * Medical/ volume (>295). { number of patiens, ICU number, Age, Gender, SAPS surgical/coronary II mean}. carecombined * Neurologic/ neurosurgical ICU combined * Other mixed

* Surgical ICU only * Surgical/trauma ICU combined.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 112 -

Patients with AMI who are Pays et/ treated by high-volume Patients characteristics: Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir admitting physicians are more N° ou région Méthode age,Variables gender, predictedcontrôlées 30- Résultats Multivariable analyses of likely to survive at 30-days and étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Retrospective cohort study day or 1-year mortality, social d’étude the physician volume effect 1 year. Physician * Hospitals were classified as low using linked administrative economicus status, Income, were also conducted using Cardio, Mortality risk Volume and (҅33 AMI cases per year), medium databases containing patient Clinical status admission ; 72 1992 Canada, 98,194 patients random effects hierarchical AMI (Acute rates for 30 daysHospital AMI (34-99 cases per year), or high admission information. Physician Characteristics: Ontario - 5,374 logistic regression models -1998 and 1-year after Volume. (҆100 cases per year) volume. Speciality (family Physician, physicans. Myocardial that adjusted for patients and Canada or post-AMI * Annual Physician volumes : General internist, Cardiologist), Infarction). physicians characteristics. (Crude mortality (AMI cases per year) age, sex ; The multilevel models were t and risk-adjusted * 1 to 5 ; Hospitalfi Characteristics: using the software package mortality) * 6 to 13, hospital volume, teaching, MLwiN and took into account Length of stay. * 14 to 24, availability of on-site the hierarchical nature of the Physcian* and more PCI than volume 24 AMI was cases calculated treated by revascularization facilities, Physician data, with patients clustered counting per year. all 1997 Medicare claims for Rural hospital. Patients and Hospital within physicians and Cardio, PCI angioplasty, stent, or artherectomy. 73 During US 167,208 Outcome Volume rates physicians clustered within (percutaneous Hospital PCI volume was calculated by 1997 patients aged * Rates of of Medicare hospitals. coronary counting all claims for coronary 65 to 99 years CABG surgeryPCIs. interventions) angioplasty, stent, or artherectomy. 6,534 folowing a PCI, in the Era of Physicians and Hospitals volume physicians 30-day mortality Coronary Stent. stratified into : low, intermediate, and 1 003 Hsopitals. *The combined high volume categories. Medicare patients treated end point of Analysis of data from Patients characteristics andPatients characteristics : by high-volume physicians CABG or Medicare Claims History lesoutcomes were compared age, sex, race, urgency of and at high-volume centers 30-day mortality fi used both Part A (hospital) across volume strata. Logisticadmission (elective, urgent, experience better outcomes during the index and Part B (physician) claims.regression models were usedemergency), acute myocardialfollowing PCIs. episode of care. to adjust for patient infarction (AMI), comorbidity, characteristics while and the number of vessels assessing the relationship. revascularized (single vessel or Used X² test to assess multivessel PCI). Comorbidity differences across categoriesusing the Dartmouth-Manitoba and analysis of variance testmodi fication of the Charlson were used to assess Physician Medicare PCI volume per Comorbidity index score differences among means. Year : (0 ; 1 or ҆ 2); Stent use. Following multivariate (< 30; 30-60; > 60). Hospital Medicare PCI volume per Year : analysis (< 80; 80-160; > 160). with logistic regression, direct standardization techniques were used to generate adjusted rates of outcomes adverses.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 113 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir In general, outcomes were N° ou région Méthode BaselineVariables Characteristics contrôlées : Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) better in high volume centers, age, sex, race, risk factors d’étude Univariate differences although in the stent group, {Diabetes, Hypertension, Study from both randomizedamong low, medium, and there was no clear relationship * Same Cerebrovascular disease, CABG by Stent Patients were subdivided into two trials and administrative high volume institutions between volume and out 74 1996 US 100,318 admission Patient Chronic pulmonary disease, and PTCA. distinct groups : with a principal databases. Data were were compared using the come. These results support 19 states angioplasties hospital death with and Renal disease, Mutivessel diagnosis of AMI and without such a derived from the NationwideChi-Square for trend statistic. earlierfindings that hospital 191 hospitals * same without AMI angioplasty}, Primary Payer diagnosis. Inpatient Sample (NIS) Outcomes for stents and mortality and particularly same 43,966 admission (codes410.xx) {Medicare, Private insurance, Hospital were categorized as low from Health care Cost and conventional PTCA were admission surgery rates are involved stent CABG surgery and hospital Medicaid, Other}; State. volume if they performed҅ 200 cases Utilization. Project Release compared with the Chi- lower with stenting. Altougth the placement. * length of stay volume rates. per year, medium volume as 201-4005, which contains hospital Square test for categorical volume outcome association The results of analisys con rm cases per year, and high volume as discharge abstract data. variables and the Student’s for stenting was less clear thanfi that pancreatic head resection * Median charges. >400 cases per year. t-Test for continuous ones. in Medicare these result do Quantitative : Annual numberscan be performed with low TreatmentLogistic regression strategies was vary not mean that the fundamental of different types of surgery mortality in specialized units. Questionnaires were frequently,used to determine partially if because stent volume outcome relationship Pancreatic HeadMortality rates Hospital Annual Hospital Volume ranged from 14 for each malignant and Variations in indications, completed by seven high- ofassociated the lack ofwith evidence- better has been changed by stenting. 1,454 Resections. after PHT. volume. to 52 (10 to 43 in Malignancy) cases/ benign disease ; after PHR ; operative technique, and Germany : volume surgical departmentsbasedoutcomes. data. Performed a 75 1999- Pancreatic year. and morbidity including perioperative care may Berlin, regarding quantitative multi-institutional analysis 2004 Head reoperation, type and techniquedemonstrate the lack of Dresden, and qualitative aspects of evaluating current numbers Resection, 944 of pancreatic anastomosis, evidence-based data and/ Essen, pancreatic head resections indications, techniques, and for malignacy, use of abdominal drains, and or personal and institutional Freïburg, in the period from 1999-2004complications rates of PHR. 7 surgical median lenght of postoperativeexperience. The low number Göttingen, (Five prospective and two departments. hospital. Evaluated the of patients receiving adjuvant Mannheim, retrospective institutional percentage of different tumor therapy after resection of Münster database). entities, of superior mesentericpancreatic cancer suggest that portal vein resection, the typemore efforts must be made of lymphadenectomy. The typeto establish novel adjuvant and frequency of neoadjuvanttherapies under randomized and adjuvant therapies study conditions. after resection of pancreatic adenocarcinoma.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 114 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) d’étude There were no clinically For major resection : age, sex, Individual hospitals were placed into significant differences in Hospital comorbidity score 76 1990 Canada, 1,072 patients Cancer Rectal Operative low-volume (11 or less), medium- Administrative database Volume appropriate, chi- treatment measures operative procedure square, Kruskal-Wallis, or For resection with pathology mortality, and long-term survival Ontario incident cases cancer. mortality and volume (12 to 17) or high volume (18lacked or details on the presence volume or Wilcoson two-sample testingavailable : N° of cases, Node among the hospital groups of rectal long term overall more) groups based on the number ofof metastic disease, use of adenocarcinoma Teaching measured for differences positive, Full wall invasion, according to both univariate survival. major resections perfomed during theadjuvant therapy, incidence of Hospital among the groups in patientModerate differentiated, and multivariate analyses. In study period. local tumor recurrence, and diagnosed (Major status. and tumor variables and in Ulcerating morphology, conclusion, the absence of a cancer-related mortality. resections) ; treatment and outcomes Vascular/lymphalic invasion. hospital volume or teaching 418 procedures And into teaching (training of physicians) measures. Logistic status effect on treatment and with pathology or nonteaching institutions. Volume regression models analyzed outcome measures suggests available cutoffs created three evenly sized operative mortality, whereas that for rectal cancer surgery, 124 hospitals. groups. proportional hazards models centralization of procedures assessed long-term survival. into high-volume or teaching Patients ages, clinical state, Unplanned readmission within The effect of independant centers is unlikely to improve The category and volumes histologic differentiation and co-30 days following discharge variables on the readmission surgical quality. Breast cancerHospital divided into three categoriesof hospitals and surgical morbidity (Absent, present). were significantly higher in Korean 1,351 patientsBreast cancer Unplanned of patients to hospitals were 77 1994- operation based on breast cancer operation information were collected hospitals with a lower surgical operation. readmission examined using multivariate 1996 in 24 hospitals, cases Hospitalcases: low-volume hospitals (≤ 50 using questionnaires volume. rates within 30 analysis, especially by logistic Surgery performed by 6 nurses. volume. annual procedures), medium-volumedesigned for each category Patient demographics days of surgery. Usingregression univariate analysis. Logistic high-volume surgeons was hospitals (51-99 annual procedures), (age, etnicity), payer status Aof statewidehospital. hospital regression models to associated with a 48% and high-volume hospitals (≥ 100 Insurance, hospital type, 6,181 patients ;Uterine Cancer. In-hospital deathSurgeon and The primary independant variables indischarge database was evaluate for significant factors reduction in the risk of hospital 78 1994- US, annual procedures). Discordance surgeon or rates Hospital case study were individual surgeon uterineused to identify women associated with in-hospital death. Incresead efforts to 2005 Maryland - 894 surgeons ; attending physician, ICU length volume. cancer case volume and individual undergoing primary fot mortality and access to high concentrate the surgical US 49 hospitals. of stay. hospital uterine cancer case volume. Uterine cancer. The study volume care. management of uterine cancer Surgeon case volume was categorizeddesign was a cross-sectional by high volume surgeons at as low (҅99 cases/12 years) or High҆analisys ( of hospital discharge high volume centers should be 100 cases/12 years). data from nonfederal acute undertaken. Hospital case volume was categorizedcare hospitals collected by as low (҅199 cases/12 years) or HighMaryland HSCRC. (҆200 cases/12 years).

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 115 -

Pays et/ There is a relationship Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode CovariatesVariables include contrôlées (Patient’s between Résultatshospital volume and Logistic regression was used étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) This Population based cohort Age at initial surgery, Co - reoperation rate. Institution type d’étude to evaluate the importance study included all newly morbid diagnoses (exclusive onlyof influenced reoperation of hospital type, hospital Hospital and The volumes of ovarian cancer diagnoses ovarian cancer cancer), an indicator variable rate. Statistically significant 3,815 women. Ovarian Cancer. 30-day volume, surgical specialty, 79 1992 - Canada, Physician surgeries by hospital were classified inas all women in Ontario. for metastatic disease, and anassociations were found postoperative and surgeon volume of 1998 Ontario - Volume of 15 surgeries per hospital per year (lowHopitalization and surgical indicator variable for Emergent,between surgeon specialty mortality, ovarain cancer operations on Canada procedures. volume centers), billing databases were used. Urgent versus Elective surgery.and all three outcome outcomes. reoperation 16-99 (Intermediate volume centers), Comorbidity using Charlson variables. Adjusted survival is rate, and overall and greater than 100 (High-volume Comorbidity Score. improved when the operation survival. centers). made by gynnecologist or The number of ovarian cancer gynecologic oncologist. The operations per year by physicians volume of surgery performed showed three zones in the curves : The by an individual surgeon only group of physicians who performed influencedreoperation rate. 10 or more operations per year (high volume), those that performed 3-9 operations per year (Intermediate volume), and those that performed 1-2 operations per year (low volume).

Cancer : SEER- Medicare linked Co-morbidity, age, cancer stage. These data support the 5,013 patients Retrospective Cohort - Pancreatectomy,In-hospital Hospital database. hypothesis that when complex 80 1984- US - aged 65 years study using Surveillance, - mortality rates. procedures 1993 (New York, or older. surgical oncologic procedures volume. Epidemiology and end results California, are provided by surgical teams 32 Hospitals. Esophagectomy, (SEER) registry. Maryland, in hospitals with specialty - Pneumonectomy, Californian), expertise, mortality rates are - Liver resection, England - lower. Adjustement for case-mix - pelvic Scotland and other patient factors did not exenteration. change this results.

Despite individual operator A single center, over 3 yearsBaseline demographics (age volumes below those The clinical and coronory and gender), clinical indication, Cardio, Individual evaluation. Chi-square and currently being considered 81 1993- US 1,389 * Incidence of interventional procedures comorbidities (congestive Coronary operator Fischer test. for credentialing, the overall 1995 procedures; complication ; were prospectively entered heart failure, previous CABG, coronary institutional outcome was 9 operators. Interventions. * In-hospital into a database for all the previous PTCA, or AMI. interventional excellent in a diverse and mortality. patients who underwent Baseline procedural variables volume. complex patient population. operation between 1993- (no of lesions, etc.). 1995.Very detailed data on 82 1997- US 31 academic patients. 1998 level 1 or 2 trauma centers The result indicate that a strong High volume (> 650 trauma admissions/Retrospective cohort study. Logistic (for mortality) andAge, sex, mechanism of PAI admission Trauma care. * Inpatient Trauma association exists between linear (for LOS) regression.injury, transfered from other 478 patients ; Centers year) mortality center, Shock, massive bloodtrauma center volume and volume. Low-volume ҅( 650 admissions/year). * Hospital length transfusion. outcomes, with significant of stay. improvements in mortality and LOS when volume exceed 650 cases per year. The benefits 541 Multisystem are evident in patients at high blunt trauma. risk for adverse outcomes.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 116 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats Age, sex, race (white), étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) New York state trauma Multivariate analyses d’étude co-morbidity (0 or >1), Hospital and Hospital and surgeon volume were registry data. performed to adjust the diagnosis (Osteoarthritis, defined as the number of primary relation with logistic Perioperative surgeon 83 Jan - AugUS 80,904 patientsTotal knee and revision total knee replacement regression. R. arthritis, other), medicaid Patients managed at hospitals mortality and volume. 2000 78,745 primaryreplacement. performed in hospital or by the surgeon eligible. (more than 200 compared with complications in total knee in Medicare recipients. less than 25) and by surgeons with the rst 90 days replacements. fi Hospital volume : greater volumes (more than 50) postoperatively. 1-25 ; 26-100 ; 101-200 ; > 200; of total knee replacement have lower risk of perioperative adverse 5 outcomes : Surgeon volume : events (risk of pneumonia, death, * Death 1-12 ; 13-25 ; 26-50 ; > 50. * Acute MI, infection) following primary total knee replacement. These myocardial data should also be integrated infarction into the policy debate about the * Deep infection advantages and drawbacks of * Pneumonia regionalizing total joint replacement * Pulmonary to high-volume centers. embolus.

Age, age squared, sex, motorUnable to document an inverse Used Logistic regresion relationship between hospital Trauma. Risk-adjusted Hospital Three volume measures used : responses of patient, blood 26,973 to identify significant volume and inpatient mortality 84 1994- US, volume rates.* Total annual volume of trauma cases pressure, ICISS. inpatients inpatient independant predictors of rate for trauma centers in New 1995 New-York ҆1,200 and total annual volume҆240 43 trauma mortality rate. mortality, their weight, and York state. Low volume centers US for patients with Injury Severity Score centers. the probability of in-hospital (not meeting ACS criteria) had (ISS)҆15 mortality for each patient. lower, although not significant, * Total annual volume of patients with risk adjusted mortality rates. ISS҆15, and Used standard logistic Age, sex, surgical complexity In-hospital Hospital Discharge abstract data from * total annual volume of cases in the categories. Cardiovascular volume. Statewide research system.3regression techniques. 85 1990- US, New- 31 hopitals deaths. Registry (inpatient with ISS҆ 9). procedures : cohort patients. 1995 York 97,137 * CABG US operations. * elective repair In AAA operation, a signi cant of AAA; fi inverse relationship between * repair of hospital volume and in-hospital congenital deaths was determined. For cardiac defects. congenital cardiac defects, a significant inverse relationship (which was most pronouced for neonates) was found between volume and death. However, lack of such a relationship for CABG. This observation may be largely explained by the qualtity improvement program in New- York state for bypass operations since 1989. If so, these results have important implications for expanding the scope of quality improvement efforts.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 117 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) d’étude Adjusting for comorbidity for Patients who have a TSA or Retrospective cohort study Logistic regression with HA performed by high-volume Surgeon TSA and HA Hospital volume : charlson index, age, race, level III-2. generalized estimating surgeon (more than 4 a year) 86 1998- US 12,594 patientsTotal Shoulder * Inhospital and hospital < 5 ; ҆5 - < 10 ;҆10 household income, and sex. equations and multiple linear or in a high-volume hospital 2000 total shoulder Arthroplasty mortality rate, caseload Surgeon Volume : < 2҆ ;2 - < 5 ;҆ 5. regression models were used (more than 10) are more likely arthroplasty ; (TSA), * Length of stay volume. to estimate the adjusted to have a better outcome. 17,999 patientsHemiarthroplasty in the hospital, hemiarthroplasty. association between surgeon (HA). * Disposition of and hospital volume and the patient on outcomes fot TSA and HAS. Little evidence that all patients discharge, and Hospital characteristics { Neurowith major trauma do better in * In-hospital surgeon, CT scanner, Traumahigher volume departements, Using the stratified W postoperative Annual case volume of hospital variedProspective follow-up team, N senior doctors}, there was evidence that England 14 English Trauma care. Mortality 6 Annual case statistic, an age and severity 87 1990 - complications. between hospitals from 5 to 96 a year.study of all major trauma Patient characteristics {Age, months after thevolume of adjusted measure of patients with complex needs, 1993 emergency patients (with injury score ISS mean, % died, % with blunt incident. hospital . outcome, the relationship such as the multiple injured or department ; > 15) arriving alive, with no injuries, with multiple injuries, between volume of those with head injuries, had 2,190 patients. invariably fatal injury. % with multiple injuries, % cases and outcome was better outcomes. head injuries, % of case out of initially examined using hours}. the Spearman correlation coefficient. Multiple regression analisys was used to explore further Covariates including age , sex,Hospital volume was not a Thethe relationship, covariates were after entered good proxy for outcome. Low- Hospital volume was derived from Data of this study were mechanism of injury, Injurity Trauma care. Mortality and Patient intoadjustement a multiple for logistic hospital-level volume centers appeared 98,245 patients the annual number of admissions perobtained from the patient Severity Score and Trauma 88 1998- US, length of hospitalhospital regressioncharacteristics. model to assess to have outcomes that were admitted after center : < OR҆ 1,200/year. ISS < 15, discharge data of the State of center designation. 1999 California- stay. volume. inpatient mortality and a comparable to centers with injury, California. US ISS > 15. multiple linear regression higher volumes. Perhaps 38 Lever I model to assess hospital institutional outcomes rather (n = 12) or length of stay. than volumes shoul be used II (n = 26) as a criterion for trauma center trauma center. verification.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 118 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Analized claims data of d’étude Hospital and Analyses of Hospital volume were Medicare recipients to Outcome adjusted for surgeon volume, and identify cases of primary Total Hip Surgeon (҅90 days inversly. and revision total hip 89 1995- US 58,521 replacement Volume. For each outcome, Age, gender, comorbidity index,Patients treated at hospitals postoperative) Divided hospital and surgeon volumereplacement. 1996 procedures Primary (PTH) constructed 2 sets of speci c arthritis diagnosis, and by surgeons with * Mortality, into clinically sensible categories. fi for elective and Revision multivariate models, one Income, and Medicaid eligibilityhigher annual caseloads of * Dislocation, For PTH: primary (PR). examining the association indicator. primary and revision total hip * Deep infection, Surgeon volume : total hip between outcomes and replacement had lower rates * Pulmonary 1-5 ; 6-10; 11-25; 26-50; >50; replacement hospital-volume strata and of mortality and of selected embolus. Hospital volume : or 12,956 the other examining the complicaitons. These analyses 1-10 ; 11-25; 26-50; 51-100; >100. procedures association between of Medicare claims are limited revision outcomes and surgeon- by a lack of key clinical total hip volume strata. The stregth of information such as operative replacement. the association between details and perioperative functional status.

volume and outcome expressed with an adjusted odds ratio, and the test for linear trend was performed with an ordinal variable representing the volume stratum. Used additional logistic regression models of the surgeon effect within each * Length of 9,973 patients,Rotator Cuff hospital volume stratum. 90 1997- US, New- hospital stay, Covariates : Patients 4,146 surgeonsrepair. Multivariate regression 2000 York * Disposition of Surgeon and Hospital volume were characteristics (age, sex), less 15 Hospital and models were used to Patients operated by low volume US divided into Low, Intermediate, and High Mean Deyo score, and patient procedures. Surgeon estimate the risk-adjusted surgeons had significantly higher volume categories. comorbidity Charlson index. Volume. association between provider probability of an extended stay, Surgeon volume : < 15 ; 15- 29 ; ҆30 Mean number of diagnosis on volume and outcomes. longer times in operating room, patient discharge procedures. this discharge. compared with those operated * Operating room Hospital volume : < 75; 75-199; ҆200 time. by high volume surgeons. procedures. Used HES database for Surgical Annual Volume of Hospital trust into clinically sensible categories for eachEngland in NHStrust between Adverse volumes of UK 281,360 Hip Total joint intervention : 1-50 ; 51-100 ; 101-250 ;fi nancial years. In England, there are fewer 91 1997 & outcomes Hip/Knee Used Logistic regression Age, sex were controlled for as Replacement : 251-500 and >500. adverse events following TJR 2002 replacement ; * 30-day replacement. analisys to investigate oddspotential of confounders. Hip (THR) and in High volume centres and in 211,099 knee Inhospital in-hospital death within 30-daysType of admission (Elective, Knee (TKR). orthopaedic training centres. replacement. mortality, of the operation. Univariate Emergency, Unknown), Adverse events are rare, but * length of stay, logistic regression looked atCharlson Co-morbidity Index, significant relation between * Readmission the association of the outcometrainig centre status. with each exposure, and a adverse outcomes and low within a year, multivariate regression analysis volume units. * Rate of surgical was carried out controlling for revision within confounders. Likelihood ratio 5 years. tests examined interactions. Poisson regression modelling was used to calculate the rate of readmission within a year.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 119 -

Patients treated by providers Pays et/ Models were adjusted for Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir with lower caseload N° ou région MéthodeLogistic and multiple comorbidity,Variables age, contrôlées gender, Résultats Prognostic study, Level II-1 volumes had higher rates of étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) regression models were usedrace, household income, d’étude The Hospital volume, was the total retrospective study mortality following total knee Hospital and to estimated the adjusted and procedure, Discharge * In-hospital number of primary and revision total Performed a secondary arthroplasty. Proposing volume Total Knee Surgeon association of surgeon or disposition. mortality rates knee arthroplasties performed in eachanalisys of a analysis of standards could decrease 92 1997 US 277,550 Arthroplasty- volume hospital volume with rates * In-hospital hospital or hospital system, divided intoa national administrative patient mortality following this patients Primary and procedures. of in-hospital mortality, postoperative four groups: database. procedure. (50,874 Revision. pulmonary thrombosis, complications < 85 ; 85-149 ; 150-249 ; ҆250 inpatients deep venous thrombosis * length of stay in procedures. Primary in the lower extremity, and the hospital. Surgeon volume : < 15 ; 15-29 ; 30-59 ; total knee postoperative wound infection ҆60 procedures. Arthroplasty ; Cross-sectional study basedas well as length of hospital Surgeon Surgeon volume divided in low, 4,636 Bariatric Readmission on data from inpatient stay. 93 2003 New-York 7,868 Bariatric and Hospital intermediate, and high volume inpatients operation. at 30 days of discharge database. procedures volume. procedure/year :҅ 25; 26- 150 ; >150. US Revision discharge. patients ; Hospital volume, number of procedures/ total knee 597 Bariatric year : Arthroplasty). procedure ҅100 ; 101-200 ; 201-300 ; >300. patients readmitted. Chi-square and Fischer’s Controlling for demographics There is an important exact test were used to (Age, gender, race) ; relationship between surgeon examine the bivariate comorbidity and length of indexand hospital volume and short- relationship between hospitalization. term readmission after bariatric nonordered categorical factors operation. and readmission rate. The Cochran-Armitage test for trend was used to determine if there were any linear trends in readmission rates for ordered categorical variables. A multiple logistic regression model was constructed to examine the association between provider volume and readmission within 30 days after adjusting for patient and postoperative characteristics. C-Statistic was used to assess the predictive ability of the models.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 120 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Controlling for demographics The likelihood of postoperative d’étude (Age, gender, race or ethnicity)complications ; from bariatric Chi-Square analyses were Using New-York State’s and comorbidities. procedures is greater for Surgeon Surgeon volume cutpoints of 25 ; 50 ; used to examine the bivariate Bariatric Postoperative inpatient discharge database. patients with low-volume 94 2003 New-York 147 Surgeon, and Hospital 100 ; 150 procedures. relationship between patient operation. complications Using a Cross-sectional study surgeons or in low-volume US 7,868 Bariatric volume. Hospital volume cutpoints of 100 ; 125 ; risk factors and postoperative rates. based on data from New- hospitals. procedures 150 ; 200 procedures. complications and to York State’s 2003 inpatient patients, improve the parsimony of the discahrege database. 78 Hospitals. multivariable models. After separate multiple logistic regression, equations were constructed for each of the surgeon and hospital volume threshold. C-Statistic was used to assess the predictiveAdjusted age, sex, race, the This study suggest that hepatic ability of the models. number o f comorbidities. resections performed at high- Maryland discharge databaseThe statistical analysis of volume center are associated Surgeon Discharges were stratified into high andwere used to study. dichotomous and categorical with improved safety and * Average length Hepatic or provider low volume provider groups : variables using the Chi- similar total charges compared 95 1990- Maryland - 606 liver of stay, Resection Volume. High volume : >15 / year ; Low volume : square statistic. Continuous. to low volume hospitals. US resections * Average total 1996 (Liver). ҅15/year. variables were compared Conclude that Hepatic performed ; hospital charges, An alternative strati cation performed using Student’s test. resection can be performed 52 nonfederal fi * In-Hospital for relative risk analysis of mortality One assumption of linear more safely and at comparable acure care mortality. High volume : >15 / year ; Medium regression models is that cost at high-volume referral hospitals. volume : 7-15/year ; continous variables must be centers. Low volume : < 7/ year. normally distributed. Multiple linear regression models were performed to assess the relationship between hospital volume and average length of stay and average total charges, adjusted for age, race, sex and the number of comorbidities. Poisson regression used to examine the relationship.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 121 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) d’étude These results of this study Hospital Prospective study. Univariate comparisons suggest that the expected Total Hip * Inhospital 96 1991- Minnesota -1,810 patients and surgical used Chi-square test for effects of seeing better Arthroplasty. complications 1992 US admissions Volume. nonparametric variables outcomes among teaching 27 hospitals. * Posthospital and t-tests for parametric hospitals and among outcomes at variables. Multivariate institutions and surgeon who the time of the analyses used logistic did higher volumes of THAs did follow up. regression for dichotomous not occur systematically. For variables specific cementless prothesis group, complications and ordinary significantly more operative least-squares regression for complications were associated continuous variables. with being in HMO or with insurance other than Medicare. General complications were assiociated positively with a higher caseload per surgeon for patients receiving cemented protheses. Hospital volume had no significant relationship to the Large number of confoundersInverse relationship between 28-daygeneral mortality complication and therate. Hospital Hospital : <1,000 births versus >1,000 1) Kaplan-Meyer survival including data on pregnancy, Very low birth Neonatal volume of perinatal unit. 97 1991- Germany 4,379 infants volume. births ; NICU : < 36 VLBW versus > 36 estimates for infants from perinatal conditions, maternal weight. mortality. The relationship was more 1999 ages 24 to VLBW admissions. small versus large NICU, andrisk factors. 30 week and compares them with log-rank significant with perinatal unit <1,500gr in test. Controls for confounders (NICU) volume rather than total 107 delivery (case-mix). hospital birth volume. units. 2) Multivariate Generalised logistic regression for controlling Modeled as a logistic This experience defines a Low average SVI : 1 - 3 cases per the effect of clustring. US 2,255 patients,Breast Cancer Failure rates for Surgeon learning curve for lymphatic 98 1994- month ; regression curve. Sentinel Lymph each surgeon. monthly mapping in breast cancer 1998 16 Surgeons. Med Average SVI : 3 - 6 cases ; Node Mapping. volume. patients. Data suggest that High Average SVI : > 6 cases. increasing volumes lead to decreased failure rates. These data provide surgeon performing SLN biopsy with a new paradigm for assesing their skill and adequacy of training and describes the relationship between volume of cases performed and success rate of SLN detection.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 122 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) d’étude California’s hospital system is Hospital volumes (compared Hospital Examines the impact of Leapfrog’s far from being regionalized. AAA, Procedure to Leapfrog criteria) are volume. volume criteria for metropolitan areas: Although academic hospitals 99 2000 California-Nonacademic CABG, volume as a analysed to examine the 500 for CABG; 30 for AAA; 100 for CEA; appear better positioned than US hospitals : 705 ;CEA, proxy for quality. impact of regionalisation in 7 for ECR. nonacademic hospital. The Academic ECR. California. hospitals : 60 ; vast majority of all hospitals do CABG (32,210), not meet Leapfrog’s volume AAA (3,389), criteria. As efforts to use CEA (11,432) ; volume as proxy measure ECR (335) of quality gain momentum, 4 procedures. hospitals and physicians will be forced to measure and report quality. As such, surgeon need to decide between accepting volume as an adequate measure of quality and devoloping other possibly more direct and reliable methods. Using X² test to compare Injury severity. Regardless of overall volume 100 1990- US 30,930 patientsTrauma care. Mortality rates. Volume of Classified as : Low volume (LV, 100-500 proportion of patients with of patients encountered, there centers. cases) ; 1994 records, is a consistent proportion of Mid volume (MV, 501-1000 cases) ; high severe injury and mortality. 37 hospitals. severe injury. The increasing volume (HV, > 1000 cases). mortality with the most severe injuries seen in the high volume centers may reflect overdemand on resources. Surgeon Years of experience. Outcome after trauma seemed Trauma care. Overall mortality ; 101 1995 andLouisville- * 2,230 trauma annual trauma to be related to severity of overall 1996 US admissions ; volume. injury rather than annual 14 surgeons ; complication rate volume of cases per surgeon. Length of stay * 2,207 trauma per attending admissions ; surgeon. 13 surgeons.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 123 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) d’étude Retrospective Cohort study. Generalised models takingAge and gender. There is a volume outcome Surgeon Volume modelled first as a continuous Cataract surgery. Postoperative into account clustering of relationship for cataract surgery volume. variable. Then in 4 category : 1) 50-250 Canada Retrospective and this relationship persist even 102 2001- adverse events 2) 251-500 3) 501-1,000 4) more than patients. for very high volume surgeons. 2003 cohort. 284,797 until 14 days 1,000. cataract after surgery. Ex : surgeries for > vitrectomy, lens 20 years old. In extraction,fluid 70 hospitals or exchange… eye centers. Study indicates that the Low - Medium - High. Administrative data - Cohort Adjusted patient Surgeon patients of surgeons with Outcomes : demographics : age, gender, 103 1994 - Maryland - 377 surgeons; Shoulder and Hospital Study Characteristics. higher average caseloads of * Risk of race, marital status, insurance 2000 US 1,868 patients.Arthroplasty. volume. total shoulder arthroplasties complications Type (Medicaid, Medicare), and hemiarthroplasties have rates ; income ; comorbidity score by decreased complications rates * Length of stay ; Charlson index, Diagnosis. and hospital length of stay * Average Total compared with the patients of Hospital surgeon who perform fewer of Charges. these procedures. 104 4 years Niagara Period- Falls- US 1997 Despite having a very low 486 thoracic Thoracic : Operative Hospital Retrospective review of Age, length of stay ; stage volume of thoracic surgical surgical pneumonectomy,mortality volume. thoracic surgical operations. of disease, benign disease cases the community procedures ; esophagectomy. (elective, emergency). * In-hospital death, hospital had crude outcomes 10 * 30-day mortality. comparable to those reported esophagectomy. from high volume tertiary hospitals. This suggests that the surgeon may be a more important factor in the hospital volume-outcome relationship than previously throught. Nevertheless, complex thoracic surgical operations are ideally performed by an experienced surgeon, and in a high volume hospital.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 124 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Physician, Practice, Patient andPhysicians with high-volume d’étude Time use during patient visitsVisit Characteristics, By patientpractices are more efficient Data on the contents of Physician Physician with High, Medium, and Low as measured with the DavisVolume. than those with low-volume outpatients visits were Clinical volume volume pratices. Observation Code (DOC). A practices in providing similar 105 2 days Ohio - US 3,893 patients,Primary care collected using direct outcomes : Patients Patient volume. one-way analisys of variance services in a shorter amount 138 practice : observation, patient and The rates of volume. was used for continous of time. This greater apparent Physicians. Nursing. physician questionnaires and preventive outcome measures and efficiency may come at a cost medical record review. services Chi-square statistics were of lower rates of preventive delivery, patients used to compare categorical services delivery, lower patient satisfaction, and measures. satisfaction, and less positive time use during doctor-patient relationship. patients visits. Health care plans and physicians setting productivity goals should consider the trade-off inherent in high- volume practice.

106 1998- 18 States 1999 - of US 2000 Controlling for : Patient age, Patients with a diagnosis of SAH on their discharge records Hospital case volume Quartiles. Retrospective analysis, Both univariate and sex, Medicaid status, hospital 16,399 patientsAfter SAH In-hospital Hospital who initially presented through administrative database. multivariate analyses region, Data source year, Hospitalizations(Subarachnoid mortality rate. volume of the emergency department were used to assess the comorbidity conditions index. for SAH ; Hemorrhage). SAH cases. of a hospital with a high case volume-mortality rate 9,290 admitted volume of SAH cases had relationship. through significantly lower mortality Proportions across the emergency rate. Concentrating care for this departments quartiles of hospital SAH disease in high-volume SAH 1,546 hospitals. case volume using one-way treatment centers may improve analisys of variance or Chi- overall survival. square test appropriate. Used GEEs to adjust in-hospital 916 patients mortality. 107 1998- Kaohsiun - The operative volume of 4 surgeons. 2002 Taiwan Multiple Logistic and linear Selected covariates individual surgeons not only A retrospective secondary Surgeon- Each surgeon’s volume was measured regression models were used(Comorbidity with Charlson had a positive impact on clinical data analisys. Clinical and Laparoscopic Clinical speci c by the total number of procedures. to examine the relationship.index) ; Patient demographicsoutcomes, but also greater fi economic information for Cholecystectomy.outcomes : volume. A = 502 ; B = 192 ; C = 147 ; D = 75 used : age, gender, disease effect on conservation of health each patient was abstracted * complications cases. diagnosis, coexixting care resources. Thefindings from medical charts and the rates, conditions, route of admission,validate the theory of Practice financial division. Economic and pre-operative hospital makes perfect. outcomes : days. * length of stay, * hospital charges.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 125 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) d’étude Cohort study of a populationMutlivariable, hierarchical, Thefindings of this study in the Hospital Hospital volume for nonelderdy and based nonconcurrent. mixed effects, logistic context of previous research Trauma acute Two measures of elderly adults was de ned as the suggest that relationships 108 1995- California 39 Nonfederal volume. fi regression analyses. care. hospital quality - US hospitals average annual number of trauma between trauma volume and 1999 care : 84,211 trauma admissions. Three measures of within- outcomes exist but depend on * In-Hospital patients. hospital volume were defined for each which patient populations are mortality hospital as the monthly, quartely and studied and how the data are * Hospital yearly deviations from mean monthly, analyzed. Furthermore, trauma readmission. quartely, and yearly volume. centers may be subject to the detrimental effects of high temporal volume overextending existing services and capacity. Since this study found that both between-hospital volume and within-hospital volume measures are associated with outcomes. Primary * Mortality, 109 1991- California 222,684 total knee * Infection, 2001 - US primary Arthroplasty. This study confirm the total knee The models were adjusted for Multivariate logistic relationship of lower hospital arthroplasty. Data for all hospitalizations patient age, sex, race/ethnicity, regression models were volumes to a higher OR for Hospital Surgical volume was determined as obtained from discharge insurance type, Charlson performed to examine the complications after total knee surgical the average number of primary total California’s Office of comorbidity index, Hospital association between hospital arthroplasty. volume. knee arthroplasties performed yearly Statewide Health Planning type and size and teaching * Thrombophlebitis, volume and complications. during the study period. Hospital dividedand Development (OSHPD). status, year of surgery, and * Pulmonary into 3 tiers as High, Intermediate, Low wether patients underwent embolism. Volume. Hospital size divided into 4 bilateral procedures. categories based on the number of beds (< 100, 100-199, 200-299, and >299).

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 126 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) This study demonstrates that d’étude Independant variables increased hospital experience examined in the analisys is associated with a marked Clinical and Using plubicly available Poisson regression was used 4,561 patients;Complex included patient age, gender, decrease in hospital mortality. 110 1989 - Maryland - Economic Hospital Hospital stratified into 4 volume groupsdischarge : data. to assess the relationship 51 hospitals, Gastrointestinal race, admission status The decreased mortality at 1997 US Primary volume. High-volume provider : 201 or more between in-hospital mortality 6 procedures Surgery. (elective, urgent or emergent ..),the high-volume provider outcomes : procedures/year ; and hospital volume after codes. diagnosis (benign, malignant),was also associated with * In-hospital Medium-volume group : 21 to 50 casemix adjustment. 6 procedures, time period shorter lengths-of-stay and mortality, procedures/year, Multiple linear regression ( scal years), payment source,lower hospital charges. *Average length Low-volume group : 11 to 20 models were used to assessfi and place of residence, These ndings were more of hospital stay, procedures/year ; differences in average fi Charlson comorbidity score. * Average total Minimal-volume group : 10 procedures length-of-stay and average pronounced for malignant or less/year. total hospital charges among diagnoses than for benign hospital volume groups. conditions. Characteristics hospital charges. of the high-volume provider thought to contribute to improved outcomes include overall experience level of Nothe signiphysiciansficant differences and staff ; that Adjustement following risk dospecialized not favor staff the ;hypothesis facilities, and factors : age, sex, previous thatequipment there is in increased the operating surgical Prospective, multicenter The comparison of patient The number Number of cases/year: 3 group: tumor, peripheral vascular riskrooms and and worse intensive survival care in units. Operative Spanish study. characteristics among the Cancer: of hospitals * 1 - 43; disease, weight loss, systemiccenters having a lower volume Las palmas19 hospitals mortality, different groups was carried 111 1993 - Bronchogenic intervention * 44 - 54; arterial hypertension, diabetes,were found in this Spanish de Gran departments, Long term out using on-way analisys of 1997 carcinoma (BC) volume. * ҆55. level of dyspnoea, symptoms,multicenter study. Canaria - 2,994 patients. survival, variance (ANOVA). The survival Thoracotomy for to be bedridder, COPD, Spain Mortality analisys was performed lung cancer. employing the Kaplan-Meier ischemic cardiopathy, and ҅30-days. test. Comparisons of survival type of surgery (exploratory, curves were examined using thoracotomy, and incomplete the log-rank test. Mortality surgery). Hospital Hospital volume was calculated as the was analyzes by means of a Mechanical All cause of binomial logistic regression 13,846 medical volume mean annual number of ventilation 112 1998- Province ventilation. mortality (death) model. patients ; ventilation. episodes performed at each hospital 2000 of Ontario - within 30 days 6,373 surgical during the study period. It as grouped in Canada of initiation of patients ; 5 volume categories : mechanical 126 hospitals. ventilation.

For surgical patients requiring Estimated were adjusted for Population-based Multivariable logistic mechanical ventilation for >2days, patient demographics (age, retrospective cohort study regression analisys was used hospital volume had no effect on sex), mean number of Charlson using administrative to examine the relationship. mortality. For medical patients, comorbidities, diagnoses, and databases. higher mortality may occur in a admissions status, Hospital subgroup of low-volume hospitals region and rural location ; andthat do not routinely transfer (<100; 100-199; 200-299; 300-699; ҆700). accounted for clustering withintheir patients to larger-volume hospitals. facilities. this ndingsfi needs further investigation in a larger- sized study.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 127 -

Pays et/ For Gastrectomies and Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode AdjustingVariables for patientscontrôlées colectomies, Résultats risk -adjusted étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) demographics (age, mortality is substancially The association of in-hospital d’étude gender, race), Insurance lower when performed by mortality and subspecialty Hospital and The combined Hospital/Surgeon volume status (Medicaid), Surgical subspecialty interested and Gastrectomy In-hospital training/interest was examined 113 1998- New-York 55,016 Surgeon measure had the following 4 categories: subspecialty training. trained surgeons, even after and Colectomy.mortality. using a logistic regression 2003 US inpatients Volume. *Low hospital volume/Low surgeon Comorbidities, ischemic heartaccounting for hospital and model. The X² test was 68 surgeons in volume; disease, airway obstruction, surgeon volume and patient used to assess the bivariate Oncology ; * Low hospital volume/High surgeon congestive heart failure, organcharacteristics. Thesefindings association between each risk 61 surgeons in volume; metastasis, peripheral vascularmay have implications for factor and in-hospital mortality. colorectal. * High hospital volume/Low surgeon disease, chronic obstructive surgical training programs and The correlation assessed by volume; pulmonary disease, diabetes,for regionalization of complex Spearman-rank correlation * High Hospital volume/High surgeon and dysrhythmia. surgical procedures. volume. coefficient. A multivariate Hospitals with more Hospital volume Quartile : 1-27 ; 28-54 ; generalized linear model for Adjusting for patients Admistrative discharge data. Logistic regression analysis years of experience with Patients55-140 ; >were 141; divided into binary outcomes was used characteristics (age, gender) Pancreatico- Inpatient Hospital was used to examine pancreaticoduodenectomy 114 1988- California 6,652 patients approximativelySurgeon volume quartiles Quartile :according 1-4 ; 5-9 ; to examine the risk-adjustedand comorbidities index with 500 hospitals. duodenectomy. mortality rates. annual differences in inpatient had lower rates of inpatient 1998 - Florida - to10-20 each ; >21.hospital’s annual procedure relationship. Charlson Index. procedures mortality among hospitals mortality. However, higher US volume: volume. with different procedure procedure volume has played * very low = 1 ; low = 2 or 3 ; medium = volume and years of a larger role than increased 4 -9 ; high = 10. Patients characteristics Bariatric surgery performed at Clinical database. Additionnalexperience. analyses were (age ; gender ; race) adjusted hospitalsexperience with in reducingmore than in 100 Hospital divided into 3 groups : based patient death rates. Annual performed to determine morbidity, Laparoscopic cases,cases annually is associated on the average number of bariatric with a shorter length of stay, US 24,166 Bariatric. In-Hopital hospital differences between high Severity class. 115 1999- operations performed within the 4-year Mortality ; volume. and low volume hospital with lower morbidity and mortality, 2002 patients; period : Perioperative Pearson X² tests. Bivariate and decreased costs. This 93 academic High volume : > 100 cases/year ; outcomes and analyses with one-way volume-outcome relationship centers. Medium-volume : 50-100 cases/year, cost : Length analisys of variance were is even more pronounced for Low volume : < 50 cases/year. of Hospital performed to determine a subset of patients older than stay ; Overall differences in length of stay 55 years, for whom in-hospital complications ; and costs between the 3 mortality was 3-fold higher 30-day groups. at low- volume compared readmission ; with high-volume hospitals. Mean cost ; High-volume hospitals also Disposition. have a lower rate of overall postoperative and medical care complications, which may be related in part to formalization of the structures and process of care.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 128 -

Pays et/ Individual surgeon experience Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir Patient age, race, gender, N° ou région Méthode Variables contrôléesis signi cantly Résultats associated with étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) comorbidity score, thyroid fi The distribution of patient complication rates and length of d’étude Using administrative data. diagnosis and procedure, Surgeon Surgeon volume groups per year : characteristics amoung stay for thyroidectomy. Length A cross sectional analisys insurance status, time period, Thyroidectomy. Short-term volume. * A : 1-9 cases provider groups were of stay and complications 116 1991 - Maryland - 5,860 patients ; was conducted using a place of residence. clinical and * B : 10-29 cases ; were more determined by US 52 nonfederal compared using analysis of 1996 computerized statewide surgeon experience than acute care Economic * C : 30-100 cases variance for the continous hospital discharge. hospital volume, which had hosptials ; outcomes : * D : > 100 cases . variables and the X² statistic no consistent association with 658 surgeons. * In-hospital for categorical variables. complications, Bivariate analyses were outcomes. * length of stay, used to determine which * Total hospital variables were associated charges. with our outcomes. Multiple Linear regression was used to assess how length of stay and mean total hospital charges differed between Tumor size and nodal status. From these population data surgeon volume groups after Surgeon A population-based study ofDescriptive statistics we conclude that few patients 577 Diagnosis ; Cancer : GastricOverall survival adjusting. 117 1991 - Northern Volume. Tumor -node relationship. were calculated, variable presents with «curable» gastric 1997 Alberta Cancer. rates (operative distributions tabulated and cancer, Node negative or small - Canada. 30 days mortality survival curves estimated gastric cancer survival is not and hospital using the Kaplan-Meier influenced by nodal stage, mortality). technique. The survival Positive margin resection distributions between survival is better than bypass or the prognostic groups no surgery in stage IV but not were compared and stage III disease,and surgeon the significance levels volume does not appear to determined by use of the influence patient survival. logrank test. 118 1996 & US 13,887 patients 1997 507 hospital in 1996 ; 536 hospitals in 1997.

This study from a Unadjusted and Case-mix representative national NIS Administrative Database. Univariate comparison usingrisk-adjusted analyses were database documents that Hospital For AAAs, high volume was found to be Cardio : After In-hospital death Chi-square test, Wilcoxon performed. Risk adjustment HVHs have a significantly volume. greater than 30 procedures per hospital Abdominal Aorticrate and length rank-sum test, Student test,included demographics lower death rate than LVHs each year. Hospital assigned either high t Aneurysmectomy.of stay. simple logistic regression, (age, gender, race) comorbid for repair of both and ruptured or low volume status for each year, and and simple linear regression.disease, nature of admission AAA. These data support the volume was encoded as a dichotomous Mutliple logistic regression of(elective, urgent, or emergent),regionalization of patients to variable. the hospital death rate was and ruptured versus intact HVHs for AAA repair. used to test its association AAAs. with hospital volume after adjusting patient case mix variables.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 129 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) d’étude Persistence of the volume- 34,4617 Administrative discharge Adjusted for patient and Florida; None, continious volume. For each Looks if the link between outcome relation and 119 3 time patients Six Cancer : colon,Operative In- Annual data. Logistic regression. hospital characteristics: New-Jersey hospital, the number of times each volume and mortality has increasing hospital and Period 6 cancer rectum, pumonaryhospital mortality.hospital and age, gender, race, nature of New-York speci c operation was performed wasremained stable over time. surgeon volumes explain 1988- operations. lobectomy, Surgeon fi admission, health insurance, calculated for each year. much of the decline over time 1991 ; pneumonectomy, volume. state of residence, cancer esophagectomy, in inpatient mortality for five 1992- stage, and Charlson Wipple procedure. of the six cancer operations 1996 ; comorbidity index. Patientsstudied. Concentratingwho undergo elective cancer 1997- Adjusted patients AAAresections repair amongthat is performedhigh-volume by 2000 Retrospective cohort study Univariate analyses of characteristics (Age, gender, vascularproviders or should cardiac lead surgeons to further Charlson comorbidity index) ; The Annual surgeon AAA volume with linked administrative patients and surgeon havereduced signi inpatientficantly lowermortaltiy. Annual Cardio : Elective30-days risk divided into 3 terciles : databases. characteristics were type of hospital ; teaching mortality rates than patients surgeon 120 1992- Ontario- 5,878 cases ; abdominal aorticadjusted * < 5 Cases/year ; conducted initially. A status ; transfer status ; Medianwho have their aneurysms AAA volume 1996 Canada 63 vascular aneurysm. mortality rate. * 5-13 cases/year ; multivariate logistic annual volume of index repaired by general surgeons. and Hospital surgeons; * > 13 cases/year. regression model was procedures. These results provide evidence annual 53 general Annual Hospital volume of elective AA created to determine the that surgical specialty training volume. surgeons; surgeries performed into : independent impact of each in vascular procedures leads to 14 cardiac Large, > 15 cases/year ; of these factors on 30-days better patients outcomes. Individual surgeons performed 1 to surgeons. Surgeon Medium, 7-15 cases/year ; mortality rates after AAA Cancer : 30-days 163 operations over 4-years (median, California- 28,644 and Hospital Small, 1-6 case/year. surgery. 121 1996 to Colorectal Postoperative interquartile range, 1-13). The number US patients ; volume. 1999 Cancer. mortality and of operations at individual hospitals 397 hospitals ; 6-year overall ranged from 1 to 425 over 4 years 2,993 survival (overall (median, 41 ; interquartiles range 11- surgeons. mortality). 112). Entire Cohort :

Greater surgeon and Hospital Adjusting patients volumes were associated characteristics (age, gender, with improved outcomes for Cohort study from the X² test was used for race/ethnicity ; median patients undergoing surgery for California Cancer Registry. categorical variables and household income, and colorectal cancer. ANOVAs were used for Charlson-Deyo comorbidity age. Survival by quartiles index. Tumor site/stage. of hospital and surgeon volume was estimated with the Kaplan-Meier method and compared with the Surgeon volume : 1-12 ; 13-24; 25-40; log-rank test. Multivariable >40; analyses. Cox proportional Hospital volume : hazards models analyze ҅83 ; 84-151 ; 152-219 ; >219. overall survival. Used Logistic regression, adjusting standard errors to control for clustering patient by hospitals.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 130 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir Patients at the High-volume N° ou région Méthode Variables contrôléesOPCAB facilities Résultats had étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Age, sex, smoker, history d’étude significantly lower rates of The p-values were calculatedof tobacco use, chronic major complications than those Adminsitrative database Volume of Cutt-off of 100 OPCAB operations: for each population patient obstructive pulmonary at the low-volume OPCAB Clinical containing patient, clinical, 122 1999- US 16,988 Cardio : Off- OPCAB. divided into High and Low volume disease, insulin-dependent outcomes : and hospital using Student’s t facilities. High volume OPCAB 1999 patients ; Pump Coronary off-pump facilities. (> or < 100 case perprocedural, and outcome and noninsulin-dependent Mortality, test. Used Logistic regression sites were significantly more 72 hospitals. artery Bypass. year). data. diabetes melitus, acute 11 Complications techniques and reported the likely to discharge their patients Case-mix data. renal failure, chronic renal risk. odds ratio and the p values directly home than were low- for the High volume OPCABfailure, unspecified renal volume OPCAB sites. coefficient. 12 logistic failure, cardiogenic chock, The results suggested that regressions with dependenthypertension, AMI, old MI, surgical team experience variables being patient cardiomyopathy, congestive and choice of approaches to mortality and each of the 11heart failure, peripheral performing CABG had impact complications. vascular disease, endocarditis.,on patient outcomes. procedural characteristics, The surgical volume of the Adjusting for age at diagnosis Hospital characteristics (LOS,hospitals did not affect the 10- and clue for detection, Cancer Data from the OSAKA cancerThe cumulative observed hospital surgery volume) ; year survival rate significantly, Cancer : Breast 10-years survivalHospital Each Hospital divided into four stage (localized, regional, 123 1985- Osaka- 4,333 female registry (OCR). survival rate was estimated discharge disposition. except for the very low volume Cancer. rates. volume. categories : High, medium, low, very distant, unknown), Range of Japan patients 30-64 using the Kaplan-Meier hospitals in Osaka, Japan. 1991 low. surgeries per hospital/per year. years old ; method relative to the However, the study of these screnning detected. 200 hospitals. Hospital surgical volume. The relationships should be relative 5-year and 10-year continued and expansed in survival rates were calculated future to include . by the ratio of observed to expected survival. The Elderer II method was used. Survival differences were analyzed by Cox’s proportional hazards models adjusting.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 131 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Hospitals that perform a high d’étude Adjustment for volume, age, volume of esophagectomies Using two samplet-test for Data obtained from the race, comorbidity score, have better results with Hospital Hospitals were stratified to low volume the continuous variables and Massachusetts Health Data urgency of admission, sourceearly of clinical outcomes and 124 1992- Massachussets1,193 patients ;Oesophagectomy. Primary Volume. hospitals (<6 cases per year) Consortium on discharge x-Square tests for categoricaladmission, year of admission, 2000 64 hospitals. outcomes : and High volume hospitals (>6 cases marked reductions in mortality - US information for all acute carevariables. Multiple linear andpayer type, and place of mean and per year). compared with low volume Hospitals. logistic regression models residence were performed for median length hospitals. were used to assess the thefinal regression models. of stay, length independent effect of hospital of intensive volumes. Patients undergoing care unit (ICU), Adjusted : age, gender, race, esophageal resection at In-hospital Risk-adjusted analyses were Hospital Hospital volume was converted into nature of admission, operatingLVHs were at a markedly Esophageal *mortality In-hospital ; and performed using multiple Maryland - 366 patients ; volume. dichotomous variables using a threshold physician, vital status at increased risk of postoperative 125 1994- resection. discharge mortality ; logistic regression. Test of US 10 for » low » and « high » volume. discharge, comorbid disease complicaitons and death. 1998 *destination. 10 univariate association were complications, adjusted for using Romano’s Pulmonary complications are postoperative performed using the X² test, 52 nonfederal modification of the Charlson particularly prevalent at LVHs complications. simple logistic regression, acute care comorbidity index. and contribute to the death t-test, Wilcoxon rank-sum, hospitals. Comparisonof patients having of the surgery AMI at Japanese, European, Chi-squareand simple linearanalysis regression. used to compare dependence of databasesthose centers. suggest there is In-Hospital ; PCI volume The PCI volume was divided into American AMI databases Cardio: primary categorical variables. a relationship between the 3 japaneses hospitals. 3 categories : and world wide databases of 126 1998- Japan, percutaneous 30-day mortality. primary PCI rate and annual patients ; * small low -volume hospital acute coronary syndrome. 2000 Europe, Coronary PCI caseload in each country. 4 European, (< 200 cases/year), America, intervention. It is interesting that in Japan 4 American, * Intermediate volume hospital World- even low-volume PCI hospitals 2 World-Wide. (200-400 case/year), Wide. have comparable numbers of * Large high volume hospitals primary PCI cases. (> 400 cases/year) ; In the JSIC hospitals divided into : * < 17 cases/year : small ; * 17-48 cases/year : intermediate ; * > 48 case/year : High.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 132 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir There was no signi cant N° ou région Méthode Variables contrôlées Résultatsfi étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Adjustment : patients relationship between hospital d’étude volume and in-hospital Data from the Japanese Differences in demographic,characteristics (age, gender, Hospital Hospitals divided into terciles according outcome among AMI patients Coronary intervention Studymedical, angiographic, and previous myocardial infarction, In-hospital volume. to the annual PCI volume : undergoings PCIs. 127 1997 Japan 129 hospitals,Cardio : (JCIS). Using the contemporyprocedural varaibles were Hypercholesterolemia, mortality, * low = 10 {1-16}, medium = 33{17-55}, 2,491 patients.Angioplasty for statistically assessed by smoking, hypertension, and high = 89{56-370}. database. acute MyocardialIn-hospital Chi-square test and Student’sdiabetes mellitus, renal failure, Infarction. bypass surgery t-test. The relationship cerebrovascular disease, prior following PCI examined in terms of odds PCI, prior CABG, number of (CABG), ratio using multiple Logistic diseased vessels, attempted the combines regression analysis. lesion, LVEF, types of devices, endpoint of In- and backup cardiac surgery. hospital mortality Adjusted : age ; sex ; cardiac Prospective cohort study. Chi-square test for trendfunctionbaseline and (acutedemographics, myocardial or CABG. Hospital and surgeon volume divided Hospital and infarctionmedical, angiographic. (AMI), angina, The rates of : into quartiles : trend analysis were used to 128 2000 - Taiwan 4,724 patients;Cardio: After Surgeon congestive heart failure (CHF), * In-hospital (IH), * The highest quartiles : 4 ; test for bivariate relationship 2001 34 hospitals, Coronary Artery Volume. ventricular brillation and * 30-day after * The lowest quartiles : 1&2; between volumes levels fi 203 surgeons.Bypass Grafting. discharge * Median quartile : 3. and patients characteristics,fl utter), co-morbidity (peripheral (AD30), In-hospital characteristics vascular disease (PVD), and mortality rate. After cerebrovascular disease, renal * In-Hospital The relationship of higher- adjustment by stepwise failure, and shock) and in- plus 30-day volume hospitals or surgeons logistic regression was usedhospital complications. after discharge. with lower mortality rate for to analyze the association of patients undergoing CABG is a Model 1 & 2 &3. general phenomenon in Western and Chinese societies. However, the combination of the AD30 and IH mortality rates has to be considered when investigating procedural mortality rates in Chinese society. Age, gender, race/ethnicity, No other outcomes were signi cantly associated with Prognostic study, Level II-1 Using Fisher’s exact test, and comorbidities by Charlson fi hospital volume. Hospital Hospital volume per quarter year retrospective study. Kruskal-Wallis test data Index. * Death within divided in tertiles included : Thefindings that greater 129 1996- New-York 1,307 patients. Cardio : Total volume. with a post hoc Mann- 60-days, *Low volume : centers in which fewer hospital volume decreases 1999 US Shoulder Withney U test for pair-wise risk of readmission may Arthroplasty * Readmission than one total Shoulder arthroplasty comparison. Using Cochran have important public health (TSA). within 60-days, 1 - 15 cases. Mantel-Haenszel Non zero * Length of Stay * Middle volume hospitals : 16-47 cases ; Correlation Coefficient to implications, but additional * Revision * High volume hospitals : 48 or more test liear trend. Multivariate research is needed before surgery within cases. analysis. Collinearity implementing policy changes. 24 months, measured and used linear * Hospital cost. regression plots.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 133 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôléesElective surgery Résultats of the étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) After adjustment for baseline The propensity scores were ascending aorta/arch was d’étude Surgeon High and Low volume Surgeon. Demographics data and differences, era surgical risk/ estimated using multiple associated with low mortality. Volume. perioperative data for the complexity procedure. Cardio : ThoracicOutcomes after logistic regression. Baseline Outcomes after emergency 130 1992- UK 296 procedures period were collected by aorta/aortic. Surgery : characteristics were surgery conformed to 2003 carried out on retrospective case note * Intraaortic compared using the contemporary. Only limited 291 patients ; review. Cohort study. balloon pump ; chi-squared of Fisher’s exact differences were identified 296 surgeons ; * Re-operation test or Wilcoxon rank sum both with respect to the case 37 hospitals. for bleeding; test. Using Cox proportional profile and early clinical Rewiting ; hazards regression. outcomes. Better outcomes * Neurological in the mid-term in the higher complication, volume group persisted despite *Renal adjustment for differences in complication ; caseload and are worthy of * In-hospital further study. death ; *In-hospital Cardio : * Post-operative 131 1999 - US 12,293 PCIs death rate ; Percutaneous hospital stay; 2001 performed ; * In-hospital Coronary * Patient survival 28 surgeons. composite In contemporary PCI practice Intervention of death, Patients characteristics at a large center with high- (PCI) volume operators, in-hospital MI, CABG, or Prospective database on allComparisons between (age, sex, diabetes metillus, Angioplasty. outcomes are not affected by Surgeon Operators were classified as by tercilespatient undergoing PCI at thisgroups were made with Hypertension, heart failure, operator volume, experience, volume and of annual PCI : institution. the Kruskal-Wallis test or MI, Prior Coronary bypass, or board certi cation. Rather, Experience. * Low ҅92, Wilcoxon rank sum test Prior PCI, smokers, Chronic fi * Medium : 93 to 140, for continuous variables, obstructive pulmonary disease,patients’ clinical risk score is * High : > 140. and the Chi-square or Prior stroke, Serum creatine, the overriding determinant of ExperienceTerciles of years in practice : Fisher’s test. To determine Multivessel Coronary disease)clinical outcomes. Thisfindings ҅8, 9 to 14, or >14 years. the independant effect of and Operator characteristics emphasize the power of a well- Cerebrovascular operator characteristics and(years of practice and volumeorganized high-volume system Outcomes of patients who of ҆2of 3 BRS, performed multiple Patientsof PCI, and demographics Board certi ficationto minimize the impact of Using stepwise multiple underwent primary angioplasty criteria. logistic regression. andstatus). medical and cardiac operator factors on outcomes Hospital Procedure volume : logistic regression models for AMI were improved when Total annual characteristics. Adjusted for of PCI. In-Hospital * High volume҆ ( 400 PTCA to control for potential these procedures were Cardio: primary Coronary age, medical and cardiac 132 1995 New-York 151 physicians; mortality procedures in 1995) ; confounding factors. Using performed in hospitals that Coronary Agioplasty history time to treatment, US 32 hospitals ; following * Low-volume (< 400 PTCA chi-square analysis and performed >400 PTCA cases Angioplasty with Volume of ejection fraction and stent use. 1,342 patients. Outcomes procedures/1995) ; Student’s t-test. annually. The total annual acute myocardial Physicians of Primary Physician procedure volume : PTCA volume category for infarction. and Hospitals. Angioplasty for * High volume҆ ( 75 PTCA Physicians does not appear acute Myocardial procedures) ; to predict outcome of primary infarction. * Low-volume (< 75 PTCA procedures). angioplasty for AMI.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 134 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) There was wider variation in d’étude Data of two voluntary Chi-square and t-test ; Hospital grouped into quartiles using outcome among lower-volume Hospital databases using the STS using of scatter plots with Cardio : CardiacIn-hospital death, cut-point : Low (200) and high҆500) ( California 119 nonfederal annual logarithmic trendlines. hospitals. However, many 133 1997- (CABG). Discharge ; natioanl database. hospitals ; procedural annual volume of procedures per year low-volume providers had 1999 - US transfer to 120 cardiac volume. CABG only ; excellent results. Concluded another health programs ; CABG plus & Valve only : annual that although volume is care facility. + 200 CABG. volume :҆ 100 high ; <25. clearly related to outcome patient-related factors and process variables may be more important. Performance Concludeimprovement that programs among hospitals should Assessment of risk after participatingbe developed in to this improve trial, there Logistic regression used enrollment, sex age, current Study assessed both iscommunication nonlinear relation between between Quartiles groups of annual to examine the effect of smoking status, reason for Annual procedural volume and annualproviders interventionnal with differing volume interventional volume : academic status and annualrevascularization ; types of institutional academic status. andoutcomes. outcomes. This relation is * 90 - 554 ; interventional volume on unstable angina present ; 4,010 patients ;Cardio :CoronaryThe primary end-Volume of complex, involving variations 134 1993 - San * 555 - 1,200 ; the primary composite end treatment groups. 82 centers. Intervention. point : coronary in periprocedural infarction 1994 Francisco, * 1,201-1,492 point. Stepwise and Barward * composite of interventional rates and additional, unde ned California * 1,493 -3,300. regression methods were fi death ; procedures. institutional differences -US used to develop the final Myocardial (other than academic status) model of predictors. infarction, that result in differences in Nonelective procedural outcome. repeat coronary intervention, or nonelective CABG at 30 days, or Placement of Adminstrative database. Mutliple logistic regression Surgeon volume and an intracoronaryHospital or General vascular surgery, hospital size, In-hospital death certification are significantly Florida -US 31172 LEABs,Cardio : stent for abrupt Surgeon patient age and gender, type of surgery was used to test the 135 1992 ; rates Outcome related to better patient 45,744 CEAs, Vascular : CEA, closure during volume. with or without certification. significance on outcome of 1993 ; after CEA, outcomes for patients who 13,415 AAs. LEAB, AAA. the initial surgeon volume. 1994 ; undergo CEA and AAA. LEAB,procedure. AAA ; 1995 ; length of stay. In addition, surgeons with 1996 high volumes demonstrated consistently lower mortality and morbidity rates than the surgeons with low volumes. Hospital volume for a given procedure also is correlated with better outcomes.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 135 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode PatientVariables age, gender, contrôlées race, The incidence Résultats or rAAA does étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) not appear to be declining. Categorical variables were associated comorbidity rates, d’étude Hospital and surgeons were stratifiedIn the community-based, Although operative rAAA repair analyzed by Pearson’s operative surgeon experience Annual regarding AAA and rAAA into Low, cross sectionnal study. continues to be associated Outcomes after X² test or Fisher’s Exact with rAAA. 136 1990 & Maryland 3,820 patients ;Cardio : Repair Hospital Medium, and High annual volume. with substantial risk and rAAA Test. Continuous variables 1995 - US 527 recent of abdominal volume ; remains an especially lethal * Operative were analized by analysis rAAA cases. aortic aneurysm. and condition among the elderly, mortaltiy rates ; of variance for parametric total AAA the operative mortality rate * hospital length variables and the Mann- operative has declined in recent years of stay ; Whithney U test or Kruskal- surgeons in Maryland. Lower operative * Total Hospital volumes. Wallis test for nonparametric Floridamortality CON rates (Certi and fihospitalcate of charges. variables. Multiple logisitic Age, sex, acute myocardial need)charges laws are were associated associated with regression was used for Florida AHCA and CaliforniaLogistic regression pooling infarction (AMI), multivesel withoperations higher performedaverage PTCA by high- Annual PTCA The average hospital volume, the multivariate analysis. In-hospital OSHPD dicharge databases.the data from california andPTCA, and Stent placement, volume surgeons. Cardio : percent of hospital performing < 200 volumes relative to California 292,832 CABG surgery orprocedure Florida. Two time period Charlson comorbidity index, 137 1988- Florida and Percutaneous procedures, and the percent performing hospitals, where no such laws admissions in death. volume for dummy variables were Diabetes, Renal insufficiency, 1998 California - transluminal ҆400 procedures in each state and exist. Because a higher PTCA orida ; each hospital. included in the regression Peripheral vascular disease, US fl coronary year. volume was associated with 393,754 to compare outcomes for length of stay. angioplasty moderately better outcomes, admissions in admisssion that occurred (PTCA). CON may be marginally California. during 1991 to 1994 and Higheffective risk inpatients improving are outcomesmore Clinical database. Using1995 to hierarchical 1998 versus logistic 1988 Adjusting for expected surgicallikelyfor PTCA. to undergo CABG at Low- Hospital Hospitals were divided into terciles of Cardio : In-hospital regressionto 1990. and log-binomialrisk. Age, sex, race, surgical volume facilities where their 138 1997- California 27,355 patients annual CABG volume : mortality rates Volume. regression models. priority (elective, emergent, risk of dying is higher. - US adults Coronary Artery 1998 after CABG. * Low-volume n = 44; or salvage) previous heart 68 hospitals. Bypass Grafting. * Medium-volume n=19 ; surgery, left ventricular ejection * High-volume n = 5 fraction, extent of left main stratified across 4 risk percentiles : coronary artery disease, extent 20th ; 40 th; 60 th; 80 th. of overall coronary artery Several Physician and disease, history of angina and Patients characterisitics : Hospital characteristics are Analysis of a large severity... Hospital case Hospital CEA volume in the 4 fiscal Canadian administrative age, sex, admisssion throughdeterminants of outcome after Cardio: carotid Outcome after administrative database 139 1994 - 9 Provinces14,268 volume. years into 2 categories: hospital discharge database. the emergency department, CEA, but the negative effects of endarterectomyCEA : In-hospital using logistic regression to 1997 of Quebec patients . * Low (< 150) and high҆ (150). comorbidities, and Hospital low hospital and surgeon cases (CEA). stroke and/or correlate adverse outcome Canada and Physician related variablesvolumes, in particular, suggest death. using logistic regression to (surgical specialty), hospital that regionalization should be correlate adverse outcomes teaching status, clinical trial considered for CEA and that after CEA with surgeon and participation. surgeons with low case volume hospital characterisics. should not be performing CEA.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 136 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) d’étude Outcomes for Hospital and The risk adjusted end point Cardio: 140 1996- Florida and452,404 PTCA. Operator Divided hospital/operator pairs into 4 Retrospective cohort study Weof tested for trends over difference between classes Outcomes were best for patients. percutaneous time using the Cochrane- 2001 New-York- volume. classes based on volume classificationsdischarge records of patients narrowed over time. Patients patients receiving care from transluminal US in the the ACC/AHA. Annual operatorwho underwent PTCA at Amritage test for trend. characteristics (age, gender, class1 hospital/operator pairs, coronary and hospital volume : nonfederal hospitals. Used Logistic regression state, race/ethnicity, and an increasing proportion angioplasty Class 1: min annual operator volume of to calculate mean annual co-morbidities, admission of patients received care from (PTCA). 75 and min annual hospital volume > 400. increases in procedure source, type of admission. class 1 pairs. There were Class 2 pairs: min operator volume of 75 volume. outcomes differences within and a hospital volume of 200 to 400, or subpopulations of operators in an operator operator < 75 but a hospital classes 2 and 4, which suggest volume > 400. possibilities for alternative Class 3: < 75 and a hospital volume from volume classification guidelines. 200 to 400.

Hospital was categorized into quintiles Hospital Cardio : In-hospital or CABG volume : Patients demographics ; Patient 141 1996 - US 948,093 volume ; A retrospective cohort studyLogistic regression was Volume alone, as a Coronary ArteryMortality within very low : fewer than 125/year ; comorbidiites, admission type, 2001 Medicare Leapfrog database derived from used to adjust hospital discriminator of mortality, is only Bypass Graft 30 days after low : 125 to 204/year ; surgical priority. patients ; volume Medicare Provider and Analysismortality rates for patient slighty better than a coin flip. Surgery. CABG. 870 US category. Medium : 205 to 299/year ; Review (MEDPAR) part. A characteristics ; discrimination hospitals. High : 300 to 499/year ; public use data files. Clinicalof the volume categories was Very High (more than 449/year). and Administrative database.assessed by the c statistic. Hospital were classified by volume criterion proposed by the Leapfrog Group.

Did not nd an association Demographics/noncardiac; fi Hierarchical logistic between OPCABG volume * 30- day Based onVA continuous Cardiac, surgical priority, 142 1998- US 5,076 OPCABGCardio: Off- OPCABG Procedural volume defined in two ways : regression models evaluated with short-term mortality, operative Improvment in Cardiac current use of diuretics, 2003 surgical Pump Coronary procedural * Average hospital OPCABG procedural the relationship between perioperative morbidity, or mortality ; Surgery Program data. of digoxin, preoperative procedures; artery Bypass volume. volume in 6-month period ; OPCABG procedure volume intermediate-term (180 day) * Perioperative intravenous nitroglycerin. Graft. the most recent 6 - month hospital with risk-adjusted. 44 centers. morbidity ; mortality. OPCABG volume preceding the * 180-day all surgical procedural. Hospital-based factors, such 143 1991- * <10,2 ; 10,2 - 16,6 -30,2 ; >30,2. as greater bed capacity, use Race, gender ; Comorbidties 2002 cause mortality. of critical pathways, ability to Multivariable logistic Hospital factors were Population of the hospital perform cardiac angiography, Cardio : Carotid Hospital were strati ed as low, medium, regression was used to Connecticut- 14, 288 CEA Outcome afer CEAHospial fi established by telephone location, setting, residents, or presence of a dedicated Endarterectomy or high volume institution, de ned as analyze data. US 26 hospitals. Perioperative Volume. fi survey and validated by hospital capacity bed ; vascular recovery unit CEA. performing҅10 ; 10 to 49, or҆ 50 Categorical variables were mortality repeated survey. academic status, cardiac predict reduced preoperative analyzed by Pearson’s * in-hospital death procedures annually. angioplasty capability, cardiacmortality, stroke and cardiac Hospitals were also strati ed as low, X² or Fischer exact test. surgery capability, presence of *perioperative fi complications from CEA. Continuous variables medium or high size institutions : surgical residents, presence ofThese results suggest that stroke, analyzed by the Mann- ҅131 , 132 to 279, or ҆280 staffed surgical ICU, number of ICU hospital-associated factors do * perioperative Whitney or Kruskal- patients beds. t-test beds, size of intensive care impact surgical outcome and cardiac Wallistest. unit ; presence of an intensivist, complications, that surgeons need to optimize presence of recovery beds; these factors, extrinsic to the * length of dedicated to vascular patients ; stay(LOS) patient and surgeon, to provide Postop SICU admission, maximal quality of care. * intensive care Hospital CEA volume, Critical unit (ICU) LOS, pathway, Routine D/C. *Unadjusted

hospital charges.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 137 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) d’étude Low and High volume Hospital Mortality For the vast majority of Procedure Demographics patients (sex, Cardio : In-Hospital rate ; Logisitic regression modeling patients, low risk patients 144 1996 New-York 20,078- volume after Retrospective cross - age,) and risk factors patients Coronary Arterymortality after Risk group (very low, low, moderate, was used to explore the benefits significantly more US patients ; ajusting for sectional: Cohort analysis (hemodynamic state, ejection Bypass Surgery.CABG. high, very high) interaction between patient than high-risk patients from 32 hospitals. severity of conducted on clinical fraction, previous MI, previous 75% patients underwent surgery at databases. risk and procedure volume. undergoing CABG surgery at disease. open heart surgery, chronic high-volume centers instead of obstructive pulmonary disease,at low-volume centers. Low-risk Diabets, Hepatic failure, Stroke. high-volume centers (> 500 cases/year) patients benefits significantly and low volume҅ ( 500 cases/year). more than high-risk patients from undergoing CABG surgery at high-volume centers instead of at low volume centers. However, before generalizing thesefindings to other states, this study should be repeated using other regional population based clinical databases.

145 1992- In this analysis, only volume 1996 Level of accreditation, A retrospective analysis of A logistic regression model of patients treated had a direct Patients volume Per year - Injuries presence of in-house trauma Trauma care. Survival Volume prospectively collected data.was the created to evaluate impact on survival outcome. Pennsylvania-24 accredited volume trauma centers : surgeons, presence of a outcomes Risk of trauma the most seriously injured Accreditation, regardless of US centers ; Low volume : < 1,200 trauma surgical residency program, of mortality. admission patients. level, appears to be beneficial. 88,723 centers. admissions/year; and presence of an-site patients ; High volume ҆: 1,200 trauma medical school. Injury type. 9 serious admission/year. injuries.

146 2000 5 States Patient-, surgeon-, hospital- Higher surgeon and hospital (Florida, Data for a nationally Multivariate analyses were level variables. volumes are associated Maine, Surgeon and Hospital volume were performed to analyse the 1,277 infants. Pyloromyotomy. * Major operativeSurgeon representative sample of with better outcome among New-York, & Hospital stratified into quintiles. impact, and using models complications ; in infants from the Kid’s infants who are treated for Volumes. Surgeon Volume in procedure/year : that accounted for the * Length of stay ; Inpatients Database. pyloric stenosis. Identification 1-3 ; 4-6 ; 7-10 ; 11-14 ; >14. hierachical structure of * Hospital of aspects of medical and Hospital volume in procedure/year : patient, surgeon and hospital surgical treatment that charges. 1-5 ; 6-13 ; 14-27 ; 28-34 ; >34. Pennsylvania, level covariates. account for this finding may South lead to improvement in the Carolina) outcome of infants undergoing US pyloromyotomy.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 138 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Age, hospital stay ; type of This study supports an inverse d’étude operative procedure. relationship between hospital Differences in hospital The patients operated were grouped volume and hospital mortality Hospital Hospital mortality for the groups with 147 1987- Sweden 1,429 patientsCarcinoma ; of into hospitals assording to yearly after surgical tumour resection mortality, operation different hospital operative 1996 74 hospitals. the Oesophagus hospital operative volume into three of the oesophagus mortality 5-year survival volume. volume were tested by and Gastric groups : after surgical tumor resection rates. Fisher’s exact test. The effect Cardia. Low volume < 5; of the oesophagus or gastric of hospital volume analysed cardia. Overall 5-year survival Intermediate-volume : 5 - 15 ; by Kaplan-Meier survival High-volume : > 15 resections. was significantly higher at high- curves. Using Cox model - volume hospitals compared Relative survival model to low-volume centres. We can be fifted by Poisson believe that national authorities regression, evaluation must move to centralize the of covariates and model surgical treatment to tumours goodness-of-fit ; using the of the oesophagus and gastric methods for generalized cardia to high volume-hospitals. linear models. This is a prerequisite to achieving high quality surgical care and to facilitate research aimed at improving prognosis for these patients. 148 1992- 1999 During the 5-year period, Patients characteristics: (age,40% of the procedures gender ; weight loss) Jaundice, Groups were compared using were performed in hospitals Part A: (single institution Cardiopulmonary comorbidity, Hospital Hospital divided into four categorie the Chi-square statistic, the performing fewer than five database divided into 2 Preoperative biliary drainage, volume. based on the number of resections two tailed Fischer exact test, resections per year, and Netherlands300 patients. Pancreatico- Part A : periods) & B: (Netherlands Perioperative laboratory values. performed per year : fewer than 5 ; 5 to and the Mann-Withney test. death rate was greater than duodenectomy. *Rates of overall medical registry data on age Surgical details ; Risk factors of 9 ; 10 to 24 ; 25 or more. Correlation between factors in hospitals performing more complications ; and postoperative death of complications. was calculated with Pearson than 25 resections per year. * Death after patients. procedure ; correlation test. Using Chi- The overall death rate after Hospital Length square analyze the impact pancreaticoduodenectomy old stay ; of patient characteristics did not decrease significantly * risk factors. and surgical factors. Using a during the period, and it was multiple Logistic regression greater in low-volume hospitals model to assess their and older patients. The lower Part B : independant prognostic value death and complication rates * Postoperative for complications. in high-volume hospitals, including the single center death. outcomes, were similar to those reported in other countries and may be due to better prevention and management of complications. Pancreaticoduodenectomy should be performed in centers with sufficient experience and resources for support.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 139 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) d’étude Age, gender, height, weight, A large caseload in rectal Using X² test to evaluate tumour stage, risk factor surgery results in a significant Prospective multicentre Hospital divided into three groups by the number of cases, and and american society of reduction in permanent stoma 149 1999 Germany 75 hospitals Rectal; Cancer. Short-term Hospital observational study. two dimensional frequency formation and postoperative annual caseload : Anesthesiologist classification. 1,463 patients. postoperative annual Hospital participated on * group 1: less than 20 ; distributions were compared morbidity. outcome. caseload a purely voluntary basis. group 2: 20-40 ; using X² test. Additional volume. The questionnaires were group 3: more than 40. Tukey multiple comparisons completed by the senior test. registrer of each hospital/ department. An observational 150 1993- study. 1998 There is a signi cant (inverse) Compare case mix with fi For categorical variables, relation between hospital Netherlands310 esophagec-Esophagectomy. Hospital Hospital Centers divided into 3 categories: DNMR database anonymous. demographic data (age and using the Chi-Square statistic mortality and hospital volume tomies mortality. volume. * (43-55)Low volume centers The result of the current gender), indication for surgery for linear trend. The Mann- for esophageal resection in performed. (1-10 esophagectomies a year), study based on a (malignant, non malignant), and Withney U test used with the the Netherlands. Although * (1-3)Medium volume centers retrospective analysis of pathology results. pTNM stages as categories. hospital mortality is not the only (11-20 esophagectomies a year), data from two nationwide measure for quality of care, * 2 High volume centers databases. (> 20 esophagectomies a year). these data suggest a potential benefical effect to centralization of esophagectomy in the Low operative mortality Age group, sex, year of Netherlands. Hospital Hospital classified into quartiles groups.Retrospective cohort study. Multiple Regression analyses and length of stay were Hepatic Operative surgery, source of admission, 151 1990- California - 138 hospitals ; volume. number of operations/5 years : were used to adjust for associated with high volume Resection for mortality ; type of resection, presence 1994 US 507 patients. 1-2 ; 3-6; 7-16; ҆17. differences in patient mix. centers. These data support Hepatocellular Length of stay. of chronic liver disease and regionalization of high-risk Carcinoma. presence of other preoperative procedures in general surgery, comorbid illness. such a hepatectomy for HCC.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 140 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Speciality training in thoracic d’étude Adjusted for patient Used multiple logisitic surgery has an independant The number of cases performed demographics (age, sex, In-hospital or Hospital and regression to compare the association with lower mortality US 1 946 patients,Esophageal race), coexisting diseases, and 152 1998- Surgeon by each hospital and each surgeon risk of operative mortality after esophageal resection. But Cancer 30-day mortality admission acuity to account for 1999 1 118 surgeons Volume. during the 2-year study period. For with thoracic and general specialty training appears to Resection. before discharge. case-mix differences among (Thoracic & both hospital and surgeon volume, we surgeons, adjusting for be less important than hospital surgeons. Comorbidity with general). created three approximatively equal providing volume. and surgeon volume. size groups on the basis of terciles of Charlson Index, income group. 153 1994- volume: low, medium, high volume. 2004

Medical database. Used X² test an I² test. Adjustment patient The data on hospital volume Netherlands19,688 Pancreatic Mortality rates. Hospital 5 volume groups: and mortality after PR are too patients. Resection. volume. <2 . 2 to 5. 5 to 10. 10 to 20 . and > 20 characteristics, pathology. resections per year. heterogeneous to perform a meta-analysis, but a systematic review shows convincing evidence of an inverse relation between hospital volume and mortality and enforces the plea for centralization. The 10-year lasting plea for centralization among the surgical community dis not result in a reduction of Hospitalsthe mortality that rate perform after PRhigh or Adjusting independent volumeschange in of a easophageal referral pattern in Using X² test, simple logisticpredictor variables. Adjusted for Hospital Hospitals divided into three groups resectionThe Netherlands. have superior clinical Esophageal In-hospital demographics (age, sex, race), Maryland 1,136 patients; volume. based on annual procedural volume: regression, t-test, Wilcosin Resection. mortality, length and economic outcomes. By 154 1984- - US 52 nonfederal rank sum rtest, analysis comorbid disease, and severity of stay ; Hospital ҅3 procedures/year (low volume) ; referring these patients to 1999 acute- care of variance, and simple of illness. charges. 4 to 15 per year (medium volume) ; high volume centers, we may hospitals. linear regresion. Multiple 3 time ҆15 per year (high volume). improve quality and reduce Linear regression of Log- period : costs. 1984- transformed length of stay 1989 ; (LOS) and hospital charges 1990- was used to estimate the 1994 association of hospital 1995- volume after adjusting, 1999

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 141 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) After control factors adjusted, d’étude Adjusted factors of age, sex, Multivariate logistic patient insured by Medicaid Annual hospital volume: The Office of Statewide illness severity, neuromuscular Scoliosis. In-hospital Hospital regression models were were found to have a 155 1995- California 3,606 patients 5 or fewer ; 5.1 -25 ; 25.1-50 ; > 50. Planning and Development disease, surgical approach, volume. constructed to calculate signi cantly greater odds 1999 - US 25 years or mortality, surgical (OSHPD) California inpatient Medicaid status, and annual fi adjusted odds ratio for for complications, increased younger. complications, discharge database used a hospital volume on outcomes. outcomes. LOS. After multivariate logistic reoperations, retrospective review of all regression revelated an Length of stay. patients 25 years. inverse relationship between annual hospital volume and likelihood of reoperation, as patient treated at hospitals. The current data suggest that being insured with Medicaid in the state of California is associated with poorer outcomes after scoliosis surgery. Additionaly, this study documents a volume/ outcomes relationship in scoliosis surgery. US 12,948 156 1988- Despite the advent of less 2002 patients. Adjusted models of invasive techniques PCNL independent association Data from the Healthcare Bivariate analyses were remains a popular means of patient characteristics Hospital PCNL volume were : Cost and Utilization Project performed using the of managing stone disease. * Hospital and hospital qualities with PercutaneaousProlonged LOAS * Very low : 1 to 4 cases/year ; Nationwide Inpatient SampleChi-square and t-tests. For Although mortality was rare, it PCNL prolonged LOS after PCNL. Nephrolithotomy.and In-hospital *Low : 5 to 8 ; were abstracted. modeling purposes potential was significantly lower at high volume ; age, sex, race, insurance, mortality. * Medium : 9 to 15 ; covariates with at least a than at low volume hospitals. * Annual marginal association were admission type, region, Low short-term mortality rates *High : 16 to 27 ; hospital included in the multivariate Charlson Comorbidity inde coupled with shorter LOS and * and Very high : 28 or more ; discharge models. Logistic regressionHospital characteristics. high succes rates may make And Hospital discharge volume : volume. was performed using a PCNL increasingly palatable Very low : 1 to 8,284 discharge/year ; backward model building from a patient perspective and *low : 8,285 to 12,280 ; process. provide a potential basis for its * medium : 12,281 to 17,647; increasing use. *high :17,648 to 23,730 ; 157 1992- US 7,229 Medicare * and very high : 23,731 or more . 1995 patients.

Although volume-outcome Using Information from the Using mutlivariate logistic Major HPB Mortality rates. Hospital Average Hospital annual volume divided relationships have been Medicare claims database, regression to account Procedures volume. into quartiles : reported for many complex performed a national Cohortpotentially confounding Pancreatico- * Very low : <1/year ; surgical procedures, study. patients characteristics. duodenectomy. * Low : 1-2/year ; hospital experience is * Medium : 2-5/year ; particularly important with * High : (5 + year). pancreaticoduodenectomy. Patients considering this procedure should be given the option of care at high-volume referral center.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 142 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Pediatric shunt procedures d’étude Using Logistic regression. After multivariate adjustment Used population based performed at high-volume Hospital and Divided hospital/surgeon into quartilesnational : hospital discharge for demographics variables, hospitals or by high-volume Ventriculo- In-hospital 158 1998- US 5,955 patients surgeon * Low volume centers : database-study cohort. median household income forsurgeons were associated with peritoneal shunt.mortality rates ; 2000 admissions ; volume. < 28 admissions/year, ZIP code residence, primary lower in-hospital mortality rates, Length of stay 253 hospitals ; *High volume centers : care, type of emergency with no signi cant difference in (LOS), fi 411 surgeons. > 121 admissions/year. admission, and presence of LOS or hospital charges. total hospital * Low volume surgeon : infection, source of admission. charges. < 9 admissions/year, The volume-outcome Adjusted for age, sex, * High volume surgeon : relationship may not pertain Cross-sectional retrospectivePerformed multiple logistic Caucasian ethnicity, Medicaid Maryland Esophageal Mortality rates. Hospital Hospitals > 65 admission/year. classified by terciles of to some medical diseases 159 1992- study of hospital discharge and linear regression with status as a proxy for lower - US variceal volume. admissions as : such as esophageal variceal 1996 data. the robust generalized linearsocioeconomic status, transfer bleeding. * High (>17 cases of variceal bleeding bleeding. Alternatively, the model with the gamma familyfrom another hospital, and year per year), biases inherant in research and log link to determine of admission. * Medium : (12-17 cases per year) ; whether the outcomes were using administrative data may * or Low (< 12 cases per year) volume. related to hospital volume. make this relationship appear weaker for some medical than Very high surgeon volume is Retrospective cohort study. X²test were used to Adjusted odds ratio with surgical diagnoses in this type associated with a reduction Volume Divided into quartiles (operation patients characteristics, of study. Hospital and determine the association. A in surgical complications. Colon Cancer. 30-day per year) : surgeon and hospital 160 1992- US 22,672 patients surgeon series of logistic regression However the association postoperative * Low : 1-9 surgeons ; 1-58/hospitals ; characteristics. 1996 aged 66 years volume. models examined. between increasing hospital mortality ; * Medium : 10-16 surgeons ; 59-107 and older ; volume and post-operative 30-day PPIs hospitals 2,677 mortality appears to derive (postoperative * High : 17-26 surgeons ; 108 -150 surgeons ; mainly from a full spectrum procedural hospitals 660 hospitals. of clinical services that may intervention). * Very High : 27- 85 surgeons ; 151-341 facilitate the prompt recognition hospitals. and treatment of complications. Annual hospital procedure Hospital < 200; 200-399; 400+. Looks at the impact Longitudinal study. Age, sex, Charlson index, 161 1984- US, PTCA. In-hospital volume is associated to volume. of «learning » versus stent, LOS, AMI, multi-vessel 1996 California mortality, PTCA, PTCA volume et improved outcomes, although emergency by- improvements linked to time. squared volume. effect is small. No evidence of pass. learning by doing, rather time technology effect.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 143 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) d’étude Age, operative time. Although 5 cases represent Surgical High and low volume centers. A retrospective analysis.Comparisons were made with Pancreatico- Mortality, adult population who went very low volume compared Toronto, 5 patients volume. Wilcoxon’s rank sum 162 1993- duodenectomy. morbidity ; through the same operation. with high-volume adult centers, identified test used to calculate 2003 Ontario Postoperative minimal difference in operative as operated significance. Canada complications ; time, length of stay, and between 1993- LOS. mortality (none) were observed 2003. in this series. Pediatric patients Hospital(low volume) and surgeondo as well volume as Independant factors : age, race, haveadult apatients significant operated impact in on high in- Mortality was compared sex ; number of procedures ; hospitalvolume institutions.mortality and LOS after Hospital and individual surgeon volumeAll inpatient discharges among hospital volume levels Hospital and primary expected payer, radical cystectomy performed 163 1988- US 13,964 patients;Radical In-hospital of discharges per year were separatedafter radical cystectomy using Mantel-Haenszel Chi- surgeon Medicaid/Medicare, no charge,at a higher volume center may 1999 1,159 hospitals.Cystectomy. mortality ; into terciles : low, moderate, high. for bladder cancer from square test, while the LOS volume. self pay, bed size, location andresult in lower charges and length of stay the Health Care Utilization and charges were compared teaching status of the hospital,Lowshorter volume hospital providers stay while (LOS) and Project-Nationwide Inpatientusing ANOVA. Multivariate Patients characteristics and Multivariate analysis, and region, Comorbidities. anddecreasing surgeons the had likelihood higher of Inflation adjusted DataSample across were the included multiple in thelinear and logistic regressionclinical factors: age group, reoperationin-hospital mortality. rates to correct charge per Hospital and physicians were classi edanalysis. individual re-indentified ethnicity and comorbidity by Surgeon fidata sets representing careanalyses were used to adjust prolapse during the first subjects were longitudinallyCharlson Index, diabetes Sling surgery Postoperativeadmission. and Hospital as high or low volume providers in hospital inpatient, hospitalfor confounding factors. postoperative year. This 164 1999- US 1,356 sling tracked for 6-months beforemellitus ; chronic obstructive for Urinary complications, volume. based on their cumulative operative outpatient and physician suggests that high volume 2000 procedures the sling procedure to assespulmonary disease. Incontinence. concomitant or experience during the previous 2-year.of fice settings. providers are more likely to (5% of national comorbidities. Patients delayed prolapse For surgeons : low volume 1 to 3 slings diagnose and manage prolapse random tracked 12 months after repair, outlet per 2 years period ; mid volume 4 to 6 ; at the time of the sling, sample), operation for complications. obstruction high volume >7. For hospitals : 1 to 4 obviating the need for a second 988 providers. and repeat slings per 2 years period ; mid volume 4 operation. incontinence to 11 ; high volume >12. 165 1999- Australia surgery. 2001

In Australia, lower hospital Adjusting for demographics volume is not associated with Population-based study Investigated wheter unit characteristics such as Hospital Hospital volume birth : adverse outcomes for low risk Maternity. Rates of the National Perinatal Data size was an independant maternal age, patient insurance 750,491 volume of < 100; 100-500; 501-1000; 1001-2000 women. intervention Collection (NPDC). risk factor for each outcomeclassi fication, Aboriginal and women ; birth. and >2001 births per year. at birth and factor by conducting Torres Straits Islander status 331,147 neonatal multivariable logistic and geographic residential medically low mortality for low regression and multinomial status. risk ; risk women. logistic regression 132,696 adjusting for demographic primiparae ; characteristics. 198,451 multipare.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 144 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir There is wide variation in N° ou région Méthode TypeVariables of initial biopsy,contrôlées patient Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) diagnosis and treatment d’étude Data drom a single insurer The Chi-square test for age, TNM stage, Hospital Hospital and provider volumes were outcomes for patients with Volume of representing 25% of the independance was used volume, surgeon volume, and Outcome for grouped by the number of biopsy mammogram-detected breast Cancer : physicians regional population. Cohort to assess differences in radiologist volume as predictors 166 1995- US 224 tumors diagnosis procedures (1 ; 2-10 ; ҆11 procedures). carcinoma. Overall, practice Mammogram- and centers study databases. positive biopsy rates by variables. 1996 diagnosed by Detected Breastand treatment and radiologist provider volume. Factors that volume was correlated with EBNL, 604 Carcinoma. of breast volume. affected the use of BCS were the use of BCS but not with diagnoses ; carcinoma : evaluated in a multivariate the rate of positive biopsy rate 397 tumors. Rates for EBNL model. among high-volume of practice 2,796 patients ; and SCNB. is not a surrogate for quality 66 surgeons. in the diagnosis of breast carcinoma.

167 2000- London 187 patients. 2003 - UK

* The number Despite recent increases OesophagogastricSurgical of attempted Prospectiveconsultant Statistical analysis was Patients characteristics : in workload, high volume cancer. outcomes : oesophageal database. performed using an unpairedpre-operative diagnosis ; specialist units can deliver an * 30-day resections ; t-test. Histology of invasive efficient and timely service with mortality ; * The number tumours ; Location of invasiveboth good treatment outcomes *In-Hospital tumours, Pre-operative and minimal impact upon mortality ; TNM staging, Neoadjuvant elective surgical waiting lists of consultants. *length of therapy (Chemotherapy, and ICU provision. hospital stay ; Chemoradiotherapy). This nationwide population- Patients characteristics : Age, * postoperative Multivariate analysis, Chi- based survey of the treatment All new case of colorectal Sex, location of residence, complications. Surgeon Surgeon and Hospital were categorized squared tests and Logisitc of colorectal cancer patients cancer registred at each type of Hospital, presentation, Colorectal Postoperative and Hospital into two groups according to the number regression analyses were suggests that the delivery of 168 2000 Australia 2,383 Australian State Cancer Cancer site, preoperative Cancer. death within 30-volume. of patients with colorectal cancer seen used to determine levels of care by surgeons (the majority) questionnaires Registry were entered into diagnosis and preparation for days. by each surgeon over the survey period. statistical significance. who treat patients with rectal and 2,015 were surgery. Outcomes the survey. Generated a cancer infrequently should be returned ; including Duke’s questionnaire that was sent evaluated. * 550 surgeons. Correlations and probabilitiesState at diagnosis, age, If, as in this study, survival stage, adjuvant Theto the present treating study surgeon. is a Breast Cancer. 5-year relative Hospital Average hospital annual case volume : treatment. correlates with stage at 169 1980- US 2,409 Breast treatment. are presented. 1 to 10 ; 11 to 20 ; 21 to 30 ; 31 to 40 ;review of cases from the diagnosis and not with case 1995 cancer cases ; Survival rates. annual Breast above 40. Blue Mountain Regional volume, then improving survival 9 hospitals ; case volume. Tumor Registry and Test that requires identifying cases 5-year period. hypothesis. earlier. To do that requires improving saturation of the population at risk with effective screening and improving access to healthcare. That implies dispersing services instead of concentrating them in high-volume centers.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 145 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) Surgeon-speci c experience as d’étude Age, sex, comorbid fi measured by procedure volume All statistical test of illness, cancer stage, and A retrospective population- can have a signi cant impact Rectal Cancer. Hospital and Surgeon volume quartiles : association were corrected socioeconomic status were fi 170 1992- 5 States ; 2,810 rectal Mortality at Hospital and based cohort study on survival for patients with * 5-year case volume : for «clustering». To used to adjust for differences in 1996 and 6 cancer 30-days and 2 Surgeon utilizing the Surveillance, rectal cancer. * Very-Low : 1-5 ; examine overall survival case mix. US patients ; years, specific epidemiology and End Result * Low : 6-11 ; we constructed a Cox metropolitan420 hospitals ; Overall survival,procedure (SEER)-Medicare linked * Medium : 12-20 ; proportional hazards model surgeon The rate of volume. database. * High : 21 - 57. corrected for either within- volume abdominoperineal surgeon, or within-hospital Infant caracteristics at birth ; identified for resections. Retrospective study of of Random effects logistic Low birth weight. Mortality prior Annual correlation. mode of delivery ; sex ; race ; 2,603 patients ; Vermont Oxford Network regressions with mortality 171 1995- US 332 hospital/ number prenatal care (yes/no) ; median 1,121 surgeons. to discharge at database. as a dependent variable, 2000 NICU (40% of of NICU income and education by ZIP home. controllingfirst for patient Nicu in US), admissions of code. Hospital ownership; level and then hospital level 94,110 low VLBW infants. teaching status. caracteristics. Number of low birth infants that weight infants a hospital treats each year does born between not reliably predict futur mortality 1995 and rates. Direct measures are 2000. better. Increase in volume was associated with a large decrease in mortality. But patient volume of NICUs were significantly Hospitalsassociated operating with mortality at full rates. Poisson regressions. In model1. Individual patient Looks at the variation in daily work loadEstimates two models (100%) or over capacity may Adverse events Likelihood of an Separate regression for eachrisk : age, sex, DRG categories, Occupancy (patient volume per nurse) and AE. of probability of AE. One experience higher rates of 172 2000- US 4 hospitals (AE). AE in a given hospital. intensive care. Compares w/respect to average. controlling for volume. adverse events. 2001 in 2 states. day. Used 19 rates, In Model 2. adds days of the Random indicators of AE admission per week, emergent admission, sample of including: day, discharge workload. 24,676 adults wound infections,per day, discharged admisssion/ from Oct 2000 discharge to Sept 2001. urinary infections,turnover. 6,842 cases postoperative Plus intensity Emergency occupancy is with adverse stroke, falls, (complexity of Does not control for case-mix. Retrospective analysis/ Age, season, day of the week, Australia 1 hospital Emergency. 10-day mortality. Shifts are associated with increased 173 48 weeks events. adverse drug cases). Comparisons of two cohort timeof of the day. divided into 2 in 2002- (Canberra) ; events. patients. in-hospital mortality at 10 days 2004 about 66,500 overcrowded after controlling for seasonal, patients. or not, shift, day of the week effects. following mean occupancy.

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008 - 146 -

Pays et/ Période Echantillon Intervention Indicateur Indicateur Seuil utilisé pour définir N° ou région Méthode Variables contrôlées Résultats étudiée de l‘étude étudiée de résultat volume haut/bas volume (n) d’étude Retrospective analysis. Hazard scale based on Age, diagnosis, referral source,Hospital and ED overcrowding Australia 3 hospitals, Emergency. Deaths on day 2, Hospital occupancy was scored 1, 2, 174 July hospital and ED occupancy.urgency and mode of transportis associated with increased 7 and 30. 3 corresponding occupancy levels 2000- 62,495 to hospital. mortality. (<90%; 90-99%; >100%). Access block June admissions. occupancy scored : 1) <10%; 2) 10%- 2003 19% 3) >20%.

Both hospital volume Hospital volume (<100 versus >100).Very good case-mix control. Logistic regression. Demographic (age, sex, race) Pediatric cardiacRisk adjusted Sugeon and surgeon volume are 175 1992- US- New 7,169 cases Surgeon volume (<75 versus >75). and diagnostic risk factors surgery. mortality. and hospital significantly linked to in-hospital York (children) in 1995 volume. (previous operations, renal mortality, and these differences 16 acute care dysfunction, pulmonary persist both for high-complexity hospitals. hypertension, …) and 4 and low-complexity procedures. complexity categories (very detailed).

Laure Com-Ruelle, Zeynep Or, Thomas Renaud, avec la collaboration de Chantal Ambroise et Anna Marek Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature IRDES - Décembre 2008

Achever d’imprimer le 2 février 2009 TOP CHROMO 6, place du Village - Parc des Barbanniers 92632 Gennevilliers Cedex Dépôt légal : Février 2009 LE SERVICE DOCUMENTATION DE L’IRDES

Achats de publications Sur place ou par courrier : IRDES Service Diffusion 10, rue Vauvenargues 75018 Paris Horaires d’ouverture : Lundi au vendredi : 9 h 30 - 12 h / 14 h - 16 h

Consultation du fonds documentaire Sur rendez-vous tél. : 01.53.93.43.51 Horaires d’ouverture : Lundi, mardi, jeudi: 9 h - 12 h 15

Informations sur les travaux de l’IRDES Toute personne peut demander à être intégrée dans le fichier des correspondants de l’IRDES afin d’être régulièrement informée des dernières publications. tél. : 01.53.93.43.17/06 - fax. : 01.53.93.43.50

INTERNET : E-MAIL & WEB

Vous pouvez communiquer avec l’IRDES aux adresses suivantes : avec le service Diffusion E-mail : [email protected] avec le service Documentation E-mail : [email protected] avec le service informatique des logiciels Eco-Santé E-mail : [email protected]

www.irdes.fr notre serveur Web L’IRDES met à votre disposition un serveur sur Internet Ce service vous permet d’obtenir en direct des informations sur l’IRDES : Présentation du centre et de l’équipe Informations sur les enquêtes de l’IRDES Informations sur les logiciels Eco-Santé Consultation des bibliographies chronologiques et thématiques Description du fonds documentaire Commandes de rapports Volume d’activité et qualité des soins dans les établissements de santé : enseignements de la littérature

Si la concentration de l’offre de soins hospitaliers dans de grandes unités est souvent présentée comme un moyen d’améliorer la qualité des soins, le sens et l’ampleur de la relation entre volume d’activité et qualité des soins font toujours l’objet de débats. Ce rapport dresse un examen complet et systématique de la littérature portant sur la relation entre le volume d’activité dans les établissements de santé et les résultats des soins. Au total, 175 articles ont été évalués selon un protocole standard. Par ailleurs, nous avons étudié les concepts d’économie industrielle pour en comprendre les mécanismes sous-jacents du lien entre volume d’activité et résultats.

Cette revue montre que pour certaines procédures et interventions, en particulier pour la chirurgie complexe, la possibilité d’améliorer la qualité des soins lorsque le volume d’activité augmente est réelle. Il semble que l’effet d’apprentissage au niveau individuel (chirurgien) mais aussi au niveau de l’hôpital (transfert de connaissances, mode d’organisation) explique une grande partie de cette corrélation.

Ce lien de causalité entre volume et qualité doit par ailleurs être nuancé : les résultats sont sensibles à la nature des procédures et interventions analysées ainsi qu’aux seuils d’activité retenus. Plus l’intervention est spécifique et complexe, plus la corrélation volume-qualité est affirmée. Pour la plupart des interventions, il n’existe pas de seuil d’activité unanimement accepté. De plus, certaines études montrent que la relation volume-qualité devient marginale au-delà d’un seuil qui peut être relativement bas.

Mots-clefs : hôpital, chirurgie, volume d’activité, qualité des soins, centralisation, théorie de la courbe d’apprentissage, économie d’échelle, revue de la littérature.

* * * The volume and quality of health care in hospitals: lessons from the literature

The concentration of hospital activity in larger hospital units is often presented as a means to improve the quality of care. Nevertheless, the extent of the relationship and the direction of causality between hospital volume and health outcomes are still being debated. This report provides a comprehensive and systematic review of the literature on the relationship between hospital volume and quality of care. In total, 175 articles were analysed using a standard procedure. Furthermore, the concepts and theory from industrial economics were reviewed to understand the direction of causality, and the underlying dynamics which explain this causality.

The results of this review show that for certain procedures and interventions, in particular for complex surgery, there seems to be a real margin for improving outcomes by increasing activity volumes. The presence of a learning curve both at the individual (learning by doing) and hospital (knowledge transfer, organisation of work) levels seems to be particularly relevant for understanding this relationship.

However, the link between volume and quality should be interpreted with caution: the results are sensitive to the type of procedure or intervention studied as well as the activity thresholds. The more complex the intervention, the better confirmed the correlation “volume-quality”. For most procedures there is no single minimum volume threshold which emerges from the literature. Furthermore, some studies show that the relationship between volume and outcome might become marginal after a relatively low level of activity.

Keywords: hospital/surgeon volume, outcomes, centralization, regionalization, quality of care, theory of learning curve, scale economies, methodology, literature review.

www.irdes.fr I.S.B.N. : 978-2-87812-349-4 Rapport n° 545 (biblio n° 1734) Prix net : 30 euros