Sepsis in intensive care patients: challenges in diagnosis and prognostication
Peter Klein Klouwenberg Sepsis in intensive care patients: challenges in diagnosis and prognostication PhD thesis, University of Utrecht, the Netherlands ISBN: 978-94-6182-528-5 Lay-out: Off Page, www.offpage.nl Printing: Off Page, www.offpage.nl © Peter Klein Klouwenberg, Utrecht, the Netherlands. All rights reserved. No parts of this thesis may be reproduced, stored in a retrieval system or transmitted in any form or by any means without permission of the author. The copyright of articles that have been published or accepted for publication has been transferred to the respective journals. The research presented in this thesis was performed within the framework of CTMM, the Center for Translational Molecular Medicine (ctmm.nl), project MARS (grant 04I-201). The printing of this thesis was kindly supported by Immunexpress, SAS Institute BV, Chipsoft BV, Pfizer, and Astellas Pharma BV. Sepsis in intensive care patients: challenges in diagnosis and prognostication
Sepsis in intensive care patiënten: uitdagingen in diagnostiek en prognostiek (met een samenvatting in het Nederlands)
Proefschrift
ter verkrijging van de graad van doctor aan de Universiteit Utrecht op gezag van de rector magnificus, prof.dr. G.J. van der Zwaan, ingevolgde het besluit van het college voor promoties in het openbaar te verdedigen op vrijdag 23 januari 2015 des middags te 4.15 uur
door
Peter Marius Cornelis Klein Klouwenberg
geboren op 7 mei 1976 te Bergeijk Promotor: Prof.dr. M.J.M. Bonten Copromotor: Dr. O.L. Cremer TABLE OF CONTENTS
PART I: introdUCTION Chapter 1 General introduction 9
PART II: challengeS IN THE DIAGNOSIS OF SEPSIS Chapter 2 Classification of sepsis, severe sepsis and septic 17 shock: the impact of minor variations in data capture and definition of SIRS criteria Chapter 3 Interobserver agreement of CDC criteria for 33 classifying infections in critically ill patients Chapter 4 Presence of infection in patients with presumed 45 sepsis at the time of intensive care unit admission Chapter 5 Electronic implementation of a novel surveillance 61 paradigm for ventilator-associated events: easibility and validation
PART III: challengeS IN THE PROGNOSTICATION OF PATIENTS WITH SEPSIS Chapter 6 Intensive care unit-acquired infections after 83 admission for sepsis or non-infectious disease: incidence and attributable mortality Chapter 7 The attributable mortality of delirium in critically 101 ill patients: a prospective cohort study Chapter 8 Incidence and outcomes of new-onset atrial 123 fibrillation in critically ill patients with sepsis: a cohort study Chapter 9 Predicting evolution of disease severity in critically 135 ill patients with severe sepsis or septic shock
PART IV: diSCUSSION AND SUMMARY Chapter 10 General discussion 153
Nederlandse samenvatting 165 Dankwoord 171 About the author List of publications 173 Curriculum vitae 175
PART I
INTRODUCTION
1 General introduction
Peter M.C. Klein Klouwenberg
General introduction 11 1 . , . 6 4 1 2 . 3 . The panel defined . The panel defined 9 . 10 , which means “rotten flesh and putrefaction” putrefaction” and flesh “rotten means which σηψις, . It was soon understood that particularly patients with patients with . It was soon understood that particularly 7 . Finally, in 1914, Schottmüller reported that the release of pathogenic of pathogenic that the release in 1914, Schottmüller reported . Finally, 5 . More than 30 years ago, a group of experts proposed a clinical definition a clinical definition of experts proposed 30 years ago, a group than . More 8 The word “sepsis” was first introduced by Hippocrates (ca. 460-370 BC) and is is by Hippocrates (ca. 460-370 BC) and introduced “sepsis” was first The word Most patients with severe sepsis and septic shock are admitted to intensive care units units admitted to intensive care and septic shock are sepsis severe Most patients with patients of and the annual costs for the treatment on mechanical ventilation, and are alone US$17 billion in the United States estimated to exceed with sepsis are changing our modernof the term understanding has only “sepsis”. Nonetheless, it solely a consequence of that the morbidity of sepsis is not appreciated been recently deleterious consequences itself, but that its’ the virulent activity of the microorganism to the of the host response effects also caused by the indirect as importantly, are, disease which developed lung In 1967 Ashbaugh observed a severe invasion. microbial compliance, and lung loss of shortness of breath, severe with critically ill patients in infiltration alveolar diffuse the concept uniform clinical trials. They introduced inclusion in for sepsis to improve body alterations in (SIRS), involving response syndrome” a “systemic inflammatory of rate and leukocyte counts heart rate, respiration temperature, Celsus, a Roman encyclopaedist, described the cardinal clinical signs of localized signs of localized clinical described the cardinal Celsus, a Roman encyclopaedist, . and tumor in his extensive book De Medicina dolor, calor, rubor, acute inflammation, laesa, or added a fifth manifestation: functio Pergamon, Galen of Somewhat later, made some of the first Leeuwenhoek until the late 17th century, loss of function. Not the same time another Dutch physician, Around descriptions of bacteria, “animalcules”. the cause for sepsis. However, in the air were Boerhave, thought that toxic substances some of the founders of modern microbiology, it took another two centuries before Semmelweis, and Lister postulated the link between bacteria including Koch, Pasteur, and infection sepsis of perspective Historical its injures infection to a severe response body’s that arises when the syndrome Sepsis is a and mortality cause of in-hospital morbidity and increasing own tissues. It is a major sepsis suffered from this so-called Adult Respiratory Distress Syndrome (ARDS) and (ARDS) and Syndrome Adult Respiratory Distress this so-called from sepsis suffered In the 1980s it was reaction. of an inflammatory that its development was a result in the lungs but in the reaction was not only apparent that this inflammatory discovered whole body germs into the bloodstream was responsible for systemic symptoms and signs was responsible germs into the bloodstream derived from the Greek word word the Greek derived from sepsis as SIRS caused by clinically suspected infection, “severe sepsis” as sepsis sepsis” as sepsis sepsis as SIRS caused by clinically suspected infection, “severe dysfunction and “septic shock” as sepsis complicated by complicated by acute organ as to referred resuscitation. These definitions, generally to fluid hypotension refractory trials, sepsis in many clinical criteria inclusion as used criteria”, have been “Bone the unchanged largely and until today have remained
. Its aim is to provide solutions for for solutions . Its aim is to provide 11,12 provides an assessment of theof assessment an provides Chapter 4 Chapter 3. present a model that predicts the evolution of the evolution of model that predicts a Chapter 9 present tline utline aim and o Thesis mortality the morbidity and of sepsis, in the management advances Despite recent with sepsis are high. Patient populations unacceptably remain caused by sepsis the underlying infectious to the implicated pathogen, with regard heterogeneous outcome the which makes it demanding to improve profile, disease and their genetic of sepsis the diagnosis of sepsis and the prognostication of sepsis. In particular, and Risk the Molecular Diagnosis For these reasons, challenging. patients remain initiative and biobank study was started (MARS) data collection Stratification of Sepsis units in 2010 intensive care in two Dutch tertiary the current shortcomings in the diagnosis and prognostication of critically ill patients patients of critically ill diagnosis and prognostication shortcomings in the the current of this initiative. describes some of the results sepsis. This thesis with presumed the diagnostic challenges in critically ill patientsThe first part of this thesis focuses on in the definition andwith sepsis. Chapter 2 describes the influence of minor variations on the observed incidences of sepsis, failure of SIRS criteria and organ measurement classifying infections of for agreement interobserver and septic shock. The sepsis severe criteria in a and Prevention to Centers for Disease Control various subtypes according in described population is ICU mixed likelihood of infection in patients who were treated for sepsis upon admission to the ICU, for sepsis upon treated likelihood of infection in patients who were Chapter 5 plausibility of infection and mortality. and quantified the association between validation study of a novel surveillance system for complications of mechanical a presents pneumonia. of ventilator-associated surveillance ventilation as an alternativethe current to of important complications of sepsis in In the second part of this thesis, the prognosis a comparative analysis of hapter 6 we present critically ill patients is described. In C of secondary infections in patients admitted the incidence and attributable mortality attributable mortality caused the 7 presents to the ICU with or without sepsis. Chapter delirium, in critically ill patients. The incidence by an important complication of sepsis, sepsis is with ill patients of critically outcomes of atrial fibrillation in a cohort and in Chapter 8. provided disease for individual patients admitted for sepsis, by estimating daily probabilities for sepsis, by estimating daily probabilities disease for individual patients admitted using ICU, the in week one after failure organ multiple or death to progression of available parameters. routinely in Chapter 10. and general discussion is provided A synthesis of the results
General introduction 1 12 General introduction 13 1 J Infect Dis 1980;141(1):55-63. Intensive Care Med 2012;38(5):811-9. Intensive Care primary role of lymphoreticular cells in the cells in the of lymphoreticular primary role to bacterial of host responses mediation endotoxim. FB, Dellinger RP, Bone RC, Balk RA, Cerra et al. Definitions for Fein AM, Knaus WA, for and guidelines failure sepsis and organ therapies in sepsis. the use of innovative Conference The ACCP/SCCM Consensus College of Chest Committee. American Medicine. Care Physicians/Society of Critical Chest 1992;101(6):1644-55. sepsis and Severe T. Angus DC, van der Poll 2013;369(9):840- septic shock. N Engl J Med 51. PM, Ong DS, Bonten MJ, Klein Klouwenberg OL. Classification of sepsis, severe Cremer sepsis and septic shock: the impact of minor SIRS of definition and capture data in variations criteria. PM, Ong DS, Bos LD, de Klein Klouwenberg Huson MA, et Beer FM, van Hooijdonk RT, of Centers for agreement al. Interobserver criteria for and Prevention Disease Control classifying infections in critically ill patients. Med 2013;41(10):2373-8. Crit Care 9. 10. 11. 12.
Am J Respir Crit Care Med Care Crit Respir J Am Martin GS, Mannino DM, Eaton S, Moss M. M. Moss S, Eaton DM, Mannino GS, Martin United of sepsis in the The epidemiology 2000. N Engl J 1979 through States from Med 2003;348(16):1546-54. changing report, WHO. The world health sex and mortality history: Deaths by cause, estimates for 2002, stratum in WHO regions 2004. Lidicker WT, Angus DC, Linde-Zwirble J, Pinsky MR. J, Clermont G, Carcillo sepsis in the United Epidemiology of severe outcome, and States: analysis of incidence, Med Crit Care associated costs of care. 2001;29(7):1303-10. the Marshall JC. Sepsis: rethinking J Leukoc Biol research. to clinical approach 2008;83(3):471-82. Vincent JL, Abraham E. The last 100 years sepsis. of 2006;173(3):256-63. und Behandlung der Schottmüller H. Wesen Sepsis. Inn. Med 1914;31:257-280. Levine TL, Petty DB, Bigelow DG, Ashbaugh in adults. distress BE. Acute respiratory Lancet 1967;2(7511):319-23. RN, McGhee JR, Michalek SM, Moore SE. The DL, Mergenhagen Rosenstreich s Reference 1. 2. 3. 4. 5. 6. 7. 8.
PART II
CHALLENGES IN THE DIAGNOSIS OF SEPSIS
2 Classification of sepsis, severe sepsis and septic shock: the impact of minor variations in data capture and definition of SIRS criteria
Peter M.C. Klein Klouwenberg David S.Y. Ong Marc J. Bonten Olaf L. Cremer
Intensive Care Medicine Abstract and measurement in the definition of minor variations the effects quantify Purpose: To on the failure (SIRS) criteria and organ syndrome response of systemic inflammatory sepsis and septic shock. of sepsis, severe observed incidences study in a tertiary intensive observational conducted a prospective, Methods: We of A total 2010. and October 2009 January between Netherlands unit in the care determined lower limits the upper and included. We adults were 1,072 consecutive of the use of an of sepsis by evaluating the influence incidence of the measured in the number method of data collection, and variations manual automated versus a SIRS criteria, and duration of abnormal of values required of SIRS criteria, concurrency diagnosis. to make a particular to setting) restrictive (most 49% from varied SIRS of incidence measured The Results: sepsis and the incidences of sepsis, severe 99% (most liberal setting). Subsequently, 4 to 9% for the most from 6 to 27%, and 22 to 31%, from septic shock ranged from In non-infected settings, respectively. versus the most liberal measurement restrictive still 17 to 6% of patients without whereas patients, 39 to 98% of patients had SIRS, SIRS had an infection. incidence of sepsis heavily depends on minor variations Conclusions: The apparent a consequence, the current As recording. data in the definition of SIRS and mode of of patients into sepsis trials. uniformrecruitment consensus criteria do not ensure
Classification of sepsis 2 18 Classification of sepsis 19 2 . . 3 6
, their occurrence has proven has proven , their occurrence 2 . For these reasons, a general notion has developed that the a general notion has developed that the . For these reasons, 3 . An expanded list of signs and symptoms was introduced to to was introduced . An expanded list of signs and symptoms . According to these definitions, the diagnosis of sepsis requires requires the diagnosis of sepsis to these definitions, . According 5 1 . In addition, the extent to which SIRS criteria are present depends on the depends on the present to which SIRS criteria are . In addition, the extent 3,4 As a result of poor general acceptance by clinicians and accumulating experience poor general acceptance by clinicians and accumulating experience of As a result Despite the deceptive simplicity of this concept, the application of the SIRS criteria of the SIRS criteria simplicity of this concept, the application Despite the deceptive to be neither specific for infection, nor robust across various clinical and research research and clinical various across robust nor infection, for specific neither be to settings gathered during the conduction of clinical trials, the ACCP/SCCM definitions of sepsis the conduction of clinical trials, the ACCP/SCCM definitions of sepsis during gathered in 2001 updated were Today, computerized systems have become widely available and monitor data are are computerized systems have become widely available and monitor data Today, it remains Third, second. even every or every minute being recorded increasingly thus positive. normal to be considered undefined how long SIRS criteria should deviate from since Last, many diagnostic elements have been updated or adapted to local needs as For instance, the PROWESS study defined sepsis the first consensus conference. consensus definitions do not adequately reflect the complex pathophysiology of sepsis, sepsis, the complex pathophysiology of reflect consensus definitions do not adequately to infection host response of the staging nor prognostication allowing neither precise clinical evidence of infection and the presence of a systemic inflammatory response inflammatory systemic a of presence the and infection of evidence clinical alterations, including by specific physiological (SIRS) that is characterized syndrome rate. blood cell count, heart rate and respiratory white aberrations in temperature, uction Introd (ACCP/SCCM) Medicine Care Critical of Physicians/Society Chest of College American The Care Intensive in used are sepsis severe and sepsis for definitions conference consensus Units (ICUs) worldwide time since the start of infection as well as resuscitation. Many researchers involved in involved in Many researchers as resuscitation. time since the start of infection as well a patient, who by clinical judgment clearly sepsis trials have thus felt frustrated when met or a study because not all SIRS criteria were had sepsis, could not be accepted into in time could be recorded better describe the clinical response to infection, including the presence of elevated elevated of the presence including infection, to response describe the clinical better other ileus, and capillary refill, lactate, decreased and protein C-reactive levels of better describe sepsis Although taken together these criteria failure. markers of organ and interrater variability. subject to interpretation more as a clinical disease, they are settings. Moreover, used in research only rarely Hence, these additional criteria are First, observers may consider of variation. sources obscure more other, may be there ways when making a sepsis diagnosis. of the SIRS criteria in different the concurrency of two or three the occurrence setting sepsis would typically require In a research in clinical practice it whereas time window, physiological aberrations within a 24-hour Second, the is diagnosed only when most or all of the criteria occur simultaneously. use of computerized patient data management systems may have changed increased detected. At the time of the consensus conference the way in which SIRS criteria are were of vital signs and laboratory results in 1991, systems for automated capture typically at one-hour intervals manually, uncommon, with data being recorded in diagnosing sepsis is not straightforward. Although SIRS symptoms are clearly clearly are Although SIRS symptoms is not straightforward. in diagnosing sepsis and length of stay mortality associated with increased , 7 . 8 . 10 . In order to explore to explore . In order 1,7,8,11-13 able 1) . Antibiotic use and culture results of samples of samples results . Antibiotic use and culture 9 With regard to these issues, we aimed to quantify to what extent minor variations in in minor variations to what extent we aimed to quantify to these issues, With regard obtained from two days before admission up until the end of the first day in ICU were of the first day in ICU were admission up until the end two days before obtained from data, a final post-hoc diagnosis of infection also collected. Based on these combined according managed sepsis were with patients All of the authors (PK). made by one was based on the Surviving Sepsis Campaign guidelines to locally adapted protocols the range of measured incidences of SIRS, we investigated several clinically relevant incidences of SIRS, we investigated several clinically relevant the range of measured To analyze automated and minor variations in data definitions. methods of data recording (non- data recorded continuously we compared versus manual methods of data capture, recorded validated values with automated filtering for artifacts and outliers) and hourly number also evaluated variations in the required data (manually validated values). We versus PROWESS study definitions), the of positive SIRS criteria (consensus conference mechanical (either including or excluding respiratory score definition of the precise of positive SIRS ventilation as a separate criterion), the definitions for concurrency at any time-point in a 24-hour observation window versus criteria (criteria present derangements duration of physiological the minimum required simultaneously present), (single or 5-minute aberration versus 60-minute continuous period of abnormality), and and admission ICU since passed time at both looking (by measurements of timing the the paper: used the following definitions of SIRS throughout at the hour of the day). We whereas the KyberSept study used different cutoff values for individual measures measures values for individual cutoff study used different the KyberSept whereas the frequency, timing and method of data capture may affect the diagnosis of sepsis, the diagnosis of sepsis, may affect data capture timing and method of the frequency, sepsis and septic shock. severe Measurement of SIRS and sepsis Measurement definitions for reported screened papers were landmark sepsis trials and review Various sepsis and septic shock (T of SIRS, sepsis, severe ethods s and m Material patients Study design and performed patients admitted to a study in consecutive adult an observational We January 2009 and October ICU in the Netherlands between mixed 36-bed tertiary of < 96 hours and with an uncomplicated stay elective surgery 2010. Patients following analysis. excluded from were room other ICUs or the recovery from patients transferred collected during the and co-morbidities were Demographic data, admission diagnoses, intervals and at one-minute recorded automatically first 24 hours in ICU. Vital signs were frequently or more collected once daily, were validated hourly by nurses. Laboratory data as determined prospectively by clinical need. By the end of the first day ICU clinicians based on Dutch National whether they thought a patient had an infection recorded Evaluation (NICE) criteria Intensive Care suspected infection plus three or more SIRS criteria (as opposed to two or more) two or more) (as opposed to SIRS criteria or more infection plus three suspected
Classification of sepsis 2 20 Classification of sepsis 21 2 . 7 7 7 8 8 1 8 12 16 15 35 N/A N/A N/A N/A 2 36 12 35 7 11-13 1 7 13 17 1 11-13 35 References 1 7 11-13 35 1 7 11-13 35 1 7 11-13 35 , (2) as a reduction , (2) as a reduction 1
a /L 9 , or (3) as a systolic arterial blood pressure < < pressure blood arterial systolic a as (3) or , /L or > 10% immature forms /L or > 10% immature 9 17 /L or > 10*10 9 /L or > 12*10 9 Refractory hypotension SOFA (≥ 3 (single) (≥ 3 (single) SOFA (≥ 8 (aggregate) SOFA LODS MODS ‘PROWESS’ ≥ 3 criteria ≥ 3 criteria ≥ 2 criteria ≥ 2 criteria Refractory hypotension with other organ system failure system failure Refractory hypotension with other organ < 3.5*10 > 100/min > 90/min > 90/min Single observation above threshold Single observation above threshold Continuous period above time window Occurring within 24 hours Contemporaneous occurrence admission At any time-point following admission Early versus late after < 36.0°C or > 38.0°C < 35.5°C or > 38.5°C > 20/min or pCO2 < 32 mm Hg > 20/min or pCO2 < 32 mm Hg or MV > 30/min or pCO2 < 32 mm Hg or MV < 4*10 Automatic (continuous) recording Automatic (continuous) Manual (hourly) recording
Sources of variation in the diagnosis of SIRS, sepsis, severe sepsis and septic shock sepsis and septic sepsis, severe in the diagnosis of SIRS, of variation able 1: Sources Refractory hypotension was defined (1) as a systolic blood pressure < 90 mm Hg or a reduction reduction < 90 mm Hg or a Refractory hypotension was defined (1) as a systolic blood pressure Shock Organ failure Organ Definition of syndrome Number of criteria in systolic blood pressure of ≥ 40 mmHg from baseline, despite adequate volume resuscitation, baseline, despite adequate volume resuscitation, of ≥ 40 mmHg from in systolic blood pressure hypotension of cause other of absence the in 90 mm Hg or a mean arterial pressure < 70 mm Hg for at least 1 hour despite adequate fluid fluid < 70 mm Hg for at least 1 hour despite adequate 90 mm Hg or a mean arterial pressure in an attempt to volume status or the use of vasopressors adequate intravascular resuscitation, of ≥ 70 mm Hg of ≥ 90 mm Hg or a mean arterial pressure maintain a systolic blood pressure of ≥ 40 mm Hg from baseline in the absence of other causes for hypotension baseline of ≥ 40 mm Hg from a T collection in mode of data Differences Abbreviations: MV – mechanical ventilation, SOFA – Sequential Organ Failure Assessment, LODS – Failure – Sequential Organ ventilation, SOFA MV – mechanical Abbreviations: N/A – not applicable. Dysfunction Score, MODS – Multiple Organ Dysfunction Score, Logistic Organ White blood cell Heart rate of criteria above threshold of criteria Concurrency of measurement Timing Definition of abnormal values Temperature Respiratory rate Respiratory rate Mode of data recording time Required . To 7 /L or > /L or > 9 . 1,7,12,17 . 18 , the Logistic Organ , the Logistic Organ /L or > 12*10 15 9 , and an adapted score used in the PROWESS study used , and an adapted score 16 , the Multiple Organ Dysfunction Score (MODS) Score Dysfunction , the Multiple Organ 14
shows important patient characteristics. Cardiac surgery, neurosurgery neurosurgery surgery, Cardiac able 2 shows important patient characteristics. T shows the measured incidences of the four individual SIRS criteria by incidences of the four individual SIRS criteria by 1 shows the measured Figure incidence of SIRS by various criteria for concurrency 2 shows the measured Figure and neurology were the most common referring specialities. Hospital mortality rate rate mortality specialities. Hospital referring most common the were neurology and IV model by the APACHE to 31% as predicted was 27%, compared 10% immature (band) forms, heart rate > 90/min, respiratory rate > 20/min, pCO2 < 32 32 > 20/min, pCO2 < rate forms, (band) > 90/min, respiratory heart rate 10% immature of methods relevant clinically two the i.e., settings, extreme most the used We Hg. mm incidences of SIRS, for apparent in the highest and lowest that resulted data recording the influence next evaluated of the incidences of sepsis. We the subsequent calculation Sequential sepsis: the severe of incidence the on scores failure organ used commonly of and an aggregated of ≥ 3 (both as a single score score Assessment (SOFA) Failure Organ of ≥ 8) score assess the incidence of septic shock we used several definitions of shock of septic shock we used several definitions assess the incidence During the study period 3,062 patients were admitted to our ICU, of whom 1990 admitted to our ICU, of whom 1990 were During the study period 3,062 patients uncomplicated elective analysis. Reasons for exclusion were excluded from (65%) were (n=143; room the recovery another ICU or 57%) and transfers from (n=1,747; surgery medical patients, 345 were analyzed: 637 were 1,072 patients were 8%). As a result, patients with a complicated elective surgical patients, and 90 were surgical emergency stay. s Result Statistical analysis data using SPSS version 17.0 for Windows analyzed We USA). (SPSS, Chicago, presented Data are study variables. computed for relevant Descriptive statistics were range (IQR) as deviation (SD) or medians with interquartile as means with standard tested normality test, and compared using the Kolmogorov-Smirnov We appropriate. test for nonparametric tests for continuous variables and chi-square using groups significant. considered were categorical variables. P values <0.05 temperature < 36.0 or > 38.0°C, white blood cell count < 4*10 > 38.0°C, white blood < 36.0 or temperature Dysfunction Score (LODS) Dysfunction Score frequency and method of observation. With frequency the the exception of the leukocyte count, occurrences when defined as single five-minute frequently more SIRS criteria occurred using 60-minute intervals. This applied both to settings of than when measured The inclusion of mechanical continuous automated and hourly manual data recording. of patients fulfilling the resulted in an increase ventilation in the definition of SIRS intervals and using 60-minute 55% to 95% when measured criterion from respiratory 87% of patients In all, as five-minute occurrences. 75% to 98% when measured from mechanically ventilated during the first 24 hours of ICU admission (data not shown). were for simultaneous occurrence of the individual physiological deviations. A requirement the to a situation where in a much lower incidence of SIRS compared did not result
Classification of sepsis 2 22 Classification of sepsis 23 2
Respiratory rate Respiratory ≥3 SIRS excluding≥3 mechanical ventilation mechanical Heart rate Heart ≥3 SIRS including≥3 mechanical ventilation mechanical ≥2 SIRS excluding≥2 mechanical ventilation mechanical White blood count cell White percentiles above threshold. Bars show means, whiskers show 95% confidence Bars show means, whiskers show 95% confidence above threshold. percentiles th ≥2 SIRS including≥2 Temperature mechanical ventilation mechanical 0 10 90 80 70 60 50 40 30 20
0 100
50 40 30 20 10 60 90 80 70 percentiles above threshold. Bars show means, whiskers show 95% confidence intervals. Data Bars show means, whiskers show 95% confidence intervals. Data above threshold. percentiles Percentage of patients with SIRS with patients of Percentage 100 SIRS with patients of Percentage th Incidence of the four SIRS criteria on day 1 in the ICU by variations in threshold criteria criteria in the ICU by variations in threshold 1: Incidence of the four SIRS criteria on day 1 Figure of recordings denote automatic (continuous) bars Dark shaded data recording. of mode and of 60-minute intervals above automatic recordings light shaded bars denote single occurrences, and striped of single occurrences, Speckled bars denote manual (hourly) recordings threshold. were Single occurrences of 60-minute intervals above threshold. bars denote manual recordings as defined were 60-minute occurrences threshold; defined as 5-minute continuous intervals above 90 Incidence of SIRS on day 1 in the ICU by variations of concurrency criteria. Shaded bars criteria. Shaded bars 2: Incidence of SIRS on day 1 in the ICU by variations of concurrency Figure time 24-hour a within present separately be to required were criteria SIRS where setting a show to be required SIRS criteria were the speckled and striped bars show a setting where window, of single occurrences, Dark shaded and speckled bars denote recordings simultaneously. present Only of 60-minute intervals above threshold. light shaded and striped bars denote recordings Single gave similar results). shown (manual recordings are automatic (continuous) recordings 60-minute occurrences defined as 5-minute continuous intervals above threshold; were occurrences defined as 90 were were filtered for artifacts and outliers. for artifacts and outliers. filtered were intervals. Data were filtered for artifacts and outliers. for artifacts and outliers. filtered intervals. Data were 3 (4.5%) 8 (2.4%) 60 (18%) 23 (7.1%) 15 (4.5%) 14 (4.2%) 65 (6.1%) 90 (8.4%) 241 (73.3) 66 (20.1%) 44 (66.7%) 22 (33.3%) 88 (26.7%) 62 (18.8%) 70 (50 – 91) 330 (30.8%) 284 (26.5%) 329 (30.7%) 147 (44.7%) 345 (32.1%) 159 (14.8%) 295 (27.5%) 336 (31.3%) 376 (35.0%) 652 (60.8%) 637 (59.4%) 1,072 (100%) 4.9 (2.4 – 9.7) 21.7 (10.5 – 39.7) 60.9 (48.0 – 70.9) Gram positives Gram positives Gram negatives Fungal/yeast Predicted Observed Pulmonary tract Abdomen Blood stream Urinary tract Soft tissues Central nervous system Other Community acquired Hospital acquired ICU (days) ICU (days) Hospital (days) Emergency room Emergency Hospital ward Operating Room Other Medical Medical Elective surgery surgery Emergency Patient characteristics at admission able 2: Patient characteristics Data are median (IQR) or number (%). (%). median (IQR) or number Data are Health Evaluation. - Acute Physiology and Chronic APACHE Abbreviations: Confirmed bacteremia Case fatality (hospital) Site of infection APACHE IV score IV score APACHE Confirmed infection Length of stay Age (years) ICU readmission Admission source T patients Number of Male gender Admission type Admission type
Classification of sepsis 2 24 Classification of sepsis 25 2 65 (6.1%) 47 (4.4%) 523 (48.8%) 234 (21.8%) Most restrictive setting Most restrictive 98 (9.1%) 329 (30.6%) 294 (27.3%) 1,056 (98.5%) 3 criteria simultaneously present using manual recording at at recording manual using present simultaneously criteria 3 ≥ Most liberal setting shows the observed incidences of SIRS, sepsis, severe sepsis and septic septic sepsis and incidences of SIRS, sepsis, severe able 3 shows the observed Using the most extreme settings for the detection of SIRS we next explored the the settings for the detection of SIRS we next explored Using the most extreme We next explored whether the timing of observations influenced the observed observed influenced the timing of observations whether the next explored We T 2 criteria transiently present during a 24-hour period of automatic recording to to during a 24-hour period of automatic recording 2 criteria transiently present Incidences of SIRS, sepsis, severe sepsis and septic shock on day 1 in the ICU ICU sepsis and septic shock on day 1 in the able 3: Incidences of SIRS, sepsis, severe
T SIRS Sepsis Severe sepsis Severe Septic shock during criteria transiently present two or more For SIRS and sepsis, the liberal criteria required criteria or more three criteria required the restrictive a 24-hour period of automatic recording; as defined was sepsis Severe intervals. hourly at recording manual with present simultaneously Septic to PROWESS (liberal) or MODS (restrictive). according failure sepsis (see above) plus organ refractory hypotension with refractory hypotension (liberal) or shock was defined as sepsis plus (restrictive). system failure other organ hourly intervals excluding mechanical ventilation. Using the most liberal settings all ventilation. Using the most liberal settings all hourly intervals excluding mechanical to 71% of patients when using the most 329 infected patients had SIRS, compared still setting. In non-infected patients, 39-98% of patients had SIRS, whereas restrictive infection. 17-6% of patients without SIRS had an 22-31% in occurred sepsis that found We sepsis. for incidences observed of range failure definitions of organ the use of various we explored of patients. Subsequently, of (≥ 3), or as a total score scores (both as separate and shock. The use of SOFA in resulted (≥ 8)), MODS, LODS, and PROWESS scores systems organ the six different of 73%, 30%, 54%, 56%, and 80%, respectively. failure of organ frequencies apparent sepsis severe with diagnosed been have would that patients of fraction the Accordingly, 4 to 9%. 6 to 27%. Likewise, the incidence of septic shock ranged from ranged from deviations occurred disparately within a 24-hour time period. An automated versus automated versus An time period. 24-hour a within disparately occurred deviations similar results. produced of data recording manual setting other settings, the of and sepsis (data not shown). Irrespective incidences of SIRS passed time with decreased criteria SIRS positive more or two having of probability admission 74% of patients had the first two hours following ICU since admission: during to 68% during the period that extended compared positive SIRS criteria, two or more sepsis and severe In contrast, the incidences of sepsis 12 to 24 hours (p <0.001). from The hour of day was of the first day of admission. stable during the course remained or sepsis, and also morning of SIRS evening versus with the occurrence not correlated in significant differences. did not result 6 p.m.) measurements (e.g. 6 a.m. versus restricted most the most liberal versus the shock on the first day in the ICU for 99% when defined as The incidence of SIRS ranged from modes of data capture. ≥ as when defined 49% shows the characteristics of patients who were diagnosed as having sepsis as having sepsis diagnosed patients who were characteristics of 4 shows the able When we applied the ACCP/SCCM definitions, we unexpectedly found that the that the we unexpectedly found the ACCP/SCCM definitions, When we applied definitions and measurement also found that even use of the most restrictive We T ion Discussion of potentially hidden variations in the method Our study demonstrates that minor, and shock can failure and subtle variations in the definitions of organ data capture sepsis and septic the observed incidences of SIRS, sepsis, severe influence greatly we found that the SIRS criteria are shock in a general ICU population. Furthermore, a thus ultimately not very helpful in making overly sensitive, very non-specific, and current the that notion the underpin findings These ICU. the in sepsis diagnosis clinical sepsis needs to be improved. ACCP/SCCM consensus definition of in similar observed incidences of SIRS and use of a computerized system resulted vital signs, of a patient’s Continuous recording sepsis as did manual data collection. events, seems of other outcomes and untoward although important for the detection our sepsis is detected. However, the likelihood that to increase not required therefore of values used for the definition in cutoff findings did demonstrate that minor variations events, these of concurrency for requirements the as well as deviations, physiological incidences of SIRS and sepsis. For instance, can have a huge impact on the observed patients as having sepsis about 50% more the use of ‘liberal’ settings would categorize was shock this difference sepsis and septic settings; for severe than using ‘restrictive’ in our study many clinical trials have used methodology that Interestingly, even larger. and misclassification, by the inclusion of patients would qualify as a very ‘liberal’ setting in some of these. In lack of efficacy without true sepsis, may have led to an apparent if patients, in may have been hampered contrast, the generalizability of study results could not be included because not sepsis to be present, whom clinicians considered patients we observed that Remarkably, met or recorded. SIRS criteria were required all setting, as having ‘sepsis’ using one particular measurement classified merely who were higher case fatality rates, than as well as scores, and SOFA had higher APACHE-IV liberal setting. The more sepsis’ using another, classified as ‘severe patients who were sepsis’ and ‘septic shock’. same held true when comparing patients with ‘severe a making settings could not fully compensate for the low specificity of SIRS criteria for or severe sepsis according to the most restrictive versus the most liberal definitions. In In most liberal definitions. versus the the most restrictive to sepsis according or severe stricter settings and measurement restraining using more severity classes, both disease IV scores, APACHE selection of patients who had higher resulted in the data definitions resulted in the liberal definitions contrast, using more In and length of stay. mortality, SIRS and patients into both the postoperative surgical more selection of significantly admission in differences no found We group. sepsis severe the not but sepsis groups, patients who were or number of co-morbidities among specialty diagnosis, referring by us. explored definitions of sepsis that were to the various selected according
Classification of sepsis 2 26
Classification of sepsis 27 2
Abbreviations: APACHE – Acute Physiology and Chronic Health Evaluation, SOFA – Sequential Organ Failure Assessment Failure Organ Sequential – SOFA Evaluation, Health Chronic and Physiology Acute – APACHE Abbreviations:
Mann-Whitney test. Mann-Whitney test. Chi-square (severe) septic or in septic shock in the most restrictive setting. setting. restrictive most the in shock septic in or septic (severe)
c c b
Excluding patients that were were that patients Excluding (liberal) or MODS (restrictive). Data are median (IQR) or number (%). APACHE= acute physiology and chronic health evaluation. evaluation. health chronic and physiology acute APACHE= (%). number or (IQR) median are Data (restrictive). MODS or (liberal)
a a
or more criteria simultaneously present with manual recording at hourly intervals. Severe sepsis was defined as sepsis plus organ failure according to PROWESS PROWESS to according failure organ plus sepsis as defined was sepsis Severe intervals. hourly at recording manual with present simultaneously criteria more or
For sepsis, the liberal criteria required two or more criteria transiently present during a 24-hour period of automatic recording; the restrictive criteria required three three required criteria restrictive the recording; automatic of period 24-hour a during present transiently criteria more or two required criteria liberal the sepsis, For
0.167 13.9) – (3.0 7.0 10.8) – (3.0 5.7 (days) stay of Length 0.138 16.5) – (3.7 9.0 12.9) – (3.1 7.0 0.870 17.3) – (3.5 12.8 16.7) – (4.7 9.0
c c c
Observed Observed 22 (23%) 22 81 (35%) 81 0.042 62 (27%) 62 23 (45%) 23 35 (54%) 35 28 (60%) 28 <0.001 0.243
b b b
Predicted 28 (29%) 28 91 (39%) 91 0.115 75 (33%) 75 23 (45%) 23 35 (54%) 35 29 (62%) 29 0.002 0.089
Case fatality Case
2 (4%) 2 5 (10%) 5 2 (3%) 2 13 (6%) 13 12 (5%) 12 14 (15%) 14 Neurology Neurology
5 (11%) 5 15 (29%) 15 10 (15%) 10 43 (19%) 43 44 (19%) 44 14 (15%) 14 Gastroenterology
20 (43%) 20 19 (37%) 19 23 (35%) 23 63 (28%) 63 70 (30%) 70 19 (20%) 19 Cardiovascular Cardiovascular
16 (34%) 16 10 (20%) 10 25 (39%) 25 94 (41%) 94 96 (41%) 96 36 (38%) 36 Respiratory Respiratory
Admission diagnosis Admission
Surgical admission Surgical 0.016 (28%) 65 (45%) 43 0.770 (29%) 19 (33%) 76 0.021 (21%) 10 (49%) 25
b b b
Co-morbidities (number) Co-morbidities 0.756 2) – (0 1 1.25) – (0 1 0.306 2) – (0 1 2) – (0 1 0.638 2) – (0 1 2) – (0 1
c c c
SOFA score SOFA 0.145 9) – (5 7 9) – (4 6 <0.001 12) – (9 10 8) – (5 6 <0.001 12) – (9 10 10) – (8 8
c c c
APACHE IV score IV APACHE <0.001 104) – (64 80 84) – (46 64 <0.001 118) – (86 104 89) – (57 73 0.002 135) – (95 113 115) – (67 88
c c c
0.698 68) – (51 62 71) – (50 63 0.681 71) – (51 61 71) – (45 63 Age (years) Age 0.976 69) – (51 62 74) - (45 61
c c c
(n= 95) (n= (n= 229) (n= p 234) (n= (n= 51) (n= p 65) (n= p 47) (n=
estrictive criteria estrictive R estrictive criteria estrictive R criteria iberal L estrictive criteria estrictive R criteria iberal L iberal criteria iberal L
a a a
Sepsis Severe sepsis Severe Septic shock Septic
Characteristics of of Characteristics virtual patient cohorts classified with sepsis, severe sepsis and septic shock septic and sepsis severe sepsis, with classified cohorts patient virtual 4: able T . Epidemiologic . Epidemiologic 19,20 , without elucidating , without elucidating . The PROWESS study . The PROWESS study 21-25 . The recently introduced introduced . The recently 27 7,11,19 . It is therefore unlikely that this complex disease that this complex disease unlikely . It is therefore 5,26 Sepsis is caused by an overwhelming and very complex host response to infection. infection. to response host complex very and overwhelming an by caused is Sepsis The variations in data definition and modes of data recording investigated in our our investigated in recording definition and modes of data The variations in data required the presence of three or more SIRS criteria, whereas both the 1991 consensus consensus 1991 the both whereas criteria, SIRS more or three of presence the required only two. The Kybersept study used and the OPTIMIST study required conference the by and algorithms for inclusion than proposed values for SIRS cutoff different the studies used varying definitions of (duration Furthermore, consensus conference. in inclusion criteria might explain the difference These differences failure. of) organ and might also partly explain the groups in 28-day mortality rates in the placebo studied interventions in these trials of the in efficacy differences the exact definitions that were used to establish these diagnoses. Interestingly, the the used to establish these diagnoses. Interestingly, the exact definitions that were completely sepsis, and associated mortality in our study are incidences of sepsis, severe might differences methodological indeed that suggesting reports, these with line in for the observed variation. be (at least partly) responsible is characterized by common and non-specific signs and The ensuing clinical syndrome the symptoms, the severity of which may be influenced by many factors, including and comorbidity, virulence of the pathogen, site of the infection, host susceptibility, temporal evolution of the condition could be fully captured by the relatively straightforward syndrome-based ACCP/ syndrome-based straightforward by the relatively could be fully captured it is plausible that sepsis can be better SCCM consensus definition of sepsis. Rather, relies on clinical symptoms alone, characterized and classified if its diagnosis no longer pathophysiologically based approach but also takes a more diagnosis of infection in the ICU. However, among patients with more strictly defined strictly defined with more among patients ICU. However, of infection in the diagnosis who patients among than patients postsurgical fewer significantly were there sepsis medical suggests either that of sepsis. This liberal definitions the more only met that many alternatively, patients or, disease than surgical severity of patients had greater we found that actually had SIRS, not sepsis. In addition, patients postoperative surgical first day significantly over the course of the decreased incidence of SIRS the apparent vital of the patient’s most likely caused by a gradual stabilization in the ICU. This was impact Although this did not significantly proceeded. in the ICU signs as resuscitation may this study, our in shock septic and sepsis, severe sepsis, of incidences observed the to the onset of relative situations: the timing of eligibility criteria, not be the case in all influence trial inclusion. will probably of treatment, infection and the start sepsis in recent reported effects study may partly explain some of the between-center sepsis, of sepsis, severe incidences variations in reported trials, as well as the large rates in various observational studies. For septic shock, and their associated mortality of anticoagulants efficacy clinical trials to date studying the largest instance, the three ways very different in sepsis, defined SIRS and sepsis in studies report large variations in the incidences of sepsis (ranging from 9 to 56%), 9 to 56%), (ranging from variations in the incidences of sepsis large studies report sepsis (20 to 52%) (8 to 27%), and mortality from sepsis severe PIRO model stages the severity of sepsis on the basis of predisposing factors and factors and PIRO model stages the severity of sepsis on the basis of predisposing of the underlying infection (I), the characteristics conditions (P), the nature premorbid
Classification of sepsis 2 28 Classification of sepsis 29 2
, . In . In 26,31 17,28 , polymerase , polymerase 29,30 Crit Care Med 2003;31(4):1250-6. Crit Care 1992;37(2):170-80. Respir Care incidence, morbidities and outcomes in incidence, morbidities and outcomes in Med ICU patients. Intensive Care surgical 1995;21(4):302-9. Marshall JC, Abraham E, Levy MM, Fink MP, SCCM/ESICM/ 2001 al. et D, Cook D, Angus International Sepsis Definitions ACCP/ATS/SIS Conference. East TD. Computers in the ICU: panacea or plague? LaRosa PF, GR, VincentBernard JL, Laterre al. et A, Lopez-Rodriguez JF, Dhainaut SP, recombinant human and safety of Efficacy sepsis. N Engl C for severe activated protein J Med 2001;344(10):699-709. Pillay SS, Carl BL, Eid A, Singer P, Warren the critically ill Novak I, et al. Caring for P, III in severe patient. High-dose antithrombin trial. JAMA sepsis: a randomized controlled 2001;286(15):1869-78. 5. 6. 7. 8. .
34 and finally, novel biomarkers may become available may become available novel biomarkers and finally, 26,32-34 Bone RC, Balk RA, Cerra FB, Dellinger RP, Bone RC, Balk RA, Cerra FB, Dellinger RP, et al. Definitions for Fein AM, Knaus WA, and guidelines for failure sepsis and organ the use of innovative therapies in sepsis. The ACCP/SCCM Consensus Conference Committee. American College of Chest Medicine. Physicians/Society of Critical Care Chest 1992;101(6):1644-55. Vincent JL, Le Gall JR, Sprung CL, Sakr Y, Reinhart K, Ranieri VM, et al. An evaluation of syndrome response systemic inflammatory In Acutely Ill signs in the Sepsis Occurrence Med Intensive Care Patients (SOAP) study. 2006;32(3):421-7. Vincent JL. Dear SIRS, I’m sorry to say that I 1997;25(2):372-4. Med like you. Crit Care don’t Tarara N, Li S, Rangel-Frausto D, Pittet D, Costigan M, Rempe L, et al. Systemic response syndrome, inflammatory shock: sepsis and septic sepsis, severe chain reaction techniques may increasingly be used for rapid pathogen detection for rapid pathogen be used may increasingly techniques chain reaction the near future, genetic profiling may determine host susceptibility profiling genetic the near future, biomarkers and proteomics and transcriptomics may potentially better classify the the better classify may potentially transcriptomics and biomarkers and proteomics to infection host response References 1. 2. 3. 4. Sources of support was performed within the framework of CTMM, the Center for Translational This research research MARS (grant 04I-201). MB has received project Molecular Medicine (www.ctmm.nl), (NWO Vici 918.76.611). of Scientific Research the Netherlands Organization funding from s Conclusion a clinical diagnosis of sepsis should not be used to make The ACCP/SCCM definitions When used for characterize this complex disease. appropriately because they fail to the use of the current therapy, for inclusion into clinical trials of sepsis selection of patients sensitivity to because of an extreme consensus criteria may become counterproductive in item definitions. and subtle variations minor changes in the mode of data recording unambiguous, concepts to developing new, towards should be directed efforts Future sepsis and septic shock. sepsis, severe better identify and classify patients with of the host response (R), and the extent of the resultant organ dysfunction (O) dysfunction (O) organ of the resultant (R), and the extent response of the host to detect organ failure at an early stage at an early stage failure to detect organ
N Engl Engl N J Am Coll Coll Am J Intensive Care Care Intensive 2008;12(6):R158. Crit Care Med 2002;28(2):108-21. Pinton P, Brun-Buisson C, Meshaka P, of the B. EPISEPSIS: a reappraisal Vallet sepsis epidemiology and outcome of severe units. Intensive Care intensive care in French Med 2004;30(4):580-8. M. Moss S, Eaton DM, Mannino GS, Martin The epidemiology of sepsis in the United 2000. N Engl J 1979 through States from Med 2003;348(16):1546-54. Sepsis: definition, Mackenzie I. Lever A, and diagnosis. Bmj epidemiology, 2007;335(7625):879-83. symposium 2010: Czura CJ. “Merinoff sepsis”-speaking with one voice. Mol Med 2011;17(1-2):2-3. Marshall JC. Biomarkers of sepsis. Curr Infect Dis Rep 2006;8(5):351-7. Guo SJ, FO, Chapman Khor CC, Vannberg AJ, et al. CISH and SH, Walley H, Wong diseases. infectious to susceptibility J Med 2010;362(22):2092-101. DJ, Downey T, Laramie JM, Meyer Chung TP, LH, Ding H, et al. Molecular diagnostics Tam bench. to bedside from sepsis: in 2006;203(5):585-98. Surg Becker K, Sachse S, Dessap Hinder F, Bloos F, to trial multicenter A al. et E, Straube AM, assessment for today’s critically ill patients. critically for today’s assessment Med 2006;34(5):1297-310. Crit Care What can we learn A. from Kidokoro Iba T, in anticoagulants using megatrials three the Shock 2004;22(6):508-12. sepsis? severe of Novel Opal S. Clinical Trials LaRosa SP, of Tale A Sepsis: Severe for Anticoagulants Advances in Sepsis 2004;4. Molecules. Three Lidicker WT, Angus DC, Linde-Zwirble J, Pinsky MR. J, Clermont G, Carcillo sepsis in the United Epidemiology of severe outcome, and States: analysis of incidence, Med Crit Care associated costs of care. 2001;29(7):1303-10. V, Sagredo A, Muriel-Bombin J, Blanco L, et al. Tamayo F, Gandia F, Taboada dysfunction and mortality Incidence, organ a Spanish multicentre sepsis: in severe study. H, Alberti C, Brun-Buisson C, Burchardi Martin C, Goodman S, Artigas A, et al. Epidemiology of sepsis and infection an international in ICU patients from study. cohort multicentre 24. 25. 26. 27. 28. 29. 30. 31. 19. 20. 21. 22. 23.
Resuscitation 2009;80(7):811-8. Arts D, de Keizer N, Scheffer GJ, de Jonge GJ, de Jonge Keizer N, Scheffer Arts D, de severity of of data collected for E. Quality Intensive National Dutch the in scores illness Intensive registry. Evaluation (NICE) Care 2002;28(5):656-9. Med Care Carlet JM, Masur H, Gerlach H, Dellinger RP, et al. Surviving Sepsis Cohen J, Calandra T, for management of Campaign guidelines and septic shock. Crit Care sepsis severe Med 2004;32(3):858-73. Opal S, Demeyer I,Abraham E, Reinhart K, and safety al. Efficacy Doig C, Rodriguez AL, et tissue factor pathway of tifacogin (recombinant sepsis: a randomized inhibitor) in severe trial. JAMA 2003;290(2):238-47. controlled Azoulay E, V, Annane D, Cariou A, Maxime Corticosteroid et al. JF, Timsit G, D’Honneur and intensive insulin therapy treatment for septic shock in adults: a randomized trial. JAMA 2010;303(4):341-8. controlled Peake SL, Bailey M, Bellomo R, Cameron al. Australasian A, Delaney A, et Cross PA, of sepsis evaluation (ARISE): A resuscitation cohort inception prospective, multi-centre, study. Vincent JL, de Mendonca A, Cantraine F, J, Suter PM, et al. Use of R, Takala Moreno to assess the incidence of score the SOFA in intensive care dysfunction/failure organ prospective of a multicenter, units: results “sepsis-related on group Working study. Society of of the European problems” Med Crit Care Medicine. Intensive Care 1998;26(11):1793-800. Bernard Marshall JC, Cook DJ, Christou NV, GR, Sprung CL, Sibbald WJ. Multiple organ descriptor of a reliable dysfunction score: Med a complex clinical outcome. Crit Care 1995;23(10):1638-52. Le Gall JR, Klar J, Lemeshow S, Saulnier Alberti C, Artigas A, et al. The Logistic F, new way to Dysfunction system. A Organ the intensive dysfunction in assess organ JAMA unit. ICU Scoring Group. care 1996;276(10):802-10. Abraham Marshall JC, MM, Fink MP, Levy E, Angus D, Cook D, et al. 2001 SCCM/ International Sepsis ESICM/ACCP/ATS/SIS Med Intensive Care Definitions Conference. 2003;29(4):530-8. ZimmermanMalilaDS, McNair AA, Kramer JE, Health FM. Acute Physiology and Chronic hospital mortality IV: Evaluation (APACHE) 9. 10. 11. 12. 13. 14. 15. 16. 17. 18.
Classification of sepsis 2 30 Classification of sepsis 31 2 Crit Care Med 2009;37(3):882-8. Crit Care Lancet 2005;365(9453):63-78. peripheral blood mononuclear cells in cells in blood mononuclear peripheral sepsis. Annane D, Bellissant E, Cavaillon JM. Septic JM. Septic Bellissant E, Cavaillon Annane D, shock. Vincent JL, Moreno R, Takala J, Willatts S, R, Takala Vincent JL, Moreno H, et al. The SOFA De Mendonca A, Bruining Assessment) Failure Organ (Sepsis-related dysfunction/failure. to describe organ score on Sepsis- Group On behalf of the Working Society of the European Related Problems Medicine. Intensive Care of Intensive Care Med 1996;22(7):707-10. 35. 36.
Crit Care Med 2009;37(7):2290-8. Crit Care Tang BM, McLean AS, Dawes IW, Huang Huang Dawes IW, BM, McLean AS, Tang of profiling SJ, Lin RC. Gene-expression 2007;7(3):210-7. of Biomarkers K. Reinhart JC, Marshall sepsis. compare blood culture with polymerase polymerase with blood culture compare human sepsis. in severe chain reaction Med 2010;36(2):241-7. Intensive Care McLean Eslick GD, Craig JC, BM, Tang for sepsis AS. Accuracy of procalcitonin patients: systematic diagnosis in critically ill meta-analysis. Lancet Infect Dis and review 34. 33. 32.
3 Interobserver agreement of CDC criteria for classifying infections in critically ill patients
Peter M.C. Klein Klouwenberg David S.Y. Ong Lieuwe D.J. Bos Friso M. de Beer Roosmarijn T.M. van Hooijdonk Mischa A. Huson Marleen Straat Lonneke A. van Vught Luuk Wieske Janneke Horn Marcus J. Schultz Tom van der Poll Marc J.M. Bonten Olaf L. Cremer
Critical Care Medicine Interobserver agreement for classifying sources of infection using CDC of infection using CDC for classifying sources agreement Interobserver Prospective cohort Prospective Data were collected as part of a cohort of 1,214 critically ill of a cohort of 1,214 critically ill collected as part Data were Setting and Patients: January 2011 and June two hospitals in the Netherlands between patients admitted to out of 554 patients who had assessed a random sample of 168 2011. Eight observers in the ICU. Each patient was assessed by experienced at least one infectious episode of infection the source independently scored two randomly selected observers who (rated as none, possible, system or site), the plausibility of infection organ (by affected based or definite), and the most likely causative pathogen. Assessments were probable, evidence. of all available clinical, radiological and microbiological on a post hoc review was classified as partial (i.e., of infection for source The observed diagnostic agreement system or site) or complete (i.e., matching on specific diagnostic matching on organ and scale), terms), for plausibility as partial (two-point scale) or complete (four-point or exact pathogen match. Interobserver for causative pathogens as an approximate and as a kappa statistic. percentage as a concordant was expressed agreement observed. and Main Results: A total of 206 infectious episodes were Measurements for of infection was 89% (183/206) and 69% (142/206) the source regarding Agreement in a kappa of 0.85 This resulted respectively. a partial and complete diagnostic match, diagnostic categories, 63-91% within major varied from (95% CI 0.79-0.90). Agreement with the lowest concordance within specific diagnostic subgroups, 35-97% and from In the 142 episodes for which pneumonia. observed in cases of ventilator-associated agreement of infection was obtained, the interobserver a complete match on source 65% on a 2- and 4-point scale, respectively. for plausibility of infection was 83% and exact and approximate an for 70% and 78% was agreement pathogen, causative For pathogen match, respectively. Conclusions: diagnosis aspects of the on all full concordance criteria was excellent overall. However, for some types of infection, in particular for between independent observers was rare pneumonia. ventilator-associated Abstract observational in is important of infection source of the classification : Correct Objectives and Prevention Control studies of sepsis. Centers for Disease and interventional of these used for this purpose, but the robustness most commonly (CDC) criteria are in a mixed ICU We hypothesized that ill patients is not known. definitions in critically and would population the performance reduced, these criteria would be generally of subgroups. vary amongst diagnostic Design:
Classifying infections in the ICU 3 34 Classifying infections in the ICU 35 3 . Updated . Updated 7 . 6 . Yet, correct identification of the correct . Yet, 3 . This may cause large differences in differences . This may cause large 2 . Although several studies have tested the robustness of CDC criteria of CDC criteria the robustness . Although several studies have tested 8-13 . The clinical diagnosis of sepsis is based on the presence of a systemic of sepsis is based on the presence . The clinical diagnosis 1 . The quality of experimental and observational studies in sepsis also greatly sepsis also greatly and observational studies in . The quality of experimental 4,5 Centers for Disease Control and Prevention (CDC) criteria were developed in in developed were criteria (CDC) Prevention and Control Disease for Centers inflammatory response syndrome (SIRS) and the presumption of infection. SIRS criteria and the presumption (SIRS) response syndrome inflammatory to sensitive extremely both non-specific for infection and unfortunately by themselves are and data recording minor variations in measurement versions of these criteria are nowadays more commonly used to classify infections in to classify infections in commonly used nowadays more versions of these criteria are patients, ill critically include to extended been has use their of scope the and general, for too complicated, subjective, and imprecise despite suggestions that they are practical use This study was performed the framework of the Molecular Diagnosis and Risk within observational cohort study a prospective Stratification of Sepsis (MARS) project, and aimed at the development of new molecular techniques for early prognostication of the pathogen detection in patients with sepsis. The Medical Ethics Committees for an opt-out consent method. Participants participating study centers gave approval with at ICU admission provided the study in writing by a brochure notified of were that could be completed by the patient or by his or her legal attached an opt-out card of unwillingness to participate. in case representative s and Methods Material uction Introd the body of immunological response and complex by an overwhelming Sepsis is caused to infection the apparent incidence of sepsis across settings. In addition, there are several difficulties difficulties several are there addition, In settings. across sepsis of incidence apparent the primary cause of to the establishment of infection as the that relate in diagnosing sepsis of microbiological and the interpretation source, of its presumed disease, the localization micro-organism to identify its likely causative results by focusing on a single particular infectious disease, there have been no studies that have been no studies that disease, there by focusing on a single particular infectious of CDC criteria to diagnose infections in have systematically evaluated the usability rates postulated that, in the latter case, increased general in critically ill patients. We will be that there also anticipated can be expected. We of diagnostic disagreement infection, such as ventilator within specific subtypes of agreement diagnostic reduced test these hypotheses, we performed a cohort study To associated pneumonia (VAP). for classifying infections of various subtypes agreement to determine the interobserver unit (ICU) population. in a mixed intensive care depends on the correctness of the diagnosis. Indeed, some have suggested that depends on the correctness of patients due to diagnostic misclassification hassuboptimal inclusion or stratification sepsis trials of so many recent failure been a major factor contributing to the infectious etiology is of utmost importance to guide treatment decisions in patients with decisions of utmost importance to guide treatment infectious etiology is sepsis 1988 and originally intended for the surveillance of nosocomial infections 1988 and originally intended for the surveillance that were translated and adapted to translated and adapted to that were 7-9 and infection (defined by the systemic use of antibiotics) by the systemic use of antibiotics) and infection (defined 14 We observed 1,214 consecutive critically ill patients with an expected length of stay of stay expected length ill patients with an consecutive critically observed 1,214 We We used an investigator manual providing practical diagnostic criteria for all all for criteria diagnostic practical providing manual investigator an used We the Dutch situation (see http://journals.lww.com/ccmjournal for a detailed description for a detailed description the Dutch situation (see http://journals.lww.com/ccmjournal was of all diagnostic criteria). Allocation of subjects into major diagnostic subgroups allocation based on the diagnoses of the two assessors. In case of disagreement, observer involved in data collection for was based on the diagnosis made by a third did not the primary MARS cohort. Obviously this served to classify infections and was classified rates. The observed diagnostic agreement change the (dis)agreement system or site) or complete (i.e., matching on as partial (i.e., matching on organ specific diagnostic terms). of infection was classified both on a 4-point The plausibility Data definitions classified on the basis of CDC and International Sepsis Infection diagnoses were definitions Forum (ISF) Consensus Conference of more than 24 hours, who were admitted to the mixed ICUs of two tertiary hospitals tertiary hospitals mixed ICUs of two admitted to the who were than 24 hours, of more patients daily for screened 2011. We January and June between in the Netherlands of SIRS criteria the presence and found 554 subjects who had experienced at least one infectious episode while infectious episode while who had experienced at least one and found 554 subjects the precluded this descriptive study, aimed to performin the ICU. Since we a purely that the most frequently we anticipated use of a formal power calculation. However, infections and blood stream in the ICU would be: pneumonia, occurring infections between the rates of observer agreement to compare In order abdominal infections. would be per group infection, we anticipated that 30 patients these main types of we would need we estimated that research previous for our purpose. From sufficient therefore We numbers. these obtain to patients 150 least at of size sample total a cases using a random number generator; took a random subsample of all infectious and selected the first 168 subjects. This we then sorted these in ascending order that enabled us to allocate the same number of particular number was chosen in a way was evaluated by two assessors, every patient Subsequently, cases to each observer. physicians. Each assessor a pool of eight research randomly selected from who were the system or site), organ of infection (by affected the source independently scored or definite), and the most possible, probable, plausibility of infection (rated as none, assessment was performedlikely causative pathogen. The post-hoc a median time at of 3 months after the ICU admission. items. This manual was tested for consistency and all observers were data relevant All assessors had the start of the study. before trained in its use on several occasions discussed and had at least vignettes were attended meetings in which clinical case based on a post hoc Assessments were study. six months of work experience in this data. The feasibility of available clinical, radiological and microbiological of all review in a pilot study with patients who were had been determined previously this approach not included in the final analysis.
Classifying infections in the ICU 3 36 Classifying infections in the ICU 37 3 . Data . Data 16 for ordinal for ordinal 0.65, 95% CI 0.65, 95% CI κ ) for categorical categorical for ) agreement as very good (0.81– agreement . We categorized κ categorized . We 15 0.65, 95% CI 0.51-0.77) and 65% (weighted 0.51-0.77) and 65% 0.65, 95% CI A total of 206 possible infectious episodes were observed in 168 patients. The most were A total of 206 possible infectious episodes to plausibility of infection and with regard 2b and 2c show the agreement Figures shows clinical characteristics for the 168 subjects in the random subsample 168 subjects in the random subsample able 1 shows clinical characteristics for the 0.55-0.73) on a 2- and 4-point scale, respectively. Agreement ranged from 100% for 100% for ranged from Agreement 0.55-0.73) on a 2- and 4-point scale, respectively. on the 4-point scale. For causative meningitis on the 2-point scale to 50% for VAP 70 to 78% for a species and genus level ranged from pathogens, overall agreement were analyzed using SAS 9.2 and SPSS 17. analyzed were 1.00), good (0.61–0.80), moderate (0.41–0.60), fair (0.21–0.40) or poor (< 0.20) 1.00), good (0.61–0.80), moderate (0.41–0.60), s Result T ill at severely more Patients with infection were of patients with infection in the ICU. general the than of stay length longer and rate fatality case higher a had admission, between infected patients and differences no relevant were ICU population, but there for analysis. the random subsample that was selected tract, bloodstream, the lower respiratory of infection were observed sources frequently interobserver of rate overall The 1). (Figure system nervous central and abdomen, for of infection was 89% (183/206) and 69% (142/206) the source regarding agreement ). The κ was 0.85 (95% 2a (Figure respectively a partial and complete diagnostic match, of infection, and 0.65 on main source CI 0.79-0.90) for the partial diagnostic classification The observed (partial) diagnostic agreement (95% CI 0.59-0.73) for full classification. 76% for bloodstream categories of infection, ranging from varied significantly within broad (full)infections to 91% for pulmonary and central nervous system infections. Specific to 79% for secondary peritonitis. 35% for VAP ranged from diagnostic agreement on source full diagnostic agreement its causative pathogen for the 142 episodes where of infection had been obtained. For plausibility of infection, the overall interobserver (κ was 83% agreement We calculated rates of diagnostic agreement for source of infection, plausibility of of infection, plausibility of for source agreement calculated rates of diagnostic We (κ kappa Cohen’s calculated We organism. causative and infection Statistical analysis of infection and causative pathogen), and a weighted κ variables (source variables (plausibility of infection) scale (none, possible, probable, definite) and with a 2-point scale (no infection or or scale (no infection and with a 2-point definite) possible, probable, scale (none, analyses the Similarly, possible and probable). point between with a cut-off infection, performed on an (exact match) and the species levels on were organisms of causative into 11 distinct categories been grouped pathogens had match where approximate Gram-negative Gram-negative cocci, Gram-positive rods, (Gram-positive cocci, unknown, or no and fungi, viruses, other, atypical bacteria, yeasts anaerobes, rods, for plausibility of infection and agreement calculated interobserver pathogen). We infection diagnosis. on the only if both observers had first agreed causative pathogen 13 (8) 28 (17) 44 (27) 34 (20) 29 (17) 61 (36) 31 (19) 48 (29) 101 (60) 107 (63) (n = 168) 62 (49–72) 78 (63-100) 5.8 (2.0 – 12.9) Random subsample 66 (17) 77 (20) 66 (17) 73 (19) 42 (11) 73 (19) 228 (59) 108 (28) 135 (35) 271 (70) Infection (n = 386) 0.67, 95% CI 0.59-0.76). Within 62 (51-71) 80 (62-103) 5.0 (1.9 - 9.7) 47 (7) 45 (7) 68 (10) 396 (60) 134 (20) 274 (42) 205 (31) 264 (40) 265 (40) 131 (20) (n = 660) 58 (39-82) 60 (48-72) No infection 1.3 (0.8 - 3.1) 0.71, 95% CI 0.62-0.80) and (κ Hospital ward Hospital ward room Emergency Operating Room Other Medical Medical Elective surgery surgery Emergency Although overall the observed interobserver agreement was high, we found was high, we found agreement Although overall the observed interobserver Patient characteristics at admission able 1: Patient characteristics Admission source, n (%) n (%) Admission source, (IQR) median IV score, APACHE (IQR) median ICU length of stay, n (%) ICU observed case fatality, unit evaluation, ICU – intensive care health acute physiology and chronic – APACHE Abbreviations: ICU readmission, n (%) ICU readmission, Age, median (IQR) n (%) Male gender, Admission type, n (%) T specific subtypes of infection, this ranged from 88% for CNS and bloodstream infections infections 88% for CNS and bloodstream from specific subtypes of infection, this ranged pneumonia on a species level. acquired on a genus level to 56% for community study assessing the suitability of CDC criteria of a systematic the results present We examining patients, using multiple reviewers for classifying infections in critically ill for the source agreement determined the interobserver We identical patient records. our knowledge, of infection, the plausibility of infection and its causative pathogens. To appraisal of diagnostic classification study to attempt a comprehensive first this is the criteria for of infections in the ICU using commonly used definitions and including plausibility and causative pathogen. for some subtypes of infection, in levels of concordance substantially reduced require our findings suggest that infection diagnoses that In particular, particular VAP. and radiological information of clinical, microbiological the combined interpretation to infection compared agreement associated with a lower interobserver were of less elaborate criteria and do not involve interpretation diagnoses that require match, respectively (κ match, respectively Discussion
Classifying infections in the ICU 3 38 Classifying infections in the ICU 39 3 radiological information. The diagnosis of VAP, for instance, was associated with a for instance, was associated with a radiological information. of VAP, The diagnosis and extensive the by explained be may which agreement, interrater low relatively be true complicated criteria needed to accurately establish the diagnosis. This can performed not routinely lavages are bronchoalveolar in particular in ICUs where Descriptive diagram showing the diagnostic classification of patients treated for infection for infection classification of patients treated 1: Descriptive diagram showing the diagnostic Figure is a duplicate of the number of infections (n = 206) in the ICU. The total number of observations into the main infection independently rated by two observers. Allocation since patients were on the in case of disagreement, two assessors, or, classes was based on the diagnoses of the on was defined as concordance Partial diagnostic agreement observer. diagnosis made by a third on was defined as concordance system (e.g. lungs, blood). Full diagnostic agreement the organ pneumonia). pneumonia or hospital acquired specific diagnostic terms (e.g. community acquired for that particular between the observers was agreement classification indicates that there Correct case the alternative in which was disagreement, classification that there diagnosis; incorrect into one of Infectious episodes (n = 35) that could not be classified provided. diagnostic terms are mediastinitis (n = were: not shown. Diagnoses in this group the four main diagnostic categories are tract (n = 5), oral infections (n = 4), upper-respiratory 8), skin infections (n = 6), urinary tract infections tract infections (n = 3), post-operative wound infections (n = 2), infections (n = 3), gastro-intestinal infection, BSI – bloodstream viral infections (n = 1), and unknown infections (n = 3). Abbreviations: infection, HAP – bloodstream pneumonia, CRBSI – catheter-related CAP – community acquired tract, PD – peritoneal pneumonia, IA – intra-abdominal, LRT – lower respiratory hospital acquired – ventilator associated pneumonia. dialysis, POWI – post-operative wound infection, VAP Selected sites of infection of sites Selected Selected sites of infection of sites Selected Selected sites of infection of sites Selected Agreement on plausibility on Agreement Agreement onsource of infection Agreement oncausative pathogens Organ system Organ system Organ Organ system Organ LRT Abdomen Bloodstream CNS Other HAP CAP VAP SP CRBSI LRT Abdomen Bloodstream CNS Other HAP CAP VAP SP CRBSI LRT Abdomen Bloodstream CNS Other HAP CAP VAP SP CRBSI 0 0 0
90 80 70 60 50 40 30 20 10 90 80 70 60 50 40 30 20 10 90 80 70 60 50 40 30 20 10
100 100 100 Agreement (%) Agreement Agreement (%) Agreement Agreement (%) Agreement C B A Interobserver agreement rates for source of infection (a), plausibility of infection of infection (a), plausibility of infection rates for source agreement 2A, 2B, 2C: Interobserver Figure diagnostic (b) and causative pathogen (c) within various diagnostic categories. The observed terms; diagnostic specific on matching (i.e., complete as was classified shaded light agreement system; dark shaded bars); for plausibility of infection bars) and partial (i.e., matching on organ definite; light shaded bars) and a 2-point scale (no on a 4-point scale (none, possible, probable, dark shaded bars); and for causative point between possible and probable; or yes with a cut-off (genus level; dark shaded light shaded bars) and approximate as exact (species levels; organisms CNS – central nervous system, bars). Whiskers indicate 95% confidence intervals. Abbreviations: tract. infection, LRT – lower respiratory bloodstream CRBSI – catheter-related
Classifying infections in the ICU 3 40 Classifying infections in the ICU 41 3 = 0.40) for VAP = 0.40) for VAP . Agreements varied from fair (κ varied from . Agreements 10-12,17-19 . This update would most likely not have changed the rate . This update would most likely not have changed the rate 20 = 0.42) for catheter-related bloodstream infections to substantial to substantial infections bloodstream = 0.42) for catheter-related This study has several limitations. First, we did not include reference standards, standards, we did not include reference This study has several limitations. First, The results of our study are in line with previous studies that have reported rates rates studies that have reported previous in line with of our study are The results = 0.69) for surgical site infections. These studies focused on one particular infection infection These studies focused on one particular site infections. = 0.69) for surgical κ and moderate (κ ( arise when all uncertainty and complexity that may the diagnostic and ignored in our found however no indication We considered. of infection are possible sources ICU population were life’ mixed in a ‘real data that the rates of diagnostic agreement single any on focused have that studies previous in observed those from different any occurring infections that of frequently the concordance particular disease. Importantly, pneumonia or abdominal studied, such as community-acquired not previously were study. infections, was good to excellent in our panel, which made the classification difficult for example the opinion of an expert less is, however, standard upon the diagnosis. A reference when observers disagreed to diagnostic compared agreement important in studies determining interobserver tested. Second, the generalizability are studies in which diagnostic tests or strategies performed routinely were that processes diagnostic the on depends study this of other hospitals. Interobserver from different in the two ICUs, which could be very clinical information is available due more might be higher in settings where agreement limited diagnostic with in settings whereas diagnostic approach, aggressive more a For instance, in the two study hospitals might be lower. agreement tools interobserver performed suspected patients in routinely not determinationare titers antibody of criteria excluded these therefore We of having an infection (e.g. gastroenteritis). and regularly updated are the CDC criteria CDC definitions. Third, the original from studies. our data for future that can be drawn from this may influence the inferences has been a minor update of the 2008 definitions that Indeed, in January 2012 there we used for this study when VAP is suspected, making a definite diagnosis almost impossible to obtain. to obtain. almost impossible diagnosis making a definite is suspected, when VAP determination for the of agreements interobserver we found better In contrast, in normallyinfections occurring in sterile compartments micro-organisms causative respiratory tract occurring in the to infections fluid) compared (blood, cerebrospinal between colonization to differentiate found it difficult assessors or abdomen. Possibly, significant most the select to or pneumonia, of suspected patients in infection true and results. abdominal culture with poly-microbial pathogens in patients of diagnostic concordance of agreement in sources of infection as this particular part was not modified, but it of infection as this particular part was not modified, but it in sources of agreement in the plausibility of infection changed the rate of agreement might have marginally In addition, we infections. blood stream such as catheter related for certain sources subjective too be to considered we what excluding by definitions the adapted slightly of a ‘physician diagnosis’ in case of ‘Other criteria (for instance the prerequisite 2012 the from removed was criterion this Interestingly, tract’). urinary the of infections for other update of the CDC diagnostic guidelines for this particular infection, but not Jun Jun Am J Infect Control. Control. Infect J Am Intensive Intensive Sepsis Definitions Conference. Med. Apr 2003;29(4):530-538. Care Marshall JC. Biomarkers of sepsis. Curr Infect Dis Rep. Sep 2006;8(5):351-357. GR, et al. New Cohen J, Guyatt G, Bernard strategies for clinical trials in patients with Apr Med. sepsis and septic shock. Crit Care 2001;29(4):880-886. Garner JS, Jarvis WR, Emori TG, Horan TC, Hughes JM. CDC definitions for nosocomial 1988. infections, 1988;16(3):128-140. Cohen J. The international sepsis Calandra T, definitions on forum consensus conference 5. 6. 7. 8. , we did not find indications that this had occurred in in that this had occurred , we did not find indications 10-12,21,22
Dec 19-26 2002;420(6917):885-891. May 2012;38(5):811-819. Nov 2004;32(11 Suppl):S466-494. Cohen J. The immunopathogenesis of sepsis. Cohen J. The immunopathogenesis of sepsis. Nature. DS, Bonten PM, Ong Klein Klouwenberg of sepsis, OL. Classification MJ, Cremer impact the shock: septic and sepsis severe and of minor variations in data capture definition of SIRS criteria. Intensive Care Med. A, Cohen J, Brun-Buisson C, Torres infection in J. Diagnosis of Jorgensen Crit Care sepsis: an evidence-based review. Med. JC, et al. 2001 Marshall Levy MM, Fink MP, International SCCM/ESICM/ACCP/ATS/SIS References 1. 2. 3. 4. Sources of support Molecular Medicine (http:// for Translational This work was supported by the Center funding from research MARS (grant 04I-201). MB has received project www.ctmm.nl), (NWO Vici 918.76.611). of Scientific Research the Netherlands Organization s Conclusion for classifying agreement we observed a rate of interobserver Using CDC criteria, full concordance in the ICU that was excellent overall. However, of infection sources independent observers was uncommon, in on all aspects of the diagnosis between infections. tract respiratory lower ventilator-associated for treated patients particular in observational studies and clinical trials of sepsis A post-hoc stratification of patients in CDC criteria for the majority of infections, but feasible with the current is therefore the combined require to establish (and thus difficult more for diagnoses that are information), and radiological the current of clinical, microbiological interpretation misclassifications. to prevent insufficient criteria are infections although it is stated that “physician diagnosis alone is not an acceptable an acceptable not is alone diagnosis that “physician it is stated although infections found that small several studies of infection”. Although any specific type criterion for in the estimated cause major variation definitions could to the CDC changes made incidences of infections our study. Fourth, we could not adequately determine the interobserver agreement agreement determine adequately not Fourth, we could interobserver the our study. found small sample size. We due to the relatively occurring infections for infrequently for rare agreement in our data that the rate of diagnostic however no evidence common conditions. that of more from different infections was any
Classifying infections in the ICU 3 42 Classifying infections in the ICU 43 3 Sep Mar 1977;33(1):159-174. May 1998;19(5):308-316. Feb 2004;30(2):217-224. Landis JR, Koch GG. The measurement of of Koch GG. The measurement Landis JR, data. categorical for agreement observer Biometrics. al. JR, Culver DH, et Edwards Emori TG, nosocomial infections Accuracy of reporting patients to the in intensive-care-unit Surveillance National Nosocomial Infections Hosp Infect Control System: a pilot study. Epidemiol. CS, Wetterslev Henriksen NA, Meyhoff Rasmussen LS, J, Wille-Jorgensen P, of surgical LN. Clinical relevance Jorgensen by the criteria of the site infection as defined and Prevention. Centers for Disease Control J Hosp Infect. Jul 2010;75(3):173-177. Howell J, et al. Agreement T, Mayer J, Greene Infections Among in Classifying Bloodstream Multiple Reviewers Conducting Surveillance. Clin Infect Dis. May 31 2012. Definition Surveillance CDC. CDC/NHSN Infection and of Healthcare-Associated Types of Infections in the Criteria for Specific Setting. 2012; http://www.cdc. Acute Care gov/nhsn/pdfs/pscmanual/17pscnosinfdef_ . Accessed November 30, 2012. current.pdf K, Moody B, et al. Alternative Hawkins Minei JP, case definitions of ventilator-associated patients in pneumonia identify different unit. Shock. intensive care a surgical 2000;14(3):331-336; discussion 336-337. Jacobs Nieuwenhoven CA, Schurink CA, Van for JA, et al. Clinical pulmonary infection score accuracy pneumonia: ventilator-associated Intensive Care variability. and inter-observer Med. 16. 17. 18. 19. 20. 21. 22. Psychol Bull. Bull. Psychol Jun 1992;101(6):1644-1655. Cohen J. Weighted kappa: nominal scale kappa: nominal scale Cohen J. Weighted for scaled with provision agreement credit. partial or disagreement Crit unit. Crit in the intensive care of infection Med. Jul 2005;33(7):1538-1548. Care CDC/ Andrus M, Dudeck MA. Horan TC, health care- definition of NHSN surveillance criteria for specific associated infection and setting. acute care types of infections in the Jun 2008;36(5):309-332. Am J Infect Control. Gibbons C, Reeves BC, et al. Wilson AP, wound infection as a performance Surgical of common definitions indicator: agreement 4773 patients. Bmj. of wound infection in Sep 25 2004;329(7468):720. Fourie B, Ashton V, Allami MK, Jamil W, Superficial incisional infection in PJ. Gregg of the lower limb. Interobserver arthroplasty J diagnostic criteria. of the current reliability Sep 2005;87(9):1267-1271. Br. Bone Joint Surg variability in Klompas M. Interobserver pneumonia surveillance. ventilator-associated Apr 2010;38(3):237-239. Am J Infect Control. Ashby E, Haddad FS, O’Donnell E, Wilson AP. site infection be measured How will surgical for all”? J Bone “high quality care to ensure Sep 2010;92(9):1294-1299. Br. Joint Surg Bone RC, Balk RA, Cerra FB, et al. Definitions and guidelines failure for sepsis and organ for the use of innovative therapies in sepsis. The ACCP/SCCM Consensus Conference Committee. American College of Chest Medicine. Physicians/Society of Critical Care Chest. Oct 1968;70(4):213-220. 15. 9. 10. 11. 12. 13. 14.
4 Presence of infection in patients with presumed sepsis at the time of intensive care unit admission
Peter M.C. Klein Klouwenberg Olaf L. Cremer Lonneke A. van Vught David S.Y. Ong Jos F. Frencken Marcus J. Schultz Marc J. Bonten Tom van der Poll A clinical suspicion of infection is mandatory for diagnosing sepsis for diagnosing sepsis is mandatory suspicion of infection A clinical We studied a cohort of critically ill patients admitted with clinically suspected ill patients admitted with clinically suspected studied a cohort of critically We Abstract Background: of Yet, the accuracy response syndrome. systemic inflammatory in patients with a as being Unit (ICU) to the Intensive Care ill patients presenting categorizing critically the likelihood of infection in assessed therefore unknown. We infected or not is the admission to the ICU, and quantified for sepsis upon treated patients who were plausibility of infection and mortality. association between Methods: 2011 and December ICUs in the Netherlands between January sepsis to two tertiary or probable possible, none, as categorized was infection of likelihood The 2013. risks survival We used multivariable competing assessment. definite by post-hoc analyses to determine mortality. the association of the plausibility of infection with for sepsis, 13% had a post-hoc infection Results: Among 2579 patients treated 30% of only “possible”. These percentages likelihood of “none”, and an additional crude primary suspected sites of infection. In similar for different largely were length of stay and associated with increased analyses, the likelihood of infection was higher a had infection unlikely an with patients analysis, multivariable In complications. to patients with a definite infection (subdistribution hazard mortality rate compared ratio 1.23; 95% confidence interval 1.03-1.49). analysis to show that the clinical Conclusions: This study is the first prospective of presence the with poorly corresponds admission ICU upon sepsis of diagnosis likelihood of infection does not adversely infection on post-hoc assessment. A higher influence outcome in this population.
Infection diagnosis in presumed sepsis 4 46 Infection diagnosis in presumed sepsis 47 4 . 2 . For this study, we included we included study, this For . 3 . The Institutional Review Board approved an approved . The Institutional Review Board 3,4 . The panel defined sepsis as SIRS caused by clinically sepsis as SIRS caused by clinically . The panel defined 1 Although the clinical suspicion of infection is a crucial factor in making a sepsis in making a sepsis suspicion of infection is a crucial factor Although the clinical between the prospective this hypothesis we assessed the concordance address To Data collection and definitions all data from collected relevant Dedicated and trained observers prospectively described, i.e., as previously recorded patients. An infectious event was prospectively therapeutic reasons for started was antibiotic when an Study design and population MARS (Molecular Diagnosis and Risk StratificationThis cohort study was incorporated in the centers in the Netherlands in the mixed ICUs of two tertiary referral of Sepsis) project identifier NCT01905033) (ClinicalTrials.gov Material and methods Material uction Introd infection a severe to response the body’s that arises when a syndrome Sepsis is clinical a internationalan 1992 In tissues. own its proposed panel consensus injures response inflammatory making use of the concept of a “systemic definition for sepsis, heart rate, respiration in body temperature, (SIRS), involving alterations syndrome” counts rate and leukocyte suspected infection. It further introduced the terms “severe sepsis” to describe cases cases sepsis” to describe the terms “severe It further introduced suspected infection. shock” as sepsis dysfunction and “septic by acute organ when sepsis is complicated resuscitation. These definitions, to fluid hypotension refractory is complicated by in have been used as inclusion criteria to as the “Bone criteria”, generally referred unchanged largely trials, and until today have remained many clinical sepsis infections that started before ICU admission until 48 hours after ICU admission. We ICU admission until 48 hours after ICU admission. We infections that started before also performed in which we excluded infections starting more a sensitivity analysis opt-out consent method (IRB number 10-056C). For the current study, we analyzed all first study, opt-out consent method (IRB number 10-056C). For the current Januarybetween admitted were who diagnosis sepsis a with patients adult of admissions 2011 and December 2013, with an expected stay of > 24 hours. diagnosis, little is known about the accuracy of this diagnosis in the context of critically of this diagnosis in the context of critically diagnosis, little is known about the accuracy signs and symptoms Unit (ICU) with to the Intensive Care ill patients who present it likely that in the clinical practice of an considered We of a “sepsis syndrome”. on strict diagnostic criteria for infection ICU the diagnosis of sepsis is not based on the ICU might be of sepsis the occurrence and that as a consequence thereof for estimating incidence is helpful Quantification of the discordance overestimated. of antibiotic use. possible reduction rates in epidemiological studies and the of diagnosis post-hoc the physicians and bedside by made diagnosis sepsis clinical addition, we assessed the using strict criteria. In infection made by clinical researchers with outcome. association of the likelihood of infection . shows the main able 1 shows the main 7 . A competing risks analysis . A competing risks analysis 8 . The plausibility of infection was included . The plausibility of infection was included 9 . All patients with sepsis were managed according to locally adapted to locally adapted managed according were . All patients with sepsis 3,5,6 We assessed the effect of infection plausibility on ICU mortality using a competing using a competing of infection plausibility on ICU mortality the effect assessed We Demographics 7347 Over the 3-year observation period we studied 6944 patients during 912 ICU hospitalizations, contributing a total of 8259 ICU episodes, of which (37%) received analysis. At admission, 2738 patients from excluded were readmissions for a clinical suspicion of infection, of whom 2579 (94%) had therapeutic antimicrobials thus diagnosed with sepsis. T at least two SIRS criteria and were s Result to the ICU with a clinical diagnosis of sepsis. characteristics of patients presenting (community- pulmonary of sepsis were suspected primary sources The most frequent Statistical analyses (IQR) or numbers and ranges as median and interquartile presented are All results analyzed using non-parametric data were Continuous as appropriate. percentages, test. Chi-squared analyzed using the and categorical data were test a Kruskal-Wallis test for equality of cumulative the Gray’s and The Cochran–Armitage test for trend used. incidence functions were risks survival analysis to account for informative censoring than 48 hours before admission, because we anticipated that the likelihood of these of these that the likelihood we anticipated admission, because before than 48 hours plausibility of admission. The before the ones starting longer from differed infection was determined definite) based on all post-hoc, (none, possible, probable, infection to the according and radiological evidence and available clinical, microbiological, (CDC) and the International Sepsis Forum Prevention and Control Centers for Disease (ISF) criteria as a dichotomous variable (none/possible vs. probable/definite). We adjusted for We adjusted for vs. probable/definite). as a dichotomous variable (none/possible chosen a priori based on their expected associations with infection confounders that were on clinical expertise. and consideration of the literature and mortality after careful malignancy, disease, immunocompromise, cardiovascular These included age, gender, sepsis recent surgery, renal insufficiency, insufficiency, diabetes mellitus, respiratory health evaluation (APACHE) site of infection, and acute physiology and chronic severity, to be statistically significant. All analyses were considered values <0.05 were P IV score. R version 3.10 (www.r-project.org). NC) and performed using SAS 9.2 (Cary, provides two measures of association: the cause-specific hazard ratio (CSHR), which ratio (CSHR), which of association: the cause-specific hazard measures two provides and death), ICU discharge on outcome (both of infection effects estimates the direct separate all of measure summary a is which (SHR), ratio hazard subdistribution the and and can be used to calculate the cumulative incidence of the cause-specific hazards (i.e., death in this study) outcome of interest protocols based on the Surviving Sepsis Campaign guidelines based on the Surviving protocols
Infection diagnosis in presumed sepsis 4 48 Infection diagnosis in presumed sepsis 49 4 ). An additional 30% had an infection likelihood of possible, had an infection likelihood of possible, ). An additional 30% able 1 T shows the plausibility of infection after post-hoc analysis for the five post-hoc analysis for the five shows the plausibility of infection after
shows the plausibility of infection after post-hoc analysis for the whole whole analysis for the infection after post-hoc the plausibility of shows Plausibility of infection stratified by clinical severity upon presentation in patients with in patients with Plausibility of infection stratified by clinical severity upon presentation Figure 2 Figure ccuracy of infection diagnosis diagnosis ccuracy of infection Figure 1: Figure sepsis presumed patients with lung infection qualified as definite, which was significantly lower than in the definite, which was significantly lower than in patients with lung infection qualified as analysis, whereas this proportion was 62% for patients presenting with organ failure and and failure with organ was 62% for patients presenting this proportion analysis, whereas p (Cochrane-Armitage test for linear trend 66% for patients with shock, respectively <0.001). infections of definite and probable of infection. The proportion sources most prevalent 16% of of infection. Of note, only sources similar in patients with different was largely hours before admission resulted in a similar distribution (n = 2117): 15%, 32%, 25%, 25%, 32%, 15%, 2117): = (n distribution similar a in resulted admission before hours respectively. and definite infections, possible, probable, classed as none, and 28% were 1 Figure infection diagnosis The plausibility of a correct cohort, and stratified by sepsis severity. Only 48% sepsis severity. with greater to the post-hoc adjudication increased according or definite infection on post hoc had probable failure of patients without acute organ A upon likelihood of “none” infection 13% had an for sepsis, treated all patients Of post-hoc analysis ( infection likelihood (25% probable a higher than half scored slightly more whereas within 48 diagnosed were Limiting the analysis to infections that and 33% definite). acquired and hospital acquired pneumonia, n = 1292), abdominal (peritonitis, n = n = 1292), abdominal (peritonitis, pneumonia, n = acquired and hospital acquired blood catheter-related and stream primary blood (endocarditis, stream 414), blood The (n = 118) infections. and skin or soft tissue tract (n = 162), n = 230), urinary stream, at other sites. 363 patients had infections remaining
<.001 198) (75, 118 158) (65, 94 157) (69, 100 167) (68, 101 171) (70, 104 Creatinin
<.001 270) (78, 170 234) (47, 125 181) (19, 86 102) (8, 36 229) (35, 114 protein C-reactive
0.88 20.5) (8.5, 14.5 20.1) (9.6, 14.2 19.0) (10.4, 14.6 18.8) (9.9, 13.5 19.8) (9.6, 14.2 count cell blood White
<.001 38.7) (37.1, 37.9 38.6) (37.1, 37.9 38.5) (37.0, 37.7 38.4) (36.9, 37.6 38.6) (37.0, 37.8 temperature Core
0.08 101) (60, 79 100) (61, 79 96) (58, 75 101) (58, 74 100) (66, 77 score IV APACHE
<.001 (31%) 265 (19%) 122 (24%) 186 (27%) 88 (26%) 661 admission Surgical
Admission characteristics Admission
0.99 (19%) 161 (19%) 119 (19%) 143 (19%) 63 (19%) 486 mellitus Diabetes
Immunocompromise Immunocompromise 0.002 (27%) 228 (27%) 173 (21%) 159 (20%) 68 (24%) 628
e
Malignancy Malignancy 0.02 (11%) 90 (11%) 68 (8%) 61 (6%) 20 (9%) 239
d
Renal insufficiency insufficiency Renal 0.02 (15%) 130 (11%) 70 (13%) 98 (9%) 31 (13%) 329
c
Respiratory insufficiency insufficiency Respiratory 0.05 (14%) 121 (19%) 120 (18%) 141 (15%) 50 (17%) 432
b
Cardiovascular disease disease Cardiovascular 0.07 (21%) 178 (22%) 142 (26%) 204 (24%) 81 (23%) 605
a
Charlson comorbidity index comorbidity Charlson 4.6 (0.0, 9.7) (0.0, 4.6 9.4) (0.0, 4.6 9.4) (0.0, 2.5 8.1) (0, 1.5 9.1) (0, 3.5 0.02
Co-morbidities
Body mass index > 30 > index mass Body 479 (19%) 479 166 (22%) 166 (16%) 53 0.004 (20%) 167 (15%) 93
Race, Caucasian Race, 2257 (88%) 2257 0.09 (88%) 739 (87%) 552 (89%) 688 (84%) 278
Gender, male Gender, 1540 (60%) 1540 0.04 (59%) 501 (58%) 365 (64%) 493 (55%) 181
Age, years years Age, 62 (49, 71) (49, 62 71) (49, 62 71) (51, 62 72) (48, 62 71) (49, 62 0.73
Demographics
Number 2579 (100%) 2579 843 (33%) 843 (25%) 633 (30%) 771 (13%) 332 n/a
ll A robable P ossible P one N p-value efinite D
ost hoc plausibility of infection of plausibility hoc ost P
Baseline characteristics of patients admitted with presumed sepsis presumed with admitted patients of characteristics Baseline able 1: 1: able T
Infection diagnosis in presumed sepsis 4 50
Infection diagnosis in presumed sepsis 51 4
Organ failure at admission is based on the Sequential Organ Failure Assessment scores. Assessment Failure Organ Sequential the on based is admission at failure Organ
documented humoral or cellular deficiency. deficiency. cellular or humoral documented
of > 75 mg/day for at least one week), current use of immunosuppressive drugs, current use of antineoplastic, drugs recent hematologic malignancy, or or malignancy, hematologic recent drugs antineoplastic, of use current drugs, immunosuppressive of use current week), one least at for mg/day 75 > of
Immunocompromise was defined as having acquired immune-deficiency syndrome, the use of corticosteroids in high doses (equivalent to prednisolone prednisolone to (equivalent doses high in corticosteroids of use the syndrome, immune-deficiency acquired having as defined was Immunocompromise
e
Malignancy included both metastatic and hematologic malignancies. malignancies. hematologic and metastatic both included Malignancy
d
Renal insufficiency was defined as chronic renal insufficiency (creatinine >177 µmol/L) or chronic dialysis. dialysis. chronic or µmol/L) >177 (creatinine insufficiency renal chronic as defined was insufficiency Renal
c
mechanical ventilation, oxygen use at home, or severe pulmonary hypertension). hypertension). pulmonary severe or home, at use oxygen ventilation, mechanical
Respiratory insufficiency was defined as chronic obstructive pulmonary disease or chronic respiratory insufficiency with functional disabilities (chronic (chronic disabilities functional with insufficiency respiratory chronic or disease pulmonary obstructive chronic as defined was insufficiency Respiratory
b
angioplasty or bypass for arterial insufficiency). insufficiency). arterial for bypass or angioplasty
congestive heart failure (ejection fraction < 30%) or peripheral vascular disease (claudicatio intermittens, patients with percutaneous transluminal transluminal percutaneous with patients intermittens, (claudicatio disease vascular peripheral or 30%) < fraction (ejection failure heart congestive
Cardiovascular disease was defined as cerebrovascular disease or chronic cardiovascular insufficiency (New York Heart Association class 4), chronic chronic 4), class Association Heart York (New insufficiency cardiovascular chronic or disease cerebrovascular as defined was disease Cardiovascular
a
squared test. squared
Data are median (IQR) or number (%). n/a= not applicable. The four infection plausibility classes were compared by use of the Kruskal-Wallis test or Chi- or test Kruskal-Wallis the of use by compared were classes plausibility infection four The applicable. not n/a= (%). number or (IQR) median are Data
0.037 (12%) 103 (9%) 54 (9%) 67 (12%) 39 (10%) 263 Dialysis
0.93 (78%) 655 (78%) 492 (79%) 608 (79%) 261 (78%) 2016 ventilation Mechanical
Treatment at admission at Treatment
<.001 11) (5.0, 8.0 9.0) (5.0, 7.0 9.0) (4.0, 7.0 10) (5.0, 7.0 10) (5.0, 7.0 Total
Coagulation Coagulation 0.0 (0.0, 1.5) (0.0, 0.0 1.0) (0.0, 0.0 1.0) (0.0, 0.0 1.0) (0.0, 0.0 1.0) (0.0, 0.0 <.001
Hepatic Hepatic 0.0 (0.0, 0.0) (0.0, 0.0 0.0) (0.0, 0.0 0.0) (0.0, 0.0 0.0) (0.0, 0.0 0.0) (0.0, 0.0 <.001
Renal Renal 1.0 (0.0, 3.0) (0.0, 1.0 1.0) (0.0, 0.0 1.0) (0.0, 0.0 1.0) (0.0, 0.0 2.0) (0.0, 0.0 <.001
Respiratory Respiratory 2.0 (2.0, 3.0) (2.0, 2.0 3.0) (2.0, 3.0 3.0) (2.0, 3.0 3.0) (2.0, 3.0 3.0) (2.0, 2.5 0.35
Cardiovascular Cardiovascular 3.0 (1.0, 4.0) (1.0, 3.0 4.0) (1.0, 3.0 4.0) (1.0, 3.0 4.0) (1.0, 3.0 4.0) (1.0, 3.0 <.001
Central nervous system nervous Central 0.0 (0.0, 1.0) (0.0, 0.0 1.0) (0.0, 0.0 1.0) (0.0, 0.0 1.0) (0.0, 0.0 1.0) (0.0, 0.0 0.22
Organ failure at admission at failure Organ
Septic shock Septic 776 (30%) 776 325 (39%) 325 (31%) 197 (22%) 167 (26%) 87
Severe sepsis Severe 727 (28%) 727 235 (28%) 235 (31%) 198 (29%) 224 (21%) 70
Sepsis 1076 (42%) 1076 175 (53%) 175 380 (49%) 380 238 (38%) 238 283 (34%) 283
Sepsis severity at admission at severity Sepsis <.001 shows the concurrent (non- shows the concurrent able S1 T Supplementary
shows various patient outcomes in the whole population, and stratified the whole population, and stratified shows various patient outcomes in Plausibility of infection in patients with presumed sepsis upon presentation for the most for the most sepsis upon presentation Plausibility of infection in patients with presumed displays the cumulative incidence functions of mortality for the four classes of utcomes utcomes analysis revealed that this reduction was caused by a direct effect on death (CSHR on death (CSHR effect was caused by a direct that this reduction analysis revealed stay of length ICU a longer on effect indirect by the not and 0.61-0.89), 95% CI 0.73; that in this crude survival analysis plausibility of infection was also not associated with with that in this crude survival analysis plausibility of infection was also not associated analysis, mortality (p = 0.73; crude SHR 1.05; 95% CI 0.88-1.25). In the multivariable a was associated with a higher plausibility of infection (probable/definite) however, means that lower mortality (SHR 0.81; 95% confidence interval (CI) 0.67-0.97). This patients with a confirmed diagnosis actually have a lower mortality rate than infection patients with an unconfirmed or an alternative infection diagnosis. Cause-specific plausibility of infection was not associated with mortality either in the entire patient patient with mortality either in the entire plausibility of infection was not associated (21%, 18%, 20%, 20% mortality in patients population admitted with a sepsis diagnosis respectively) and definite possible, probable, with infection likelihoods of none, sites except for the lungs. infection of presumed or in any of the main subgroups 4 Figure for all four categories overlap, meaning The confidence intervals infection plausibility. sepsis) diagnoses that were present in patients by category of infection likelihood. in patients by category present sepsis) diagnoses that were O 3 Figure of infection. The sources presumed prevalent by infection likelihood and the most acquired pneumonia and hospital-acquired pneumonia); abdominal infections (primary and and (primary infections abdominal pneumonia); hospital-acquired and pneumonia acquired infections, catheter-related infections (primary blood stream secondary peritonitis); blood stream tract infections and skin/soft tissue infections. urinary and endocarditis), infections, blood stream of patients in the group no likelihoods of “none” were there whole cohort. Furthermore, with skin or soft tissue infection. Figure 2: Figure infections (community- sites of infection. Distribution of plausibility of infection for lung frequent
Infection diagnosis in presumed sepsis 4 52 Infection diagnosis in presumed sepsis 53 4
(CSHR 0.93; 95% CI 0.85-1.02). In subgroup analyses, the mortality hazard for each for each analyses, the mortality hazard (CSHR 0.93; 95% CI 0.85-1.02). In subgroup 0.85, 95% hospital was similar (hospital A: SHR 0.80, 95% CI 0.62-1.03; hospital B: SHR syndrome, distress of the adult respiratory the prevalence CI 0.63-1.13). Furthermore, and the length of stay significantly increased of acute kidney injury, the prevalence ICU-acquired of occurrence the whereas <0.001), (p likelihoods infection greater with Patient outcomes for various sites of infection stratified by plausibility of infection. stratified by plausibility of infection. 3: Patient outcomes for various sites of infection Figure unit- Intensive care crude associations. The length of ICU stay (LoS) is shown as median. Data are defined as infections that started > 48 hours after admission with infections (ICU-AI) were acquired distress Acute kidney injury (AKI) and adult respiratory a plausibility of infection of at least possible. taken into account. during ICU admission were at or occurred present (ARDS) that were syndrome results of the Cochran-Armitage Whiskers indicate the 95% confidence interval. P-values indicate the of not shown because Urinary tract and skin/soft tissue infections are test for trend. chi-square unit. ICU, intensive care after stratification. Abbreviations: small subgroups relatively . This finding was was finding This . 10 Our study is the first prospective comparison of sepsis diagnoses made by ICU comparison of sepsis diagnoses made by ICU Our study is the first prospective
Discussion determined diagnosis made by clinicians in the context the accuracy of the infection We for found that up to 43% of patients treated of sepsis upon admission to the ICU and results These assessment. post-hoc on infection an had have to unlikely were sepsis in many upon ICU admission is difficult show that making an accurate sepsis diagnosis therapy antimicrobial to withhold reluctant are cases and may indicate that clinicians disease even patients with life threatening for suspected infection when faced with when the plausibility of infection is low. physicians and post-hoc analyses of infection likelihoods based on strict diagnostic that the true incidence of sepsis upon ICU admission likely is criteria, revealing studies have specifically investigated the accuracy Only few previous overestimated. study found that 49% of patients were of sepsis diagnoses in the ICU. A French ICU the a new infection on for potentially unnecessarily treated infections did not (p-value 0.36). In the main subgroups of presumed infection sites, infection sites, of presumed main subgroups infections did not (p-value 0.36). In the with outcome parameters in this crude the infection likelihood was not associated analysis, except for pulmonary infections. Crude and adjusted cumulative incidence functions of mortality stratified by plausibility of mortality stratified by plausibility 4: Crude and adjusted cumulative incidence functions Figure was plotted by imputing average values of age, of infection. The adjusted curve (right panel) diabetes mellitus, respiratory malignancy, immunocompromise, disease, cardiovascular gender, site of infection, and the acute sepsis severity, recent surgery, renal insufficiency, insufficiency, into the model. health evaluation IV score physiology and chronic however based on the level of microbiological evidence and not on well-defined evidence and not on well-defined however based on the level of microbiological of patients the true percentage to appreciate diagnostic criteria, making it difficult of the correlation without infection in post-hoc analysis. Another study explored of presence post-hoc the with therapy antimicrobial of start the at certainty clinical
Infection diagnosis in presumed sepsis 4 54 Infection diagnosis in presumed sepsis 55 4 . 12,13 . 11 . In contrast to this previous study, the current the current study, . In contrast to this previous 3 , at present there are no biomarkers that provide that provide no biomarkers are there , at present 16 . While such biomarkers would be valuable in reducing valuable in reducing would be . While such biomarkers 14,17,18 . The primary aim of this latter investigation focused on antimicrobial use, use, on antimicrobial investigation focused aim of this latter . The primary 11 . While some biomarkers, such as procalcitonin, may aid in limiting the duration may aid in limiting the duration . While some biomarkers, such as procalcitonin, 14,15 A limitation of this study concerns the inherently somewhat complex CDC and ISFA limitation of this study concerns the inherently In a crude analysis the likelihood of infection in patients treated for suspected sepsis sepsis for suspected likelihood of infection in patients treated In a crude analysis the Multiple studies have been performed in the on the value of host biomarkers of antibiotic therapy in ICU patients namely how often administration of antimicrobials for suspected infection could be be infection could for suspected of antimicrobials often administration namely how with treated patients of proportion large a infection; of presence the by justified to the infection according (58 of the 125; 46%) actually had no empirical antibiotics specialist in the post-hoc assessment infectious diseases antibiotic use in this patient population, our current study suggests that for stratification study suggests that for stratification our current antibiotic use in this patient population, to risk for an adverse outcome, the infection diagnosis is less important. according We of infection. assessment for the presence infection definitions used for the post-hoc among the study team in a separate study, determined the diagnostic agreement therefore was found to be good and concordance infection process of prospective surveillance involved discussions among observers, discussions of prospective process physicians and with (senior) clinicians in multi-disciplinary meetings attended by critical care therefore All diagnoses were infection specialists, and continuous checks of data-integrity. an “ideal” situation with made after consensus. As such, our post hoc analyses represent our study availability of all diagnostic data collected after the acute event. Consequently, as an analysis of the adequacy of clinical action in the ICU, but should not be interpreted withrather as an attempt to assess the true incidence of infection in patients admitted surveys the rapid it is important to note that in large suspected sepsis. In this respect sufficient diagnostic accuracy to withhold antibiotics as initial therapy in ICU patients diagnostic accuracy to withhold antibiotics as initial therapy in ICU patients sufficient infection with suspected was not associated with mortality. Since several factors that impact on ICU mortality were on ICU mortality were Since several factors that impact mortality. was not associated with we performed survival analysis multivariable between groups, unequally distributed mortality. with increased likelihood of infection was associated and found that a lower post-hoc infection who had no infection in a presumed patients with In other words, Our design to patients with an infection. mortality rate compared analysis had a higher mortality rate in patients the increased behind reasons the was not suited to explore speculative remains “no infection”. It therefore with sepsis with a post-hoc diagnosis of aggressive other or antibiotics unnecessary of harmfulby caused is whether this effects of an alternative due to delayed recognition treatment a delay in effective treatments, alternative,the that simply or disease, in these patients is diagnosis non-infectious, these data Furthermore, infectious diagnosis. injurious than the rejected even more outcome ICU to contribute not does infection of absence or presence the that suggest response notion that the injurious host current to a significant extent and fit with the that from by infection in the context of sepsis is not fundamentally different triggered such as trauma and major surgery elicited by non-infectious critical illness discrimination between infectious and non-infectious causes of critical illness in the non-infectious causes of critical illness in the discrimination between infectious and ICU . 3 , suggesting that the, suggesting 19,20 events. Feasibility and validation. Am J Respir Med. Apr 15 2014;189(8):947-955. Crit Care Horan TC, Andrus M, Dudeck MA. CDC/ NHSN surveillance definition of health care- associated infection and criteria for specific setting. types of infections in the acute care Jun 2008;36(5):309-332. Am J Infect Control. Cohen J. The international sepsis Calandra T, definitions on forum consensus conference unit. Crit of infection in the intensive care Med. Jul 2005;33(7):1538-1548. Care Levy MM, Rhodes A, et al. Dellinger RP, Surviving Sepsis Campaign: international sepsis guidelines for management of severe Med. and septic shock, 2012. Intensive Care Feb 2013;39(2):165-228. R, Grundmann H, et al. M, Vonberg Wolkewitz Risk factors for the development of nosocomial 5. 6. 7. 8.
Jun 1992;101(6):1644-1655. Bone RC, Balk RA, Cerra FB, et al. Definitions Bone RC, Balk RA, Cerra FB, et al. Definitions and guidelines failure for sepsis and organ for the use of innovative therapies in sepsis. The ACCP/SCCM Consensus Conference Committee. American College of Chest Medicine. Physicians/Society of Critical Care Chest. sepsis Severe Angus DC, van der Poll T. and septic shock. N Engl J Med. Aug 29 2013;369(9):840-851. Bos LD, et PM, Ong DS, Klein Klouwenberg of Centers for agreement al. Interobserver criteria for and Prevention Disease Control classifying infections in critically ill patients. Med. Oct 2013;41(10):2373-2378. Crit Care MS, Ong PM, van Mourik Klein Klouwenberg of a novel implementation DS, et al. Electronic surveillance paradigm for ventilator-associated Another limitation involves the fact that this study was performedAnother limitation involves in two centers in the general ICU practice. not reflect Netherlands and may benefit of early antibiotic treatment in patients with infection is greater than the potential is greater patients with infection in early antibiotic treatment benefit of relative Notably, therapy in those without infection. harm antimicrobial of unnecessary criteria for definiteonly few pneumonia cases fulfilled the of infection, to other sources for pneumonia strict definition most likely caused by a relatively infection. This was 1. 2. 3. 4. References Sources of support Molecular Medicine (http:// for Translational This work was supported by the Center funding from research MARS (grant 04I-201). MB has received project www.ctmm.nl), (NWO Vici 918.76.611). of Scientific Research the Netherlands Organization ion Conclusion in patients of the accuracy of the infection diagnosis analysis This first prospective shows that the clinical diagnosis of sepsis with suspected sepsis on ICU admission as defined by CDC/ISF of infection, poorly with the actual presence corresponds suggest that a substantial portion of patients being diagnostic criteria. These results clinical sepsis trials may in fact not have infection, which may negatively in enrolled benefit of certain sepsis treatments. impact the power of such trials to show administration of broad spectrum antibiotics to patients with clinically diagnosed septic clinically diagnosed to patients with spectrum antibiotics of broad administration in survival increase with a time-dependent shock is associated
Infection diagnosis in presumed sepsis 4 56 Infection diagnosis in presumed sepsis 57 4 Feb 6 2010;375(9713):463-474. 6 Feb Lancet. Bouadma L, Luyt CE, Tubach F, et al. Use of et al. Use of F, Luyt CE, Tubach Bouadma L, to patients’ exposure to reduce procalcitonin (PRORATA units care intensive in antibiotics randomised controlled trial): a multicentre trial. A, Brunkhorst FM, Schlattmann C, Prkno Wacker for as a diagnostic marker Procalcitonin P. meta-analysis.and review systematic a sepsis: Lancet Infect Dis. May 2013;13(5):426-435. Fan X, et al. Accuracy of F, Wang Y, Wu sepsis diagnosis in plasma sTREM-1 for patients: a systematic systemic inflammatory Nov 29 meta-analysis. Crit Care. and review 2012;16(6):R229. et R, Martin-Loeches I, Phillips G, Ferrer reduces treatment antibiotic Empiric al. sepsis and septic shock mortality in severe a guideline- results from the first hour: from program. based performance improvement Med. Aug 2014;42(8):1749-1755. Crit Care KE, et al. Kumar A, Roberts D, Wood initiation of Duration of hypotension before therapy is the critical antimicrobial effective determinantshock.septic in human of survival Med. Jun 2006;34(6):1589-1596. Crit Care 16. 17. 18. 19. 20. Jan 1 2014;5(1):154-160. Crit Care. 2010;14(1):R15. Crit Care. Virulence. Trends Mol Med. Apr 2014;20(4):195- Trends Nov-Dec 2012;7(9):672-678. J Am Statist Assoc. Jun 1999;94:496-509. Deutschman CS, Tracey KJ. Sepsis: current KJ. Sepsis: current Deutschman CS, Tracey dogma and new perspectives. Immunity. Apr 17 2014;40(4):463-475. Marshall JC. Why have clinical trials in sepsis failed? 203. Pierrakos C, Vincent JL. Sepsis biomarkers: a review. Reinhart K. Rapid diagnosis of Bloos F, sepsis. pneumonia and mortality on intensive care care and mortality on intensive pneumonia models. of competing risks units: application Apr 2008;12(2):R44. Critical Care. Hazards Proportional A RJ. Gray JP, Fine of a CompetingModel for the Subdistribution Risk. B, Dupont H, Gauzit R, Veber Montravers P, Lepape A. Strategies of initiation Bedos JP, of antibiotic therapy in and streamlining units. Crit Care. intensive care 41 French 2011;15(1):R17. S, Sprung CL, et al. Levin PD, Idrees use in the ICU: indications Antimicrobial trial. J Hosp and accuracy--an observational Med. 12. 13. 14. 15. 9. 10. 11. 7 8 7 4 4 4 3 3 5 4 4 49 34 11 10 93 22 17 17 11 11 10 14 11 81 35 90 112 843 210 310 Perforation Perforation bleeding Compl. of previous Upper GI bleeding Surgery ischemia Vascular Sepsis Sepsis Pneumonia Endocarditis infection for Surgery Soft tissue infection Pleural effusions ARDS Atelectasis Airway obstruction Emphysema Coma Subarachnoid hemorrhage Intracranial hemorrhage Overdose Subdural hematoma Congestive heart failure arrest Cardiac Aortic valve replacement surgery Cardiovascular Abdominal aortic aneurysm efinite Definite Gastrointestinal Infectious Respiratory Neurologic Cardiovascular 5 4 4 8 6 4 9 7 4 4 8 7 6 6 4 4 4 3 29 12 65 15 11 22 81 72 82 161 633 253 191 Subarachnoid hemorrhage Stroke Seizures Intracranial hemorrhage Subdural hematoma ARDS Emphysema Atelectasis Pleural effusions Airway obstruction Pneumonia Sepsis Endocarditis infection for Surgery Soft tissue infection arrest Cardiac Congestive heart failure Cardiomyopathy shock Cardiogenic Hemorrhage Perforation surgery Compl. of previous Hemorrhage Surgery ischemia Vascular Not shown are less frequently occurring metabolic, hematologic and genito-urinary organ systems, systems, occurring metabolic, hematologic and genito-urinary organ less frequently Not shown are trauma, and transplantation. Neurologic Respiratory Probable Infectious Cardiovascular Gastrointestinal 9 9 7 7 4 7 4 3 5 8 79 34 13 10 17 92 43 26 16 13 12 16 13 13 11 771 203 100 156 119 118 Cardiac arrest Cardiac Congestive heart failure CABG surgery Cardiovascular shock Cardiogenic Perforation ischemia Vascular surgery Compl. of previous Obstruction Hemorrhage Pneumonia Sepsis Renal infection Infection/abscess Soft tissue infection Subarachnoid hemorrhage Seizures Coma Stroke Subdural hematoma Emphysema Atelectasis Airway obstruction Pleural effusions Pulmonary embolus Possible Cardiovascular Gastrointestinal Infectious Neurologic Respiratory 6 5 5 9 4 3 3 2 7 5 4 3 5 4 3 3 4 2 26 13 16 12 22 11 38 65 50 39 332 117 Y DATA Hemorrhage surgery Compl. of previous Perforation GI surgery Obstruction Pneumonia Sepsis Endocarditis Renal infection Emphysema Pleural effusions Astma ARDS Atelectasis Subarachnoid hemorrhage Seizures Coma Intracranial hemorrhage Overdose Cardiac arrest Cardiac Congestive heart failure surgery Cardiovascular Rhythm disturbance Cardiomyopathy Concurrent diagnoses in patients admitted with sepsis by infection likelihood diagnoses in patients admitted with sepsis able S1: Concurrent ARDS adult respiratory distress syndrome; CABG coronary artery bypass grafting; Compl. artery bypass grafting; Compl. CABG coronary syndrome; distress ARDS adult respiratory shown. system are The top five of diagnoses per organ Complications; GI gastro-intestinal. T None Gastrointestinal Infectious Respiratory Neurologic Cardiovascular MENTAR SUPPLE tables Supplementary
Infection diagnosis in presumed sepsis 4 58 Infection diagnosis in presumed sepsis 59 4 7 5 4 4 8 7 4 4 4 3 3 93 22 17 17 49 34 11 10 14 11 11 11 10 90 81 35 112 843 310 210 Sepsis Sepsis Pneumonia Endocarditis infection for Surgery Soft tissue infection Perforation bleeding Compl. of previous Upper GI bleeding Surgery ischemia Vascular Congestive heart failure arrest Cardiac Aortic valve replacement surgery Cardiovascular Abdominal aortic aneurysm Pleural effusions ARDS Atelectasis Airway obstruction Emphysema Coma Subarachnoid hemorrhage Intracranial hemorrhage Overdose Subdural hematoma Infectious Gastrointestinal efinite Definite Cardiovascular Respiratory Neurologic 8 6 4 5 4 4 9 7 4 4 8 7 6 6 4 4 4 3 65 29 12 15 11 22 81 72 82 161 633 253 191 ARDS Emphysema Atelectasis Pleural effusions Airway obstruction Subarachnoid hemorrhage Stroke Seizures Intracranial hemorrhage Subdural hematoma Perforation surgery Compl. of previous Hemorrhage Surgery ischemia Vascular Cardiac arrest Cardiac Congestive heart failure Cardiomyopathy shock Cardiogenic Hemorrhage Pneumonia Sepsis Endocarditis infection for Surgery Soft tissue infection Not shown are less frequently occurring metabolic, hematologic and genito-urinary organ systems, systems, metabolic, hematologic and genito-urinary organ occurring less frequently Not shown are trauma, and transplantation. Respiratory Neurologic Gastrointestinal Cardiovascular Probable Infectious 9 7 4 3 5 8 9 7 7 4 79 34 13 10 92 43 26 16 13 12 16 13 13 11 17 771 203 156 119 118 100 Cardiac arrest Cardiac Congestive heart failure CABG surgery Cardiovascular shock Cardiogenic Pneumonia Sepsis Renal infection Infection/abscess Soft tissue infection Subarachnoid hemorrhage Seizures Coma Stroke Subdural hematoma Emphysema Atelectasis Airway obstruction Pleural effusions Pulmonary embolus Perforation ischemia Vascular surgery Compl. of previous Obstruction Hemorrhage Possible Cardiovascular Infectious Neurologic Respiratory Gastrointestinal 6 5 5 7 5 4 3 5 4 3 3 4 2 9 4 3 3 2 26 13 16 12 22 11 65 50 39 38 332 117 Subarachnoid hemorrhage Seizures Coma Intracranial hemorrhage Overdose Emphysema Pleural effusions Astma ARDS Atelectasis Pneumonia Sepsis Endocarditis Renal infection Hemorrhage surgery Compl. of previous Perforation GI surgery Obstruction Cardiac arrest Cardiac Congestive heart failure surgery Cardiovascular Rhythm disturbance Cardiomyopathy Concurrent diagnoses in patients admitted with sepsis by infection likelihood diagnoses in patients admitted with sepsis able S1: Concurrent ARDS adult respiratory distress syndrome; CABG coronary artery bypass grafting; Compl. artery bypass grafting; Compl. CABG coronary syndrome; distress ARDS adult respiratory shown. system are The top five of diagnoses per organ Complications; GI gastro-intestinal. None T Neurologic Respiratory Infectious Gastrointestinal Cardiovascular
5 Electronic implementation of a novel surveillance paradigm for ventilator-associated events: feasibility and validation
Peter M.C. Klein Klouwenberg* Maaike S.M. van Mourik* David S.Y. Ong Janneke Horn Marcus J. Schultz Olaf L. Cremer Marc J. Bonten
*Contributed equally to this work
American Journal of Respiratory and Critical Care Medicine tract Abstract hampered is (VAP) pneumonia ventilator-associated of surveillance Accurate Rationale: for ventilator- a novel surveillance paradigm criteria. Recently, by subjective diagnostic introduced. was associated events (VAE) algorithm. determinethe new VAE the validity of surveillance using Objectives: To (2011- centers medical academic Dutch two in study cohort Prospective Methods: of implemented and included assessment electronically surveillance was 2012). VAE Concordance and VAP. IVAC) conditions (VAC, ventilator-associated (infection-related) assessed, along with clinical diagnoses was surveillance VAP with ongoing prospective minor of Consequences conditions. all of mortality associated and VAEs underlying evaluated. implementation were VAE in electronic differences included. and Main Results: 2,080 patients with 2,296 admissions were Measurements surveillance to prospective according and VAP VAE-VAP IVAC, Incidences of VAC, algorithm The VAE 4.2, 3.2 and 8.0 per 1,000 ventilation days, respectively. 10.0, were VAC surveillance. by prospective patients identified detected at most 32% of the VAP most often caused by volume overload and infections, but not necessarily signals were (1.5 2.5 VAC, for - 5.3) 3.9 (95% CI 2.9 for mortality were Subdistribution hazards VAP. identified by for VAP and 7.2 (5.1 - 10.3) 2.0 (1.1 - 3.6) for VAE-VAP - 4.1) for IVAC, considerably analyses, mortality estimates varied In sensitivity surveillance. prospective algorithm implementation. in electronic following minor differences was poor. algorithm and VAP between the novel VAE Conclusions: Concordance susceptible to small differences were of VAE Incidence and associated mortality needed to characterize the clinical studies are implementation. More in electronic institutions. different comparability of rates from and ensure entities underlying VAE
Evaluation of the ventilator-associated events paradigm 5 62 Evaluation of the ventilator-associated events paradigm 63 5 . . 14 18,19 . 4,5 able 1). . Consequently, variations variations . Consequently, . This algorithm identifies . This algorithm identifies 11,12 15-17 . Surveillance networks such as the such as the . Surveillance networks . Furthermore, studies of healthcare- . Furthermore, 1-3 . Low inter-rater reliability and poor and poor reliability . Low inter-rater 10 9,11,13 . VAP case-definitions are complex, labor-intensive complex, labor-intensive case-definitions are . VAP 7-9 . The Institutional Review Board approved an opt-out consent approved . The Institutional Review Board 20 . However, establishing the diagnosis of VAP is challenging and is challenging and diagnosis of VAP establishing the . However, 6 Ventilator-associated pneumonia (VAP) has been associated with increased been associated with increased has (VAP) pneumonia Ventilator-associated This study aimed to assess the feasibility and validity of surveillance based on and validity of surveillance based on This study aimed to assess the feasibility in the form of abstracts reported of this study have been previously Some results These limitations of VAP surveillance have led to the development and led to the development and surveillance have These limitations of VAP method (IRB number 10-056C). For the current study, we analyzed all adult patients who study, method (IRB number 10-056C). For the current Methods Study design and population and RiskThis cohort study was incorporated in the ongoing MARS (Molecular Diagnosis referral units (ICU) of two tertiary in the mixed intensive care Stratification of Sepsis) project centers in the Netherlands uction Introd Monitoring care. to medical considerable burden infections add a Healthcare-associated years, important in recent have become increasingly of these infections and prevention (mandatory) public reporting along with a rise in mortality and length of stay. Therefore it is one of the major infections targeted by by one of the major infections targeted it is Therefore of stay. mortality and length surveillance programs concerns have been voiced regarding the reliability of VAP surveillance and its use as surveillance and its use as of VAP the reliability concernsbeen voiced regarding have a tool for hospital benchmarking in implementation of VAP surveillance across hospitals affect the reported VAP rates rates VAP reported the hospitals affect surveillance across in implementation of VAP comparisons valid inter-hospital and preclude associated infection surveillance and prevention are vulnerable to assessment bias vulnerable are associated infection surveillance and prevention ventilator-associated conditions (VAC) and infection-related ventilator-associated ventilator-associated and infection-related conditions (VAC) ventilator-associated using objective case definitions as entities for public reporting, conditions (IVAC) and leukocyte temperature prescriptions, based on ventilator settings, antimicrobial defines amenable to automated implementation. The algorithm further counts that are for within-hospital quality monitoring (T VAP possible and probable electronic implementation of the newly introduced VAE algorithm in a multicenter in a multicenter algorithm VAE implementation of the newly introduced electronic signals were and IVAC causes of VAC setting in the Netherlands. For this purpose, compared were results VAE to evaluate the face validity of VAE, assessed in order calculated. were estimates mortality and registration, VAP prospective ongoing to algorithm performed several adaptions of the VAE to evaluate Sensitivity analyses were implementation. to electronic of several approaches and assess the robustness National Healthcare Safety Network (NHSN) allow for benchmarking among hospitals among hospitals (NHSN) allow for benchmarking Safety Network National Healthcare programs guidance for improvement rates offers and feedback of infection for subjective interpretation and leave room correlation with histopathology have also been described correlation implementation by the NHSN of a new surveillance paradigm for ventilated patients surveillance paradigm for ventilated patients implementation by the NHSN of a new (VAE) events that aims to assess ventilator-associated
. st 16 O increase in daily in daily O increase 2 Brief definition VAP and probable possible VAP IVAC, Includes VAC, deterioration respiratory sustained, New, period of stability or after a two-day baseline on mechanical ventilation improvement VAC with evidence of infection (new antibiotics of infection (new antibiotics with evidence VAC and inflammatory signs) evidence of pneumonia with microbiological IVAC or probable) (classified as possible surveillance by prospective Evidence of VAP or probable as definite, definition (classified of clinical combination Requires possible VAP). evidence. microbiological signs, radiographic and Abbreviation VAE VAC IVAC VAE-VAP PROSP- VAP O and > 10%, respectively. In addition, the original algorithm In addition, the original algorithm O and > 10%, respectively. 2 ). See supplementary text for details. ). See supplementary text for details. 2 2012. Patients with do-not-resuscitate orders and patients ventilated in and patients orders 2012. Patients with do-not-resuscitate st
to > 5 cm H 2 The second analysis aimed to evaluate the robustness of electronic implementation implementation of electronic The second analysis aimed to evaluate the robustness In sensitivity analyses, we assessed two sets of alternativeIn sensitivity analyses, we assessed two of the implementations Overview of conditions evaluated in this study. conditions evaluated able 1: Overview of T Entity Ventilator-associated event Ventilator-associated Ventilator-associated condition Infection-related ventilator-associated condition ventilator-associated Ventilator-associated Ventilator-associated pneumonia (VAE) Ventilator-associated pneumonia the prone position were included in the analysis. Patients on rescue mechanical ventilation mechanical Patients on rescue the analysis. included in were position prone the included, but were membrane oxygenation) ventilation, extracorporeal (high-frequency the analysis. excluded from ventilation were the days on rescue 2011 and July 1 defined by either a > 3 cm H conditions are Ventilator-associated The main analysis was performed using the VAE algorithm implemented as specified as specified implemented algorithm analysis was performedmain The VAE using the extracted data (minute-to-minute using electronically NHSN protocol by the current clinical characteristics) and data use, microbiology antibiotic ventilator settings, mplementation of the ventilator-associated events algorithm Implementation of the ventilator-associated had received two or more consecutive days of mechanical ventilation between January 1 consecutive days of mechanical ventilation two or more had received minimum positive end-expiratory pressure (PEEP) or > 20% increase in fraction of in fraction of increase (PEEP) or > 20% minimum positive end-expiratory pressure (FiO oxygen inspired does not incorporate the time patients are in spontaneous breathing trials when being trials when being in spontaneous breathing does not incorporate the time patients are ignores this As ventilation. on nighttime are they when or ventilator the off weaned performed a sensitivity analysis was the patient for that day, best condition of the respiratory deterioration leading to discontinuation of weaning trials that classifies air conditions This concept was implemented by setting the PEEP to room as a VAC. than 2.5 hours (PEEP=0) if the patient was in a spontaneous weaning trial for more 1 ).(10%) of the day (intermittent rule, Figure MV specify the guidelines do not precisely algorithm, as the current of the VAE original algorithm. The first implementation aimed to identify more representative representative aimed to identify more original algorithm. The first implementation PEEP in levels of increases deterioration by varying the required episodes of respiratory and FiO
Evaluation of the ventilator-associated events paradigm 5 64 Evaluation of the ventilator-associated events paradigm 65 5
th ; this was considered the the ; this was considered 20
=21%). Right panel – example of artifact filtering by both the =21%). Right panel – example of artifact filtering by both the 2 ). VAP was classified as definite, probable or possible, probable or possible, was classified as definite, ). VAP percentile rule) from one hospital were manually reviewed manually reviewed one hospital were rule) from percentile th able S1 T percentile cut-off (thus excluding the lowest 10% of measurements, 10% of measurements, (thus excluding the lowest cut-off percentile th percentile rule – or by selecting the lowest setting maintained for at at the lowest setting maintained for rule – or by selecting percentile th Left panel – hypothetical example of a patient on intermittent ventilation under the ) – 10 percentile, PEEP – positive end-expiratory pressure, VAC – ventilator-associated condition. – ventilator-associated VAC pressure, PEEP – positive end-expiratory percentile, th . Reviewers were unaware of the patients’ IVAC and VAP status. status. and VAP of the patients’ IVAC unaware . Reviewers were
percentile (lowest 10 percent of daily measurements excluded) and the sustained settings rule excluded) and the sustained settings rule of daily measurements (lowest 10 percent percentile th lternative diagnoses eference standard eference able 2 least one consecutive hour (sustained settings rule). hour (sustained settings rule). least one consecutive Figure 1 Figure reference standard ( standard reference algorithm. of the VAE and assessment was fully independent (settings must be maintained for at least 60 consecutive minutes to qualify). The duration of the dip of the dip (settings must be maintained for at least 60 consecutive minutes to qualify). The duration 10 both 30 minutes. Abbreviations: in PEEP on day 4 was 80 minutes; the two dips on day 5 were 10 Figure 1: Figure original algorithm and the intermittent ventilation sensitivity analysis. If mechanical ventilation was set to hours (i.e. 10% of the day) then daily minimum values were than 2.5 interrupted for more FiO air conditions (PEEP=0 and room A by signal, all patients flagged VAC lead to a may conditions what to assess order In algorithm (10 the VAC R by dedicated well-trained for the development of VAP assessed daily Patients were reliability inter-rater observers with ongoing evaluation of the effect of possible artifacts in the data, we excluded outliers from the minute-to- the from outliers excluded we the data, in artifacts possible of effect the either after the minimum daily ventilator settings by selecting minute measurements application of a 10 requirements for electronic data capture. As opposed to the original implementation, implementation, original the to opposed As capture. data electronic for requirements of evaluated the effect we measurements, the use of minute-to-minute that involves which only, validated measurements manually by using hourly, frequency sampling form. to assess Second, availability of data in electronic common may better reflect third reviewer if necessary. The diagnoses that we explicitly considered are defined in defined in are The diagnoses that we explicitly considered if necessary. reviewer third T by two independent physicians (PK, MvM), with consensus through discussion with a discussion with a with consensus through by two independent physicians (PK, MvM), perc – 10 perc -- -- 1 (3) 2 (6) 5 (19) 3 (10) 5 (16) 4 (13) 4 (13) 3 (10) n = 31 14 (45) 12 (39) (%) IVAC -- 2 (2) 5 (6) 6 (7) 9 (11) 9 (11) n = 81 23 (28) 12 (15) 10 (12) 23 (28) 10 (12) 10 (12) Frequency identified Frequency (%) VAC
percentile rule. Patients may have more than one than one rule. Patients may have more percentile th Identified by of antibiotics, clinical documentation explicit clinical documentation Imaging, cultures, initiation Imaging, cultures, Bronchoscopy, imaging, imaging, Bronchoscopy, Chest tube placement Therapeutic anticoagulants Pleural drainage Explicit clinical documentation Clinical documentation, microbiology, Clinical documentation, microbiology, initiation of antimicrobials Initiation of diuretics Inotropy or afterload reduction or afterload reduction Inotropy Laparotomy, ascites drainage ascites drainage Laparotomy, Imaging & clinical findings * Respiratory tract infection Atelectasis/sputum plug Pneumothorax Pulmonary embolus Pleural effusion Aspiration New onset of SIRS/sepsis Volume overload Volume Heart failure Abdominal distension Acute neurological event Acute neurological We assessed the effects of VAEs and VAP on ICU mortality using competing-risk on ICU mortality using competing-risk and VAP of VAEs assessed the effects We Clinical conditions that occurred in the five-day window surrounding VAC and IVAC VAC and IVAC five-day window surrounding in the that occurred able 2: Clinical conditions Includes ventilator-associated pneumonia and (ongoing) other respiratory tract infections such as as such infections tract other respiratory (ongoing) and pneumonia ventilator-associated Includes alternative diagnosis. Sources of extra-pulmonary infection were: bloodstream (2), abdominal (2), abdominal bloodstream of extra-pulmonary infection were: alternative diagnosis. Sources cerebrovascular events included Acute neurological (5), mediastinal (1), and muscoskeletal (1). (1), encephalopathy (3), and meningitis (1). intracranial pressure accidents (5), cases of increased * Pulmonary conditions Diagnosis T A). events (hospital worsening pre-existing pneumonia. worsening pre-existing Systemic condition, SIRS – ventilator-associated – infection related IVAC Abbreviations: condition. VAC – ventilator-associated Inflammatory Response Syndrome, Extra-pulmonary infection Cardiac/circulatory Other Notes: VAC was identified using the 10 was identified using Notes: VAC No reason for VAC identified for VAC No reason All ICU admissions were included to assess population characteristics and concordance included to assess population characteristics and concordance All ICU admissions were surveillance, VAP prospective between Concordance methods. surveillance between VAE algorithm was assessed, both at the ICU VAP likelihood, and the stratified by or IVAC VAC, admission level and by using a window of +/- 2 days surrounding VAE-VAP and original algorithm that defines IVAC This is analogous to the VAE-VAP. VAC. based on a five-day window around estimated by the cause- on outcome were and VAP of VAEs effects analyses. The direct Statistical analyses
Evaluation of the ventilator-associated events paradigm 5 66 Evaluation of the ventilator-associated events paradigm 67 5 able S2). . All analyses were were . All analyses 21 . ). A large fraction fraction able 3). A large 22,23
levels, and four by both. There were 66 IVACs in 65 patients in 65 patients 66 IVACs were levels, and four by both. There 2 ). For IVAC and VAE-VAP the sensitivities were 18% (95% CI 18% (95% CI the sensitivities were and VAE-VAP able 4). For IVAC
able 3). Using the original algorithm, 158 VACs were detected in 152 patients (10.0/1,000 detected in 152 patients (10.0/1,000 were Using the original algorithm, 158 VACs All analyses were performed using SAS 9.2 (Cary, NC), R version 2.14 (www.r- NC), R version 2.14 performed using SAS 9.2 (Cary, All analyses were 12 – 27%) and 17% (95% CI 10 – 25%), respectively, with positive predictive values values with positive predictive 25%), respectively, (95% CI 10 – and 17% 12 – 27%) the to concordance restricting When 53%). – (25 38% and 45%) – 21 CI (95% 32% of within the five-day surveillance must have occurred by prospective window (VAP VAE 13%, 9% and 6% for were sensitivities for detecting VAP VAC), window surrounding values of 10%, 18% and with positive predictive respectively, and VAE-VAP, IVAC VAC, not temporally associated with VAP. were and IVAC 16%. Thus most episodes of VAC and VAP , IVAC of VAC Concordance VAP was 33% or definite) for detection of (possible, probable The sensitivity of VAC value of 25% (95% (95% confidence interval (CI) 25 – 42%), with a positive predictive CI 18 – 33%) (T s Result of whom 2,080 admitted to the ICU, 3,473 patients were During the study period, ventilated for at least two consecutive calendar days. patients (2,296 admissions) were Overall ICU mortality was patients. surgical Median age was 62 years and 44% were 21% (T PEEP settings, increasing by triggered were days of MV). Most events (n=149) FiO five by increased specific hazard ratios (CSHR) for each event (ICU discharge or ICU death). To evaluate To evaluate death). or ICU event (ICU discharge (CSHR) for each ratios specific hazard into (discharge) competing event on death, taking the of the events effect the direct (SHR) ratio subdistribution hazard we calculated the account, adjusted for age, gender, Acute Physiology and Chronic Health Evaluation (APACHE) (APACHE) Health Evaluation Acute Physiology and Chronic adjusted for age, gender, and VAE-VAP IVAC, and hospital. VAC, vs. medical) admission type (surgical IV score, variables included as time-dependent were VAP prospective of VAC (and IVAC) events occurred on the third or fourth day of mechanical ventilation ventilation or fourth day of mechanical on the third events occurred (and IVAC) of VAC not achieve a baseline period of stability; of (46%). Of the 2,296 admissions, 108 did was of onset of ventilation. The incidence of VAE these, 60 deceased within four 4 days to ventilation compared days of mechanical seven or more higher in patients receiving 3.6 versus 12.9 per mechanical ventilation (VAC patients with less than seven days of VAE-VAP IVAC, of MV). Rates of VAC, 0.6 vs. 5.8 per 1,000 days IVAC 1,000 days of MV, comparable between both ICUs (T were VAP monitored and prospectively (4.2/1,000 MV days). All IVAC episodes that fulfilled the antibiotic exposure criteria also criteria also the antibiotic exposure episodes that fulfilled (4.2/1,000 MV days). All IVAC 51 episodes were and/or white blood cell count definition. There met the temperature days) in algorithm (3.2/1,000 MV the VAE to according VAP possible or probable of by identified VAP were or definite probable 50 patients and 127 episodes of possible, surveillance in 115 patients (8.0/1,000 MV days) (T prospective ) and SPSS 20 (IBM Software, Armonk NY). Software, ) and SPSS 20 (IBM project.org 8.0 1(1) 127 9 (8) 6 (5) 8 (7) 76 (66) 16 (14) 16 (14) 37 (32) 17 (15) 61 (53) 23 (20) 48 (42) 31 (27) 13 (11) 12 (10) 36 (31) 20 (17) 25 (22) 33 (29) 39 (34) n = 115 62 (49 – 71) 14 (10 – 28) 15 (12 – 34) PROSP-VAP 77 (56 – 100) 0 51 3.2 3 (6) 1 (2) n=50 5 (10) 8 (16) 9 (18) 5 (10) 8 (16) 9 (18) 6 (12) 6 (12) 8 (16) 32 (64) 37 (74) 22 (44) 13 (26) 14 (28) 22 (44) 13 (26) 12 (24) -VAP VAE 57 (48 – 62) 15 (10 – 29) 18 (12 – 33) 84 (56 – 102) 66 4.2 1 (2) 4 (6) 1 (2) IVAC 9 (14) 8 (12) 8 (12) 8 (12) 8 (12) n = 65 41 (63) 12 (18) 12 (18) 45 (69) 25 (38) 14 (21) 18 (28) 15 (23) 29 (45) 10 (15) 17 (26) 18 (28) 57 (49 – 66) 15 (12 – 19) 19 (13 – 33) 79 (62 – 102) 158 9 (6) 3 (2) 10.0 VAC 14 (9) 26 (17) 18 (12) 31 (20) 25 (16) 36 (24) 91 (60) 47 (31) 51 (34) 37 (24) 17 (11) 22 (14) 64 (42) 24 (16) 18 (12) 43 (28) 56 (37) 102 (67) n = 152 13 (8 – 27) 60 (52 – 69) 14 (10 – 29) 82 (65 – 105) -- -- ll All 54 (2) 164 (7) 171 (7) 199 (9) 4 (2 – 8) 5 (2 – 9) 372 (16) 247 (11) 334 (15) 426 (19) 567 (25) 647 (28) 810 (35) 640 (27) 324 (14) 707 (31) 347 (15) 540 (24) 648 (28) 476 (21) patients 1,406 (61) 1,303 (57) n = 2,296 62 (50 – 72) 75 (57 – 97) Cerebrovascular disease Cerebrovascular Congestive heart failure Diabetes Diabetes COPD Malignancy Surgical elective Surgical Surgical emergency Surgical Medical Other ICU Other ward Emergency department Emergency Operating theatre Other, unknown Other, Internal medicine Cardiothoracic surgery Cardiothoracic Neurology or neurosurgery Neurology Other surgery Patient characteristics and incidence of ventilator-associated events and ventilator- events of ventilator-associated and incidence able 3: Patient characteristics Number of events T pneumonia. associated ) N (%) or median (IQR Incidence (/1,000 MV) Age Male Comorbidities Admission type Admission source Readmission Primary specialty APACHE IV IV APACHE Duration of mechanical ventilation Length of ICU stay Length of ICU stay Notes: VAE defined according to original algorithm, PROSP-VAP is as defined by the MARS study is as defined by the MARS study to original algorithm, PROSP-VAP defined according Notes: VAE definite). or (includes possible, probable Health Evaluation, COPD – chronic – Acute Physiology and Chronic APACHE Abbreviations: ventilator- – infection related unit, IVAC obstructive pulmonary disease, ICU – intensive care event, – ventilator-associated condition, VAE – ventilator-associated associated condition, VAC pneumonia. ventilator-associated – (prospective) (PROSP –) VAP Deceased in ICU
Evaluation of the ventilator-associated events paradigm 5 68 Evaluation of the ventilator-associated events paradigm 69 5 50 65 otal 152 able 2).able T 2,296 2,246 2,231 2,144 (> 10 vs. > 20%) 20%) vs. > (> 10 31 44 2 114 2,181 2,150 2,137 2,067 Absent 96 19 94 21 77 38 115 Any ). Among the 35 episodes the 35 episodes Among able 4). O PEEP increase) was evaluated. This was evaluated. This O PEEP increase) 83 71 12 70 13 59 24 2 rospective VAP Prospective ossible Possible 7 8 29 22 21 16 13 robable Probable 3 3 0 3 0 2 1 efinite Definite Absent Present Absent Present Absent Present Using the algorithm with handling of intermittent weaning, 241 patients with 261 In a retrospective analysis of underlying clinical conditions, pneumonia, either VAP VAP either pneumonia, conditions, clinical underlying of analysis retrospective a In When restricting the reference standard to probable and definite VAP, events of of events VAP, and definite to probable standard the reference When restricting Concordance between prospective VAP surveillance and VAE events detected by the the events detected by surveillance and VAE VAP prospective between able 4: Concordance Notes: For VAE algorithm, VAP includes possible or probable for VAE and in prospective and in prospective for VAE includes possible or probable algorithm, VAP Notes: For VAE and definite. surveillance possible, probable ventilator-associated – VAC condition, ventilator-associated related infection – IVAC Abbreviations: pneumonia. – ventilator-associated event, VAP – ventilator-associated condition, VAE Total VAE-VAP VAE-VAP IVAC VAC T the patient level. algorithm at original VAE resulted in 224 episodes of VAC in 213 patients with a sensitivity of 43% and positive in 213 patients with a sensitivity of 43% and positive in 224 episodes of VAC resulted value of 23% for the detection of VAP. predictive with the original algorithm. identified, with only 101 episodes concordant were VACs As the majority of VACs was identified by increases in PEEP settings, the effect of of in PEEP settings, the effect was identified by increases As the majority of VACs setting a higher trigger (> 5 vs. > 3 cm H daptations in the algorithm Adaptations in the algorithm resulted in 51 VACs with a sensitivity of 7% and positive predictive value of 22% for the value of 22% for the with a sensitivity of 7% and positive predictive in 51 VACs resulted of FiO increase required the lowering Conversely, VAP. of detection of probable or definite prospective VAP (32 patients) there were 25 episodes that did 25 episodes that did were VAP (32 patients) there or definite prospective of probable was no baseline period of stability there VAC, either because not fulfill the criteria for in ventilator settings (n=19). increase (n=6) or no (sufficient) (T cause of VAC the most often observed pneumonia, appeared or pre-existing value was moderate (kappa = 0.51). The positive predictive agreement The inter-rater pneumonia, hospital-acquired tract infections combined (VAP, for all respiratory of IVAC tract infection) increased pneumonia and other lower respiratory community-acquired to the five-day concordance when restricting to 66% at the patient level and 36% for detection of these combined of IVAC window (data not shown). The sensitivity was 7% and 3%, respectively. conditions, however, VAC, IVAC and VAE-VAP detected 44%, 25%, and 22% cases of VAP, respectively. respectively. detected 44%, 25%, and 22% cases of VAP, and VAE-VAP IVAC VAC, (T 11% and 12% 9%, were values predictive Positive able S3). T ). The able 5). The able 5, percentile rule, SHR 5.2 5.2 SHR rule, percentile th percentile modification yielded similar results as the results as the similar modification yielded percentile th percentile rule in the original algorithm did not change the overall algorithm did not change the overall rule in the original percentile th percentile and the sustained settings rule. percentile Estimates of associated mortality for VAC identified using the various electronic identified using the various electronic Estimates of associated mortality for VAC Of the 158 episodes identified by the original algorithm, 104 (65%) were also also original algorithm, 104 (65%) were Of the 158 episodes identified by the Using hourly (validated) measurements resulted in 152 episodes of VAC in 149 in 149 in 152 episodes of VAC resulted measurements Using hourly (validated) th association was strongest for VAC (time-averaged SHR 3.9 (95% CI 2.9 – 5.3)), and (time-averaged SHR 3.9 (95% CI 2.9 – 5.3)), and for VAC association was strongest 3.9, 95% CI 2.9 – 5.3), (SHR VAE-VAP (SHR 2.5, 95% CI 1.5 – 4.1) and lower for IVAC hazard surveillance had the highest subdistribution prospective to according VAP ratios Analysis of the cause-specific hazard ratio for death (7.2, 95% CI 5.1 – 10.3). but not the surveillance, identified by prospective and VAP that VAC (CSHR) revealed of dying (T on the hazard effect had a significant direct other VAEs, Association with mortality of hazard significantly associated with an increased were and VAP All types of VAE (T the competing events process ICU death when taking into accounting Applying the 10 being discharged of in a lower daily probability resulted In addition, all types of VAE exposing patients longer to a daily risk of dying the ICU after the onset of VAE, from of is mainly the result risk of dying in the ICU after VAE in the ICU - thus the increased was no There on mortality. of VAE effect stay in the ICU rather than the direct prolonged and the surgical) admission (medical or between the type of ICU interaction significant effect their to regards with VAP) prospective VAE-VAP, IVAC, (VAC, conditions different for slightly was a trend there on estimated associated mortality although for VAE-VAP the opposite was observed for patients whereas higher associated mortality in surgical surveillance. identified by prospective VAP 10 the for 8.4) – 4.8 CI (95% 6.3 SHR were implementations This adapted algorithm had higher sensitivity (50%) and similar positive predictive positive predictive (50%) and similar higher sensitivity algorithm had This adapted surveillance. with prospective to concordance with respect value (24%) implementation Reliability of electronic the detected VAC in 152 patients), but only 117 of (158 events incidence of VAC identical. The 10 episodes were identified by all other sensitivity analyses. Thus, although the total number of episodes episodes Thus, although the total number of identified by all other sensitivity analyses. between the algorithms differences were there identified was similar for all algorithms, identified. in the types of episodes that were original VAC algorithm with respect to concordance with VAP surveillance (data not surveillance (data not with VAP concordance to respect algorithm with original VAC with a with VAC in 157 patients sustained settings rule resulted shown). Applying the value of 25%. a positive predictive sensitivity of 34% and of 23% for the value of 30% and a positive predictive patients with a sensitivity value and 11% positive predictive at the patient level (13% sensitivity detection of VAP detected Of the 152 episodes of VAC when examining episode-level concordance). detected by the 113 identical episodes were when using hourly (validated) measured, 10
Evaluation of the ventilator-associated events paradigm 5 70 Evaluation of the ventilator-associated events paradigm 71 5 § * PROSP VAP 39/104 (37.5) 0.45 (0.34 – 0.58) 1.34 (1.11 – 1.61) 2.00 (1.34 – 3.00) 1.03 (1.02 – 1.03) 1.04 (0.86 – 1.24) 1.10 (0.90 – 1.35) 1.00 (0.99 – 1.00) 7.24 (5.09 – 10.3) * -VAP VAE 12/43 (27.9) 0.56 (0.30 – 1.05) 1.34 (1.11 – 1.61) 1.11 (0.60 – 2.05) 1.03 (1.03 – 1.03) 1.04 (0.86 – 1.26) 1.99 (1.11 – 3.58) 1.06 (0.86 – 1.13) 1.00 (0.99 – 1.00) * IVAC 17/56 (30.4) 0.47 (0.33 – 0.66) 1.34 (1.11 – 1.61) 1.03 (1.02 – 1.03) 0.98 (0.57 – 1.70) 1.02 (0.86 – 1.26) 2.51 (1.52 – 4.12) 1.09 (0.86 – 1.30) 1.00 (0.99 – 1.00) * VAC 51/134 (38.1) 0.38 (0.26 – 0.56) 1.32 (1.10 – 1.59) 1.03 (1.02 – 1.03) 3.96 (2.43 – 6.45) 1.01 (0.84 – 1.23) 3.92 (2.88 – 5.34) 1.06 (0.87 – 1.31) 1.00 (0.99 – 1.00)
(95% CI) † Admission type Hospital APACHE IV APACHE Gender (male=ref) (surgical = ref) (surgical Age Multivariable subdistribution hazards model for ICU mortality to account for competing account for competing for ICU mortality to model subdistribution hazards able 5: Multivariable No recovery from any of the events was assumed. † Time-averaged subdistribution hazard ratio ratio subdistribution hazard any of the events was assumed. † Time-averaged from recovery No SHR CSHR discharge (95% CI) CSHR discharge Notes: Definitions are based on the original algorithm. If a patient was admitted multiple times, admitted multiple times, based on the original algorithm. If a patient was Notes: Definitions are interaction between was no significant models there one admission was randomly selected. In all and event. hospital and event, or between admission type CSHR death (95% CI) T outcomes * Covariates (SHR) Crude mortality due to the time-varying nature of the event (this means that the SHR may vary depending on the vary depending on the of the event (this means that the SHR may due to the time-varying nature on day 8). § The a VAC SHR than on day 3 may have a different timing of the event, e.g. a VAC resulted in a similar SHR (7.45, 95% CI 4.12-13.50). VAP and definite inclusion of only probable CI – confidence health evaluation IV, – acute physiology and chronic IV APACHE Abbreviations: related ventilator-associated – infection ratio, IVAC interval, CSHR –cause-specific hazard – condition, VAE – ventilator-associated ratio, VAC condition, SHR – subdistribution hazard pneumonia. ventilator-associated – (prospective) event, (PROSP –) VAP ventilator-associated Discussion of a novel surveillance paradigm for ventilator- The development and implementation reliable surveillance towards associated events exemplifies the ongoing efforts infections, and in particular of ventilator-associated of healthcare-associated had events VAP and IVAC VAC, the study this In pneumonia. and complications the surveillance, especially when restricting VAP with prospective poor concordance VAE the however, Importantly events. VAC surrounding window the five-day to analysis complications range of ventilator-associated algorithm aims at identifying a broader Although a significant review. retrospective from our and this is confirmed by findings could be attributed to pulmonary infections, albeit and IVAC fraction of cases of VAC infections non-pulmonary overload, as volume such conditions VAP, to not limited of a VAE also commonly implicated. Occurrence and a variety of other causes were (95% CI 3.9 – 6.9) for the sustained settings rule and SHR 6.3 (95% CI 4.7 – 8.5) for the 4.7 – 8.5) for the (95% CI and SHR 6.3 CI 3.9 – 6.9) for the sustained settings rule (95% hourly sampling scheme. . The higher incidence of VAC in these studies in these studies . The higher incidence of VAC 24,27 . These studies have also shown that VAC reflects a broader scope scope a broader reflects . These studies have also shown that VAC 15,24-26 Several other studies have also found moderate concordance of VAC with VAP with VAP of VAC moderate concordance Several other studies have also found of clinical conditions than VAP alone of clinical conditions than VAP may have resulted from differences in implementation of the algorithm or differences in implementation of the algorithm or differences differences from may have resulted of study is the first assessing the concordance between study populations. The present prospective pre-existing VAE algorithm, comparing it to a identified by the and VAP IVAC implementation. of electronic surveillance and evaluating the reliability VAP nterpretations of findings and implications Interpretations the United States as a novel tool for algorithm has been implemented in The VAE complications, not limited to surveillance and benchmarking of ventilator-associated and their usefulness difficult rates remains of VAE interpretation pneumonia. However, helpful be could criteria Several established. been yet not has improvement quality for of outcome the all, Above entity. surveillance novel this of validity the evaluating in all aspects of what it is intended to and measure be clinically relevant should interest analysis of alternative assess. Although this cannot be formally tested, the retrospective of a diversity measure and IVAC study shows that VAC diagnoses in the present practices. ventilation of) (quality with associated be not may which of some conditions, conditions for the VAE – one of the major target the detection of VAP Moreover, or fourth day occurring on the third number of VACs The large algorithm – was poor. opposed as deterioration clinical ongoing of representative be could ventilation of respiratory to detect appeared IVACs Furthermore, care. quality of insufficient to a surveillance method should to mechanical ventilation. Ideally, infections not related in their underlying risk of patients that differ between groups also identify differences Because this study was not aimed at detecting of developing the event of interest. implemented were VAP for interventions no and hospitals between differences aspect key a Furthermore, evaluated. be not could this study, of period the during VAE of conditions identified by the to be assessed is the preventability that remains was associated with an increased risk of death in ICU, however not as strongly as as however not as strongly of death in ICU, risk with an increased was associated both IVAC Interestingly, surveillance. by prospective identified of VAP the occurrence of ICU with lower likelihoods associated algorithm were VAE the defined by and VAP infection that other conditions than (pulmonary) possibly indicating mortality than VAC, Furthermore, associated mortality of VAC. for at least part of the responsible were two academic of the algorithm is feasible in implementation although electronic the implementation affect of electronic in the method centers, subtle differences In the absence of mortality. by the algorithm and their associated events identified using both detailed minute- the algorithm was implemented detailed specifications, interestingly practical hourly sampling scheme; and a more to-minute data collection and had patients different VAE in of episodes different identified both data sources surveillance. VAP prospective with similarly poor concordance with ICU mortality and at the patient level and some association of VAC occurrence length of stay
Evaluation of the ventilator-associated events paradigm 5 72 Evaluation of the ventilator-associated events paradigm 73 5
2 than the current than the current 17 . 25 . Using a decreased FiO . Using a decreased 28 . Although the VAE algorithm uses objective objective uses algorithm VAE the Although . 11 , thus care must be taken when comparing rates rates comparing when taken be must care thus , 29 increases. In our ICUs we have implemented the higher implemented the higher In our ICUs we have increases. 2 . However, as opposed to the retrospective study setting, the study setting, the as opposed to the retrospective . However, 20 protocol from the ARDSnet guidelines the ARDSnet guidelines from protocol 2 Introduction of the VAE algorithm was driven by a desire for more objective, efficient objective, efficient for more algorithm was driven by a desire of the VAE Introduction In the present study, the great majority of VACs were identified by increases in in by increases identified were majority of VACs the great study, In the present collected through manual surveillance with electronic surveillance. Since standardized Since standardized surveillance. manual surveillance with electronic collected through surveillance is not yet universally implemented, comparability across electronic remains systems manual or implementations electronic different using institutions from Finally, in the future. questionable and these concerns will need to be addressed to case-mix developments with regard the perspective of benchmarking, additional comparisons can be made. valid inter-hospital necessary before are correction current process of prospective surveillance involves discussions among observers, observers, surveillance involves discussions among of prospective process current by critical discussions with (senior) clinicians in multi-disciplinary meetings attended All physicians and infection specialists, and continuous checks of data-integrity. care criteria and is amenable to automated implementation, our sensitivity analyses implementation, our sensitivity analyses criteria and is amenable to automated important implementation lead to in electronic modifications small demonstrate that In addition, mortality. in events detected and estimates of associated differences often different have shown that manually collected variables are studies previous electronically collected those from cut-off improved sensitivity for the detection of VAP with a similar positive predictive predictive positive similar a with of VAP detection the for sensitivity improved cut-off alternative to the original be a preferable setting, may therefore value, which, in our during the study unknown algorithm was largely the VAE algorithm. Importantly, clinicians’ decisions to Netherlands, therefore period and has not been adopted in the expected to be fully independent of the new algorithm change ventilator settings are and in compliance with local protocols. of complication of mechanical ventilation measures and reliable algorithm and their effect on patient-centered outcomes. Intervention trials evaluating trials evaluating Intervention outcomes. on patient-centered and their effect algorithm to answer this needed are IVAC and VAC at targeted programs quality improvement of such the design may help to improve study present the from and results question – clinical practice guideline implementation VAP post-hoc analysis, studies. In a recent rates but not VAC compliance - modestly decreased overall guideline with increasing VAC rates and associated with lower alone was measure No specific preventive IVAC. not evaluated were specifically at VAE interventions targeted Limitations in this study, used standard This study has several limitations. The reference vulnerable to the disadvantages of VAP surveillance, is inherently VAP prospective was done prospectively the assessment However, surveillance described previously. algorithm. of the VAE and completely independent multiple well-trained assessors by between raters was high overall (89%), but A prior study found the agreement (35%) lower for VAP PEEP/lower FiO manual assessment of VAP occurrence occurrence VAP of assessment manual PEEP as opposed to FiO PEEP as opposed . Furthermore, . Furthermore, 11 . However, we would not expect much better concordance expect much better concordance we would not . However, 30,31 and reliability in settings with higher VAP rates. Finally, the competing-events analysis analysis the competing-events rates. Finally, in settings with higher VAP and reliability not as confounders, but did hospital and baseline APACHE adjusted for age, gender, confounding may remain. thus some residual include time-varying confounders and for all the entities compared. identical the adjustment methods used were However, We would like to acknowledge Cristian Spitoni for statistical advice and Mark de Jong would like to acknowledge Cristian Spitoni for statistical advice and Mark de Jong We mechanical ventilator data. for retrieving um Members of the MARS consorti Academic Medical Friso M. de Beer and Lieuwe D.J. Bos (Department of Intensive Care, (Department of Frencken the Netherlands), Jos F. University of Amsterdam, Center, Medicine and Julius Center for Health Sciences and Primary Care, Intensive Care Acknowledgements Conclusions the novel surveillance paradigm for by VAP This study shows (i) that detection of a broad represent (ii) that events detected as VAE events is poor, ventilator-associated and thus do measures be liable to preventive range of clinical conditions that may not in electronic and (iii) that small differences quality of care, not necessarily represent incidence rates and associated mortality of implementation can considerably affect underlying entities clinical the establish to needed studies are events detected. More that rates obtained and ensure events, develop methods for case-mix adjustment VAE an comparable prior to considering these events as institutions are different from paradigm given these important concerns, the VAE established quality metric. Finally, of surveillance for VAP. should not be used as a sole method prospective diagnoses were therefore made after consensus. In addition, the reliability the reliability In addition, made after consensus. therefore were diagnoses prospective studies reported similar to previously study was of the retrospective although the participating centers did not routinely perform bronchoalveolar lavage lavage perform bronchoalveolar did not routinely centers participating the although for the diagnostic this is very representative clinical suspicion of VAP, with a in patients of our findings. worldwide and thus adds to the generalizability practices in most ICUs tract (SDD) all patients are decontamination of the digestive In our setting of selective collecting endotracheal specimens according pathogens by for respiratory screened Respiratory and at least twice weekly. on admission protocol to a standardized prior to the introduction suspected of VAP all patients obtained from specimens were or in all definite confirmation was present therapy and microbiological of antimicrobial potentially regimen SDD the Second, cases. possible of thirds two in and cases probable lowers the risk of VAP
Evaluation of the ventilator-associated events paradigm 5 74 Evaluation of the ventilator-associated events paradigm 75 5 Oct 1999;54(10):867-873. the National Healthcare Safety Network and the National Healthcare criteria.Physicians Chest of College American Med. Jan 2012;40(1):281-284. Crit Care Uckay I, Ahmed QA, Sax H, Pittet D. as a pneumonia Ventilator-associated quality indicator for patient safety? Clin Infect Dis. Feb 15 2008;46(4):557-563. PL, Ferguson SM, Fakhry CP, Michetti R. Ventilator- FO, Gross Cook A, Moore associated pneumonia rates at major with a national trauma centers compared benchmark: a multi-institutional study of May Surg. Acute Care J Trauma the AAST. 2012;72(5):1165-1173. Horan TC, Andrus M, Dudeck MA. CDC/ NHSN surveillance definition of health care- associated infection and criteria for specific setting. types of infections in the acute care Jun 2008;36(5):309-332. Am J Infect Control. variability in Klompas M. Interobserver pneumonia surveillance. ventilator-associated Apr 2010;38(3):237-239. Am J Infect Control. A, et al. N, Ewig S, Torres Fabregas Clinical diagnosis of ventilator associated validation comparative revisited: pneumonia using immediate post-mortem lung biopsies. Thorax. Klompas M. Eight initiatives that misleadingly lower ventilator-associated Jun pneumonia rates. Am J Infect Control. 2012;40(5):408-410. 8. 9. 10. 11. 12. 13. http://www.apic.org/Resource_/ Mar-Apr Public Health Rep. Mar-Apr Dec 2009;37(10):783-805. Lancet Infect Dis. Apr 24 2013. APIC. HAI Reporting Laws and Regulations. APIC. HAI Reporting Laws and Regulations. 2011; TinyMceFileManager/Advocacy-PDFs/HAI_ map.gif. Accessed February 11, 2013. CL, Jr., JR, Richards Klevens RM, Edwards et al. Estimating health care-associated infections and deaths in U.S. hospitals, 2002. 2007;122(2):160-166. Webster WH, Greene RS, Gitomer RJ, Panzer PR, Landry KR, Riccobono CA. Increasing JAMA. demands for quality measurement. Nov 13 2013;310(18):1971-1980. et al. JR, Peterson KD, Mu Y, Edwards Safety Network (NHSN) National Healthcare data summary for 2006 through report: 2008, issued December 2009. Am J Infect Control. van der Kooi TI, Mannien J, Wille JC, van nosocomial of Benthem BH. Prevalence infections in The Netherlands, 2007-2008: studies. J of the first four national results Hosp Infect. Jul 2010;75(3):168-172. RH, et al. Melsen WG, Rovers MM, Groenwold Attributable mortality of ventilator-associated pneumonia: a meta-analysis of individual prevention randomised patient data from studies. Skrupky LP, McConnell K, Dallas J, Kollef Skrupky LP, MH. A comparison of ventilator-associated to pneumonia rates as identified according 1. References 2. 3. 4. support Sources of Medicine (http:// Molecular by the Center for Translational This work was supported funding research MB has received MARS (grant 04I-201). project www.ctmm.nl), (NWO Vici 918.76.611). Scientific Research of the Netherlands Organization from study; collection, in the design and conduct of the play a role The sponsors did not or review, of the data; and preparation, and interpretation management, analysis, the manuscript; and decision to submit the manuscript for publication. of approval University Medical Center Utrecht, the Netherlands), Gerie J. Glas, Roosmarijn T.M. T.M. J. Glas, Roosmarijn Netherlands), Gerie the Medical Center Utrecht, University Straat, Lonneke A. Marleen Laura R.A. Schouten, Mischa A. Huson, van Hooijdonk, Luuk Wieske, Maryse A. Wiewel,van Vught, Esther Witteveen and of (Department the Netherlands). University of Amsterdam, Center, Academic Medical Intensive Care, 5. 6. 7. Cochrane Cochrane http://www.ardsnet.org/system/ Dec 2012;40(12):3154-3161. Hayashi Y, Morisawa K, Klompas M, et al. al. et M, Klompas K, Morisawa Y, Hayashi the impact surveillance: improved Toward complications on of ventilator-associated in patients and antibiotic use length of stay Feb units. Clin Infect Dis. in intensive care 2013;56(4):471-477. et al. Heyland DK, T, J, Sinuff Muscedere of Preventability The Clinical Impact and Critically Conditions in Ventilator-Associated Patients. Chest. Ill Mechanically Ventilated Sep 12 2013. Senkal S, Gajic O, Ding S, Kilickaya O, of Trends G. Temporal Li RD, Hubmayr Pneumonia Incidence Ventilator-Associated of Implementing Health-care and the Effect Chest. Community. Bundles in a Suburban Nov 1 2013;144(5):1461-1468. Klompas M, Magill S, Robicsek A, et al. for definitions surveillance Objective pneumonia. Crit Care ventilator-associated Med. ARDSnet. Mechanical Ventilation Protocol Protocol ARDSnet. Mechanical Ventilation Summary of the NIH NHLBI ARDS Clinical Network. . files/Ventilator%20Protocol%20Card.pdf Accessed September 25, 2013. Bosman RJ, Oudemane van Straaten HM, The use of intensive care Zandstra DF. information systems alters outcome prediction. Sep 1998;24(9):953-958. Med. Intensive Care Brazzi V, Torri S, Pifferi R, A, D’Amico Liberati to reduce L, Parmelli prophylaxis E. Antibiotic tract infections and mortality in respiratory care. intensive receiving adults 2009(4):CD000022. Database Syst Rev. epidemiology: Healthcare MJ. Bonten preventing pneumonia: Ventilator-associated the inevitable. Clin Infect Dis. Jan 1 2011;52(1):115-121. 24. 25. 26. 27. 28. 29. 30. 31.
PLoS One. 2011;6(3):e18062. Dec 2008;61(12):1216-1221. The National Healthcare Safety Safety The National Healthcare Nov 2013;41(11):2467-2475. J Am Statist Assoc. Jun 1999;94:496-509. . New York: Springer; 2012. . New York: Beyersmann J, Gastmeier P, Wolkewitz M, M, Wolkewitz P, Gastmeier J, Beyersmann mathematical proof easy M. An Schumacher showed that time-dependent bias inevitably estimation. J Clin leads to biased effect Epidemiol. M. Schumacher A, Allignol J, Beyersmann Competing risks and multistate models with R NHSN. Module: Network Device-Associated Protocol. Event Ventilator-Associated Mourik MS, Ong PM, Van Klein Klouwenberg of OL, Bonten MJ. Validation DS, Cremer a new surveillance algorithm for ventilator associated pneumonia. ECCMID. Berlin2013. PM, Ong Mourik MS, Klein Klouwenberg Van of the and assessment DS, et al. Validation new surveillance paradigm for ventilator- Resistance associated events. Antimicrobial 2013;2(Suppl 1):O64. and Infection Control PM, Ong DS, Bos LD, et Klein Klouwenberg of Centers for agreement al. Interobserver criteria for and Prevention Disease Control classifying infections in critically ill patients. Med. Oct 2013;41(10):2373-2378. Crit Care Hazards RJ. A Proportional Gray Fine JP, CompetingSubdistribution of a the for Model Risk. Lin MY, Bonten MJ. The dilemma of The dilemma of Bonten MJ. Lin MY, research. bias in infection control assessment May 2012;54(9):1342-1347. Clin Infect Dis. al. et K, Kleinman Y, Khan M, Klompas of a novel surveillance Multicenter evaluation of mechanical paradigm for complications ventilation. for Disease Control Atlanta, USA: Centers January 2013. and Prevention; R, et al. Developing Magill SS, Klompas M, Balk to surveillance national approach new, a events*. Crit Care for ventilator-associated Med. 22. 23. 16. 18. 19. 20. 21. 14. 15. 17.
Evaluation of the ventilator-associated events paradigm 5 76 Evaluation of the ventilator-associated events paradigm 77 5 O or a > 20% increase of the daily minimum fraction of daily minimum fraction of of the O or a > 20% increase 2 Y DATA ), both sustained for at least two calendar days, are classified as as classified are days, calendar two least at for sustained both ), 2 settings were used for the VAC algorithm. algorithm. used for the VAC settings were 2 Source: NHSN. The national healthcare safety network device-associated module: safety network device-associated module: NHSN. The national healthcare Source: Possible or probable ventilator-associated pneumonia (VAP) pneumonia ventilator-associated Possible or probable tract respiratory for infection was collected from evidence Definition: Microbiological purulent requires A possible VAP an IVAC. time window surrounding in the cultures epithelial cells per low and < 100 as determined by > 25 neutrophils secretions respiratory respiratory sample. a from power field or a qualitative or semi-quantitative positive culture and (semi-)quantitative secretions is diagnosed by purulent respiratory VAP A probable pleural fluid, histopathology from positive cultures or samples respiratory from results viruses. samples or positive diagnostic tests for Legionella spp or respiratory results status was performed using microbiology Implementation: Assessment of VAP records. electronic and sputum characteristics extracted from and Atlanta, USA: Centers for Disease Control associated event protocol. Ventilator- 2013. Prevention; : The presence of VAC with the occurrence of hypothermia or fever, leukocytosis leukocytosis of hypothermia or fever, with the occurrence of VAC Definition: The presence antibiotic therapy for at least four calendar days or leukopenia and the initiation of new event. The first two days of mechanical the VAC in the two-day window surrounding on admission. inclusion of conditions present excluded to prevent ventilation are and clinical Implementation: Informationantibiotic use, clinical chemistry results, on IVAC the in considered Antibiotics records. electronic from extracted were signs Network (NHSN) with the Safety as described by National Healthcare algorithm were purposes or prokinetic solely for prophylactic exception that antibiotics prescribed or low-dose regimen selective digestive decontamination (routine excluded were to The algorithm was implemented independently by two researchers erythromycin). implementation. correct ensure Infection-related VAC (IVAC) VAC Infection-related : A sustained increase in ventilator requirements after a baseline period of a baseline period of after requirements in ventilator Definition: A sustained increase in daily minimum positive end-expiratory an increase Either stability or improvement. (PEEP) of > 3 cm H pressure MENTAR SUPPLE methods Supplementary entities and implementation algorithm of the VAE Brief descriptions condition (VAC) Ventilator-associated inspired oxygen (FiO oxygen inspired a VAC event. Of note, PEEP settings < 5 are considered equivalent to 5. equivalent to 5. considered settings < 5 are event. Of note, PEEP a VAC ICU extracted from ventilator settings were Implementation: Minute-to-minute the minimal PEEP day of ventilation (> 60 minutes of ventilation), databases. For each and FiO † 1,101 7,363 † Evident radiographic † Evident radiographic
29 (3.9) 67 (9.1) 33 (4.5) 71 (9.6) Hospital B Definite CPIS > 6 Evident abnormalities on radiographic examination OR of Radiographic evidence lung abscess or empyema Histopathologic evidence Histopathologic evidence of pneumonia (abscess with PMN concentration and positive tissue culture) OR If empyema, positive of aspirate. culture
) 3
N (rate/1,000 MV days) † ) or PSB (10 4 Evident abnormalities Detection of pathogen Probable CPIS > 6 on radiographic on radiographic examination in quantitative culture in BAL (10 airway culture from OR with Positive blood culture pathogen also isolated
Includes possible, probable and definite VAP. 127 events events 127 VAP. and definite probable possible, Includes ||
66 IVAC events in 65 patients, § Includes both possible and events in 65 patients, § Includes both possible and 66 IVAC 8,499 1,195 † 22 (2.6) 60 (7.1) 33 (3.9) 87 (10.2) Hospital A ratio, infiltrates on X-thorax and microbiology). 2 N (rate/1,000 MV days) /FiO 2 || on radiographic on radiographic examination in semi-quantitative culture secretions respiratory from (endotracheal aspirate aspirate) or bronchoscopic Possible CPIS > 6 Dubious abnormalities Detection of pathogen † § Incidence of VAE as implemented using the original algorithm by hospital using the original algorithm by hospital as implemented able S2: Incidence of VAE Definition of definite, probable or possible VAP by prospective surveillance (reference (reference prospective surveillance VAP by probable or possible able S1: Definition of definite, Prospective VAP Prospective events in152 patients, * 158 VAC VAE-VAP probable VAP. 51 events in 50 patients, events 51 VAP. probable in 115 patients. ventilator-associated – VAC condition, ventilator-associated related infection – IVAC Abbreviations: pneumonia. – ventilator-associated event, VAP – ventilator-associated condition, VAE N ventilator days events* VAC T N ICU admissions IVAC events IVAC Clinical criteria* Radiology T standard). Microbiology Notes: All events must occur during ICU admission and > 48 hours after onset of mechanical and > 48 hours after onset of mechanical Notes: All events must occur during ICU admission temperature, (based on tracheal secretions, score ventilation. * CPIS – clinical pulmonary infection leukocyte count, PaO abnormalities are defined as new or progressive infiltrates, consolidation, cavitation or pleural infiltrates, consolidation, cavitation or pleural defined as new or progressive abnormalities are effusion. lavage, PMN – polymorphonuclear leukocyte, PSB BAL – bronchoalveolar Abbreviations: PM, Ong DS, Bos LD, et al. specimen brush. For details: Klein Klouwenberg – protected of CDC criteria for classifying infections in critically ill patients. Crit Care agreement Interobserver Med 2013;41:2373-2378. Supplementary tables Supplementary
Evaluation of the ventilator-associated events paradigm 5 78 Evaluation of the ventilator-associated events paradigm 79 5 ------0.98 (0.96-1.00) 1.00 (1.00-1.00) 0.45 (0.34-0.58) 1.01 (1.00-1.01) 1.00 (0.90-1.10) 0.83 (0.75-0.91) 0.95 (0.86-1.05) 1.00 (1.00-1.00) 1.01 (1.00-1.01) 0.99 (0.89-1.09) 0.81 (0.74-0.90) 0.96 (0.87-1.06) 1.02 (1.01-1.04) 1.00 (1.00-1.00) 0.47 (0.33-0.66) 1.00 (1.00-1.01) 0.98 (0.88-1.08) 0.83 (0.75-0.91) 0.98 (0.89-1.08) 1.00 (1.00-1.00) 1.01 (1.00-1.01) 0.98 (0.89-1.09) 0.83 (0.75-0.92) 0.98 (0.89-1.09) 1.02 (1.01-1.02)* 0.38 (0.26-0.56)* 1.02 (1.02-1.02)* 1.02 (1.01-1.02)* 0.56 (0.30-1.05)* 1.02 (1.01-1.02)* ischarge (95% CI) CSHR Discharge -- 1.00 (1.00-1.00) 1.00 (1.00-1.00) 0.94 (0.92-0.96) 2.00 (1.34-3.00) 1.01 (0.82-1.21) 1.00 (1.00-1.00) 1.00 (1.00-1.00) 0.93 (0.91-0.95) 1.16 (0.96-1.40) 1.01 (0.83-1.23) 1.05 (0.85-1.29) 0.96 (0.93-0.98) 1.00 (1.00-1.00) 1.15 (0.95-1.38) 1.00 (1.00-1.00) 0.98 (0.57-1.70) 1.03 (0.85-1.25) 1.04 (0.85-1.28) 1.00 (1.00-1.00) 1.00 (1.00-1.00) 1.11 (0.60-2.05) 1.02 (0.84-1.24) 1.14 (0.95-1.38) 1.16 (0.96-1.40) 1.03 (1.02-1.04)* 1.04 (1.03-1.04)* 1.98 (1.48-2.65)* 3.96 (2.43-6.45)* 1.04 (1.04-1.04)* 1.03 (1.02-1.04)* 1.03 (1.02-1.04)* 1.04 (1.04-1.04)* 1.04 (1.03-1.04)* 1.82 (1.36-2.44)* 1.03 (1.02-1.04)* CI) CSHR Death (95% Age*time PROSP-VAP APACHE*time Medical admission*time Age Gender APACHE Medical admission Age*time APACHE*time Medical admission*time Hospital VAC Gender APACHE Medical admission VAC*time Age*time Age Hospital IVAC VAE-VAP APACHE*time Age Gender APACHE Medical admission Age*time APACHE*time Gender APACHE Medical admission VAE-VAP*time Hospital Age Hospital Cause-specific hazard ratios ratios hazard Cause-specific able S3: PROSP-VAP IV – acute physiology and APACHE interaction included. Abbreviations: Notes – * Time*exposure – IVAC ratio, CI – confidence interval, CSHR –cause-specific hazard health evaluation IV, chronic pneumonia – ventilator-associated condition, PROSP-VAP ventilator-associated infection related – ventilator-associated ratio, VAC SHR – subdistribution hazard surveillance, by prospective paradigm. pneumonia by the VAE – ventilator-associated condition, VAE-VAP VAC T IVAC VAE-VAP
PART III
CHALLENGES IN THE PROGNOSTICATION OF PATIENTS WITH SEPSIS
6 Intensive care unit-acquired infections after admission for sepsis or non-infectious disease: incidence and attributable mortality
Lonneke A. van Vught Peter M.C. Klein Klouwenberg Cristian Spitoni Maryse A. Wiewel Janneke Horn Marcus J. Schultz Marc J. Bonten Olaf L. Cremer Tom van der Poll =.17). In multivariable competing risk survival analysis, sepsis at admission admission at sepsis analysis, survival risk competing multivariable In =.17). P Abstract been which has suppression, immune associated with prolonged Context: Sepsis is late associated with secondary infections to susceptibility increased implicated in immune stimulatory therapy in patients propose investigators to prompting mortality, mortality analyses of the incidence and attributable comparative with sepsis. However, unit (ICU) with or care in patients admitted to the intensive of secondary infections not available. without sepsis are determine is associated with an whether a sepsis admission diagnosis Objective: To to admissions infections relative ICU-acquired the development of risk for increased ICU- of mortality attributable the compare to and illness, critical non-infectious for disease. admitted with sepsis or non-infectious infections in patients acquired observational study in the mixed ICUs of Design, setting and patients: Prospective January 2011 to July 2013, Netherlands from two tertiary teaching hospitals in the admissions) with an ICU length of stay of > involving 3,168 consecutive patients (3,640 48 hours. Attributable mortality was determined taking time using a multistate model before competing risks (death or discharge to acquisition of secondary infection and infection) into account. infections : Incidence and characteristics of ICU-acquired measures and outcome Main and their attributable mortality. infections in patients admitted with sepsis was Results: The incidence of ICU-acquired sepsis at admission (19 vs. 23/1,000 days, similar to the incidence in patients without respectively, infection (subdistribution risk of ICU-acquired was associated with a decreased in patients with sepsis infections 0.73, 95% CI 0.60-0.89). ICU-acquired ratio hazard infections or viral often caused by Gram-positive bacterial more at admission were gram-negative bacterial often by in non-infectious admissions more reactivation; infections ICU mortality of ICU-acquired infections. The overall adjusted attributable with a sepsis admission diagnosis (10%). was 13% by day 60, but was lower in patients with the : Sepsis at admission is negatively associated Conclusions and relevance infections. Secondary infections in patients admitted for development of ICU-acquired to non-sepsis admissions. sepsis have a lower attributable mortality compared
ICU-acquired infections in critically ill patients 6 84 ICU-acquired infections in critically ill patients 85 6 . In light of 3,4,6 , which includes , which includes 13,14 . However, remarkable remarkable . However, 10 . Sepsis patients who survive the first days. Sepsis patients who 1,2 . Both ICUs used standardized and protocolized care care protocolized and ICUs used standardized Both . 11,12 . However, immunostimulatory therapy could have deleterious . However, 7,8 . Several mechanisms have been implicated in sepsis-induced immune have been implicated in sepsis-induced . Several mechanisms . It has been suggested that immune suppression and secondary infections that immune suppression . It has been suggested 3-5 3,4 representing a paradigm shift from the use of anti-inflammatory agents in many a paradigm shift from representing and should ideally only be used in patients who might benefit of it. and should ideally only be used in patients 9 3-5 Reported incidence rates of ICU-acquired infections vary between 9% and 37%, 37%, and 9% between vary infections ICU-acquired of rates incidence Reported The primary objective of this study was to determineThe primary objective of this study was admission whether a sepsis occurring as a consequence thereof are important denominators of late sepsis mortalityimportant denominators of late sepsis are thereof occurring as a consequence in the ICU the application of nonabsorbable antibiotics in the oropharynx and the administration and the administration the application of nonabsorbable antibiotics in the oropharynx effects effects of intensive care unit (ICU) admission frequently enter a state of prolonged immune enter a state of prolonged frequently unit (ICU) admission of intensive care and viralvulnerable for secondary infections them which may render suppression, reactivation depending on the populations studied and the definitions used depending on the populations studied suppression, including apoptotic depletion and exhaustion of immune cells and an increase increase an and cells immune of exhaustion and depletion apoptotic including suppression, critical cells; in patients with non-infectious suppressor and myeloid-derived in T regulatory limited extent detectable or to a more not are illness these alterations methods, including selective decontamination of the digestive tract unsuccessful sepsis trails these findings immune stimulation has been suggested as a novel treatment strategy for a novel treatment stimulation has been suggested as these findings immune sepsis Study design and population Study design and population This study was conducted as part of the Molecular Diagnosis and Risk Stratification of observational study in the mixed ICUs of two prospective a large Sepsis (MARS) project, in the Netherlands between January 2011 and January 2014 tertiary teaching hospitals (ClinicalTrials. 24 hours stay > an expected length of encompassing all patients with NCT01905033) identifier gov Methods uction Introd patients andmortality in hospitalized of morbidity and the leading cause Sepsis is mortality rates globally associated with high little knowledge is available with regard to whether these numbers differ between between numbers differ to whether these little knowledge is available with regard disease. If the assumption holds true that sepsis patients admitted with sepsis or other ICU-acquired that expected is it suppression, immune to leading condition unique a is to non- often in patients admitted with sepsis compared infections will occur more with immunostimulatory for the design of clinical trials sepsis patients. Moreover, incidence of secondary infections is important, therapies not only knowledge of the but also of their attributable mortality. risk for the development of ICU-acquired increased diagnosis is associated with an secondary The illness. critical non-infectious for admissions to relative infections infections in the attributable mortality of ICU-acquired objective was to compare disease. patients admitted with sepsis or non-infectious . All . All 11,12 . The likelihood . The likelihood 15 . A competing risk . A competing risk 19 and International Sepsis . Patients with an infection an infection with . Patients 16 11 , as described in detail described in , as 17 . The SOFA score for the central nervous system (CNS) subgroup was was for the central nervous system (CNS) subgroup score . The SOFA 18 Organ failures were defined as a score ≥ 3 on the Sequential Organ Failure Failure ≥ 3 on the Sequential Organ as a score defined were failures Organ For the current analysis all consecutive patients admitted from January 2011 to July 2011 to July January admitted from all consecutive patients analysis For the current not included. Shock was defined by the use of vasopressors (noradrenaline) for for (noradrenaline) the use of vasopressors by was defined Shock not included. during at least 50% of the ICU day. hypotension in a dose of > 0.1mcg/kg/min likelihood of possible, probable or definite within 24 hours after ICU admission were or definite within 24 hours after ICU admission were likelihood of possible, probable to regarded all other ICU admissions were classified as having sepsis on admission; to define was used of therapeutic antibiotics The date of initiation be non-infectious. infection was defined as any possible, probable the start of infection. An ICU-acquired after ICU admittance and had to be unrelated or definite infection starting > 48 hours with onset of their infectious episode between to the admission diagnosis. Patients be determined excluded, for it could not were 24 and 48 hours after ICU admission for ICU admission or was the primary reason with certainty whether this infection a on classified based were The most likely causative microorganisms ICU-acquired. team. post-hoc assessment by the research ≥ 1 for which a score failure except for cardiovascular score, Assessment (SOFA) was used patient data were encrypted for privacy reasons. for privacy reasons. encrypted patient data were Forum (ISF) consensus definitions A multivariable competing risk survival model was used to evaluate the risk factors factors risk evaluate the to used was model survival risk multivariable competing A and death are infections taking into account that ICU discharge for ICU-acquired competing risks for the development of a secondary infections Statistical analysis Definitions other than none) diagnosed as suspected infection (with likelihood Sepsis was defined one additional parameter ICU admission accompanied by at least within 24 hours after 2001 Internationalas described in the Conference Sepsis Definitions of topical antibiotics in the gastrointestinal tract and systemic prophylaxis with an with an prophylaxis tract and systemic gastrointestinal antibiotics in the of topical 4 days of ICU stay. during the first cephalosporin third-generation intravenous risk for patients at all (i.e., selected hours were > 48 stay of length ICU an with 2013 by approved method opt-out an via included were Patients infection). ICU-acquired an of the participating hospitals (IRB no. 10-056C) board the institutional review analysis provides two measures of association: the cause specific hazard ratio (CSHR) ratio (CSHR) of association: the cause specific hazard two measures analysis provides ICU of the variables on outcome (for ICU discharge, effect which estimates the direct infection), and the subdistribution mortality and the development of an ICU-acquired ratio (SHR), which describes the risk for the development of an ICU-acquired hazard in this competing events. Admission variables included accounting for the infection of infection for which the clinical team initiated therapeutic antibiotics was post-hoc post-hoc was antibiotics therapeutic initiated team clinical the which for infection of physicians research or definite by dedicated ICU classified as none, possible, probable (CDC) and Prevention using Center for Disease Control
ICU-acquired infections in critically ill patients 6 86 ICU-acquired infections in critically ill patients 87 6 test or test or able 2). able 1). able 1 and T ). In both admission able S1). In both admission =.01); all other admission =.01); all other admission =.17) (T < .05 was considered statistically significant. < .05 was considered P . 22 . We incorporated the time dependency of the event dependency of the event incorporated the time . We 20,21 =.001), whereas non-infectious admissions for respiratory non-infectious admissions for respiratory =.001), whereas shows baseline characteristics stratified according to the to the according able 1 shows baseline characteristics stratified At baseline, patients from either admission group who did or did not develop who did or did not develop either admission group At baseline, patients from The attribution of ICU-acquired infection to mortality on the ICU was analyzed was analyzed infection to mortality on the ICU ICU-acquired The attribution of All data were analyzed using R studio build under R version 3.0.2, R-package “mstate” All data were Continuous non-parametric data were analyzed using a Mann-Whitney U analyzed using a Mann-Whitney data were Continuous non-parametric ). Almost half of the remaining 3,640 admissions concerned patients admitted 1). Almost half of the remaining Figure groups, patients who developed an ICU-acquired infection were more severely ill on ill on severely more infection were patients who developed an ICU-acquired groups, an infection while on the ICU, as admission to the ICU than those who did not acquire shock (T and more scores, IV and SOFA indicated by higher APACHE an ICU-acquired infection were similar with regard to age, gender, race or Charlson race or Charlson to age, gender, similar with regard infection were an ICU-acquired often developed ICU- more comorbidity index. Patients admitted with bacteremia infections (P acquired infections (P developed less ICU-acquired disorders with sepsis (1,719 or 47.2%), whereas 1,921 admissions (52.8%) were patients without patients without 1,921 admissions (52.8%) were with sepsis (1,719 or 47.2%), whereas sepsis on admission. T infection. Of sepsis on admission and the development of ICU-acquired of presence complicated by a total of 334 ICU-acquired all admissions for sepsis, 232 (13.4%) were 291 (15.1%) of admissions for non-infectious diseases were infections, whereas infections (P complicated by a total of 366 ICU-acquired equally distributed between patients who did and those who did diagnoses were infection (Supplementary T not develop an ICU-acquired We studied 5,920 patients during 6,994 ICU episodes, of which 3,725 (53.3%) (53.3%) 3,725 which of episodes, ICU 6,994 during patients studied 5,920 We excluded 85 admissions (2.3%) from hours. We admissions had an ICU-duration > 48 between 24 and 48 hours after ICU admission the analysis since sepsis was diagnosed ( s Result Study population model were age, gender, APACHE IV score, admission type (medical vs. surgical), the the type (medical vs. surgical), admission IV score, APACHE age, gender, model were 0-5) and (ranging from time of ICU admission failing at the of organs absolute number insufficiency, cardiovascular including immunodeficiency, of comorbidities, a number insufficiency. and respiratory renal malignancy, model using a multistate Kruskal-Wallis test, categorical data were analyzed using a Chi square or Fishers exact or Fishers exact using a Chi square analyzed test, categorical data were Kruskal-Wallis analyzed using a student’s All continuous parametric data were test, as appropriate. t-test or analysis of variance when appropriate. 2013, Vienna, Austria) Team (R Core (i.e., ICU-acquired infection) and the presence of competing risks (i.e., discharge or or of competing risks (i.e., discharge and the presence infection) (i.e., ICU-acquired covariables we included infection). Furthermore, ICU-acquired mortality without an For this analysis, only of disease severity. in baseline markers to adjust for differences mortality was was used. The population attributable infection the first ICU-acquired caused by secondary infections. of ICU mortality as the percentage expressed ). The source and causative ). The source able 2 T <.001) ( =.64). In a multivariable competing risk =.64). In a multivariable competing risk P P Flowchart ICU-acquired infections occurred later during ICU stay in patients with a sepsis admission infections occurred ICU-acquired As expected, antibiotic use prior to the development of an ICU-acquired infection or infection or development of an ICU-acquired As expected, antibiotic use prior to the U-acquired infections U-acquired diagnosis (median 9 days to first secondary infection) compared to patients with a non- diagnosis (median 9 days to first secondary infection) compared sepsis admission diagnosis (median 4 days, Patients with sepsis on admission developed 19 ICU-acquired infections per 1,000 infections per 1,000 Patients with sepsis on admission developed 19 ICU-acquired 23 ICU- risk days, while patients with a non-sepsis admission diagnosis developed infections per 1,000 risk days ( acquired risk for survival analysis a sepsis admission diagnosis was associated with a decreased infection (SHR 0.73, 95% CI 0.60-0.89). an ICU-acquired IC Figure 1. Figure diagnoses of sepsis or non-infectious disease. While all patients with a sepsis admission While all patients with a sepsis admission diagnoses of sepsis or non-infectious disease. still 79.0% of patients with the event, therapeutic antibiotics before diagnosis received antibiotics, although for prophylactic received a non-infectious admission diagnosis as selective decontamination of the digestive tract or surgery. such reasons a competing event (death or discharge) was different between patients with admission between patients with admission different was discharge) or (death event competing a
ICU-acquired infections in critically ill patients 6 88
ICU-acquired infections in critically ill patients 89 6
SD - standard deviation; SOFA - sequential organ failure score. failure organ sequential - SOFA deviation; standard - SD
Abbreviations: APACHE IV - Acute Physiology And Chronic Health Evaluation; ICU-AI - intensive care unit acquired infection; IQR - interquartile range; range; interquartile - IQR infection; acquired unit care intensive - ICU-AI Evaluation; Health Chronic And Physiology Acute - IV APACHE Abbreviations:
ICU-admission of unique patients, readmissions were not included. not were readmissions patients, unique of ICU-admission
Other complications include ICU-acquired weakness, pneumothorax and acute myocardial infarction. infarction. myocardial acute and pneumothorax weakness, ICU-acquired include complications Other Follow up data were calculated using the first first the using calculated were data up Follow
a a b
<.001 .14 <.001 (22.5%) 337 (37.8%) 101 <.001 (29.3%) 357 (45.7%) 86 (27.8%) 881 days 60
<.001 .02 <.001 (11.2%) 184 (26.1%) 76 <.001 (15.5%) 231 (35.8%) 83 (15.8%) 574 ICU
Mortality Mortality
b
<.001 .25 <.001 (1.4%) 23 (16.2%) 47 <.001 (4.1%) 61 (12.5%) 29 (4.4%) 160 injury lung Acute
.69 .35 <.001 (8.3%) 135 (21.3%) 62 <.001 (8.7%) 129 (25.0%) 58 (10.5%) 384 injury kidney Acute
Complications
<.001 <.001 <.001 [3-7] 3 [9-22] 13 <.001 [3-9] 5 [15-33] 22 [3-10] 6 stay of Length
utcome O
<.001 .13 <.001 (24.9%) 406 (38.1%) 111 <.001 (32.2%) 479 (44.8%) 104 (30.2%) 1,100 Shock
<.001 .01 .13 (81.6%) 1,330 (87.3%) 254 .10 (88.2%) 1,312 (94.0%) 218 (85.5%) 3,114 failure Organ
<.001 <.001 <.001 [3-8] 6 [5-9] 7 <.001 [4-9] 7 [6-11] 8 [4-9] 6 score SOFA
<.001 <.001 <.001 [49-87] 65 [57-92] 75 <.001 [62-100] 79 [72-107] 89 [56-93] 73 score IV APACHE
Severity of disease of Severity
Readmission <.001 <.001 1.00 (8.2%) 134 24(8.2%) .79 (18.2%) 270 (19.0%) 44 (13.0%) 472
Admission type (medical) type Admission <.001 <.001 .002 (49.2%) 802 (38.1%) 111 .71 (77.1%) 1,147 (77.2%) 179 (61.5%) 2,239
Charlson comorbidity index comorbidity Charlson <.001 .63 .11 [2-5] 3 [2-6] 4 .10 [2-6] 4 [2-5] 4 [2-5] 4
White race White .34 .40 .85 (89.4%) 1,338 (90.3%) 241 .95 (88.6%) 1,078 (88.8%) 167 (89.1%) 2,824
Gender male male Gender .74 .92 .15 (61.3%) 917 (65.9%) 176 .22 (60.6%) 738 (65.4%) 123 (61.7%) 1,954
Age .01 .55 .27 (17) 58 (17) 60 .25 (15) 60 (15) 59 (16) 59
emographics D
Admissions 1,487 232 3,640 1,630 291
Patients 3,168 188 1,217 P 267 1,496 P P P
and admission and AI U- IC AI U- IC o N AI U- IC AI U- IC o N AI U- IC AI U- IC o N
ll patients patients ll A
Sepsis admissions Sepsis on-infectious admissions on-infectious N Sepsis vs non-infectious vs Sepsis
Baseline characteristics and outcome and characteristics Baseline able 1. 1. able T P .01 .04 .13 .61 .01 .08 .01 .69 .002 .002 .002 .002 1.00 0.23 0.84 0.75 0.001 <.001 <.001 <.001 <.001 <.001 <.001 =.002), whereas =.002), whereas - 420 4 [3-7] 2 (0.5%) 4 (1.1%) 1 (0.3%) 6 (1.6%) (n=366) 22 (5.2%) 10 (2.4%) 10 (2.7%) 32 (8.7%) 28 (7.7%) 29 (7.9%) 28 (7.7%) 13 (3.6%) 38 (10.4%) 53 (14.5%) 71 (19.4%) 61 (16.7%) admissions 140 (33.3%) 124 (29.5%) 122 (29.1%) 177 (48.4%) 124 (33.9%) on-infectious Non-infectious 397 Sepsis Sepsis 9 [6-13] 8 (2.0%) 2 (0.6%) 9 (2.7%) 1 (0.3%) 2 (0.6%) 6 (1.8%) 4 (1.2%) (n=334) 32(8.1%) 33 (8.1%) 27 (8.1%) 18 (5.4%) 13 (3.9%) 85 (21.4%) 89 (22.4%) 64 (19.2%) 85 (25.4%) 58 (17.4%) 88 (26.3%) 53 (15.9%) 51 (15.3%) admissions 151 (38.0%) 106 (35.3%)
). Patients with a sepsis admission diagnosiswith a sepsis admission able S2). Patients ll All 818 T 7 [4-11] 2 (0.3%) 2 (0.3%) 6 (0.9%) (n=700) 54 (6.6%) 35 (4.3%) 18 (2.2%) 34 (4.9%) 26 (3.7%) 10 (1.4%) 38 (5.4%) 28 (4.0%) 85 (12.1%) 79 (11.3%) 80 (11.4%) admissions 275 (33.6%) 211 (25.8%) 225 (27.5%) 102 (14.6%) 182 (26.0%) 177 (25.3%) 149 (21.3%) 262 (37.4%) Supplementary =.002) and abdominal infection (15.3% versus 7.7%; P =.002) and abdominal able 2, a HAP VAP Bacteremia CRBSI Abdominal sepsis infection Gastrointestinal Brain abscess Primary meningitis Secondary meningitis Urinary tract Gram-positive bacteria Gram-negative bacteria Fungi/yeasts Viral Other Unknown Pulmonary Cardiovascular Abdomen Neurological Skin Other Other infections include lung abscess, sinusitis, pharyngitis, tracheobronchitis, endocarditis, endocarditis, Other infections include lung abscess, sinusitis, pharyngitis, tracheobronchitis, Characteristics of ICU-acquired infections able 2. Characteristics of ICU-acquired Onset ICU-acquired infection (day) Onset ICU-acquired Causative pathogen mediastinitis, myocarditis, unknown infection, post-operative wound infection, bone and joint infection, post-operative wound infection, bone and joint unknown mediastinitis, myocarditis, tract infection. infection, oral infection, eye infection, reproductive infection; HAP - Hospital acquired bloodstream related CRBSI - Catheter Abbreviations: associated pneumonia. - Ventilator pneumonia; VAP a T Source of infection Source were more likely to develop ICU-acquired catheter related bloodstream infections (26.3% infections bloodstream catheter related likely to develop ICU-acquired more were versus 16.7%; P pathogens of ICU-acquired infections differed between patients with sepsis or no sepsis at with sepsis or no between patients infections differed of ICU-acquired pathogens (T admission patients with a non-infectious admission diagnosis developed more pulmonary infections pulmonary more admission diagnosis developed patients with a non-infectious
ICU-acquired infections in critically ill patients 6 90 ICU-acquired infections in critically ill patients 91 6 able 3). 60 <.001), while yeasts 50 Sepsis 40 Day on ICU 30 Non-infectious 20 All 10 <.001). Gram-positive bacteria were more often involved in ICU- often involved more were bacteria <.001). Gram-positive 0 ). The attributable ICU mortality of an ICU-acquired infection was able 1). The attributable ICU mortality of an ICU-acquired 5 0 -5 25 20 15 10 30
-10 =.04). Viral infections were much more common in patients with a sepsis a with patients in common more much Viral=.04). were infections P (%) mortality ICU Attribuable Attributable mortality of intensive care unit-acquired infections. Attributable intensive infections. Attributable intensive unit-acquired Attributable mortality of intensive care =.01), whereas the opposite was true for gram-negative bacteria (22.4% versus 29.0%versus (22.4% bacteria for gram-negative true was opposite the whereas =.01), care unit mortality over time adjusted for the Acute Physiology and Chronic Health Evaluation IV IV Evaluation Health Chronic and Physiology Acute the for adjusted time over mortality unit care the total cohort, patients with sepsis at admission, and patients and age. The lines represent score AM - attributable mortality. with a non-infectious admission diagnosis. Abbreviations: Figure 2. Figure When the analysis was restricted to only patients with organ failure at admission, the at admission, the failure to only patients with organ When the analysis was restricted We performed subgroup analyses to determine the robustness of our data, focusing of our data, focusing analyses to determine performed the robustness subgroup We (T infections ICU-acquired of mortality attributable and incidence the on Subgroup analyses Subgroup 13.4% at day 60, which translates to an absolute mortality difference of 3.6% between absolute mortality difference 13.4% at day 60, which translates to an 2). The attributable in the ICU (Figure patients with and without secondary infections with a sepsis admission diagnosis. Of note, themortality was consistently lower in patients most in the first 10 days after ICU admission were negative values of attributable mortality ill patients encountering the competing event (mortality likely driven by the most severely infection. an ICU-acquired being able to develop infection) before without ICU-acquired Outcome length of stay and infection had a longer ICU an ICU-acquired Patients who developed the ICU. an infection while on who did not acquire complications than patients more one year after that was still present mortality rates, a difference They also had higher ICU admission (T (48.4% versus 25.4%; P (48.4% versus acquired infections in patients admitted with sepsis than in patients with non-infectiouswith sepsis than in patients admitted infections in acquired infections respectively, ICU-acquired versus 29.5% of all admission (38.0% disease on P respectively, 0.5%, P (8.0% of all secondary infections versus admission diagnosis and fungi were equally present as causative pathogens in both groups. pathogens in both groups. as causative equally present and fungi were AM 9.1% 8.5% 6.5% 13.4% 10.0% 25.0% 13.7% 28.3% 20.4% djusted Adjusted , versus 17.8% in our , versus 17.8% in our 60 days 25 Mortality 881 (27.8%) 443 (31.9%) 438 (25.4%) 775 (28.3%) 408 (32.1%) 367 (24.9%) 369 (35.3%) 208 (38.7%) 161 (31.8%) 14.4% 13.5% 15.1% 15.2% 14.2% 16.0% 19.5% 17.8% 21.4% . In a previous study the incidence study the incidence . In a previous Incidence ICU-AI (%) 10,23,24 (n) 537 507 3,168 1,405 1,763 2,742 1,270 1,472 1,044 Patients (n) 583 517 3,640 1,719 1,921 1,530 1,584 3,114 1,100 Admissions Sepsis admissions Non-infectious admissions Sepsis admissions Non-infectious admissions Sepsis admissions Non-infectious admissions The reported incidences of ICU-acquired infections vary widely, affected by the by the affected infections vary widely, incidences of ICU-acquired The reported Subgroup analyses on incidence and attributable mortality of ICU-acquired infections mortality of ICU-acquired and attributable analyses on incidence able 3. Subgroup T Entire cohort Entire Admissions with organ failure Admissions with organ Admissions with shock infection unit-acquired ICU-AI - intensive care AM - attributable mortality, Abbreviations: of ICU-acquired infections in septic shock patients reached 23% infections in septic shock patients reached of ICU-acquired cohort. We used strict criteria for defining infection, both at admission and during ICU used strict criteria for defining infection, both at admission and during ICU cohort. We infections for which the clinical team started therapeutic and 19% of presumed stay, to be not infectious based on post-hoc analysis by dedicated antibiotics was regarded Discussion to determine sought attributable mortality of ICU-acquired the incidence and We found that the overall life” clinical practice in the ICU. We infections in the “real infections to be 21/1,000 patient days at risk, and that the incidence of ICU-acquired comparable to that in patients without sepsis incidence in patients with sepsis was risk analysis sepsis at admission was associated at admission. In fact, in a competing indicate These results infection. risk of developing an ICU-acquired with a reduced other conditions as in similar levels of immune suppression sepsis results that severe infections was that lead to critically illness. In contrast, the incidence of ICU-acquired the severity or shock at admission, suggesting failure in patients with organ increased length of ICU stay as most influencing factors. increased of disease and the resulting definitions used and the populations analyzed incidence of ICU-acquired infections increased similarly in both patients with sepsis both patients with sepsis similarly in infections increased incidence of ICU-acquired 16%, P=.18). The incidence of ICU-acquired and no sepsis at admission (14.2% and no were but there was even higher, infections in patients with shock at admission and 21.4%, P=.15). (17.8% groups between both admission significant differences infections was lower in patients with shock The attributable mortality of ICU-acquired to compared sepsis with patients in lower significantly remained yet admission, at patients without sepsis at admission (P<.01).
ICU-acquired infections in critically ill patients 6 92 ICU-acquired infections in critically ill patients 93 6 . 26 . Although . Although 21,31 . In contrast to our to our . In contrast 11 , in patients with sepsis at admission , in patients with sepsis at admission 24,27-30 Our study has strengths and weaknesses. The strength of this study is that we this study is that we of and weaknesses. The strength Our study has strengths We observed some differences in the source and causative agents of ICU-acquired ICU-acquired of and causative agents in the source observed some differences We The attributable mortality of ICU-acquired infections was 13% by day 60, which infections was 13% by day 60, which The attributable mortality of ICU-acquired findings, a previous study reported an increased incidence of ICU-acquired infections infections of ICU-acquired incidence an increased reported study a previous findings, sepsis at admission to patients without compared ill patients with sepsis in critically the absolute mortality differences between patients with and without ICU-acquired and without ICU-acquired between patients with the absolute mortality differences mortality due to ICU- attributable similar in both cohorts, the (relative) infections were to was higher in patients without sepsis at admission compared infections acquired by the higher overall mortality in patients those with sepsis. This might be caused included that analyses subgroup in consistent was result This sepsis. with admitted mortality attributable low relatively The admission. at disease severe more with patients to mortality. more and shock) contribute failure suggests that other factors (e.g., organ followed all consecutive patients admitted to two mixed ICUs at risk for prospectively in detailed and validated documentation infection, resulting developing an ICU-acquired also comes with a of daily ICU events in clinical practice. This unbiased approach causative and sources different and heterogeneous was population the study limitation: In in the two admission groups. recorded infections were pathogens of ICU-acquired of care the Netherlands selective decontamination of the digestive tract is standard > 48 hours. While prior antibiotic patients with an expected ICU length of stay of for infections, in our cohort may influence incidence rates of ICU-acquired treatment received was not clear in the sub-analysis including only patients who such an effect so, More discharge). or death infection, (ICU-acquired event the before antibiotics ill the administration of selective decontamination of the digestive tract in all critically However, this investigation was biased due to inclusion of all patients with a length with a length was biased due to inclusion of all patients this investigation However, of non-sepsis patients. On in an overrepresentation resulted of stay > 24 hours which to time insufficient therefore and stay of lengths short patients had average these to patients at our analysis restricted therefore We infections. develop ICU-acquired In our hours. a length of stay > 48 with patients i.e., infections, risk for ICU-acquired involved < 48 hours, which predominantly discharged were cohort 46.7% of patients admission diagnosis (75.7%). patients with a non-infectious diagnosis. The former with a sepsis or non-infectious admission infections in patients to the high be related less secondary pneumonia, which could acquired group in distinguishing of patients admitted with pneumonia and difficulties percentage While the overall distribution of causative between an ongoing and a new lung infection. pathogens was comparable to earlier studies researchers making use of all clinical and microbiologic data data clinical and microbiologic making use of all researchers , Pseudomonas , Pseudomonas opportunistic pathogens like Staphylococcus epidermidis, enterococci commonly detected, hinting at possible more were aeruginosa and viral reactivation use of stay and different length of ICU increased Alternatively, immune suppression. infection could be at play. ICU-acquired antibiotics prior to development of an all nosocomial infections is comparable to other studies that included . This has prompted investigators to plea for immune stimulatory to plea for immune stimulatory investigators . This has prompted . Our study reveals that ICU-acquired infections do not occur more do not occur more infections that ICU-acquired . Our study reveals 3,4,6 3-5 While immune suppression is a well-documented phenomenon in critically ill in critically ill is a well-documented phenomenon While immune suppression therapy in patients with sepsis in order to prevent ICU-acquired infections and reduce infections and reduce ICU-acquired to prevent with sepsis in order therapy in patients late mortality Sources of support Molecular Medicine (http:// This work was supported by the Center for Translational funding from research MARS (grant 04I-201). MB has received project www.ctmm.nl), (NWO Vici 918.76.611). of Scientific Research the Netherlands Organization Friso M. de Beer, Lieuwe D. J. Bos (Department of Intensive Care, Academic Medical Academic Medical Lieuwe D. J. Bos (Department of Intensive Care, Friso M. de Beer, (Department of Intensive Care Frencken Jos F. University of Amsterdam), Center, University Medical Sciences and Primary Care, Medicine and Julius Center for Health M. van Hooijdonk, Mischa Gerie J. Glas, Roosmarijn T. Utrecht), Center Utrecht, University of Academic Medical Center, A.Huson, (Department of Intensive Care, Medicine and Julius of Intensive Care Ong (Department David S.Y. Amsterdam), University Medical Center Utrecht, Care, Center for Health Sciences and Primary Maryse A. R. A. Schouten, Marleen Straat, Lonneke A. van Vught, Laura Utrecht), Wiewel, Esther Witteveen, and Luuk Wieske Academic (Department of Intensive Care, the Netherlands). University of Amsterdam, Medical Center, um Consorti Members of the MARS patients results in a homogenous group in terms of antibiotic treatment in the first days days first the in termsin treatment antibiotic of group homogenous a in results patients of effect an purely is incidence reduced the that excluding thereby admission ICU after translation this may influence in sepsis patients, however protection better antimicrobial tract. Lastly, selective decontamination of the digestive to ICUs without of our results clinical for a diagnostic bias in which the way to correct was no possible statistical there patients infection in deteriorating an ICU-acquired to diagnose consideration was prone (e.g. non-infectious patients). antibiotic treatment not yet on therapeutic to display unique immune sepsis patients have been shown patients in general, with non-infectious in patients not or less present that are features compromising critical illness often and bear a lower attributable mortality in patients with a sepsis admission mortality in patients with a sepsis admission often and bear a lower attributable with patients with a non-infectious admission diagnosis. diagnosis when compared in sepsis patients is infections low attributable mortality of ICU-acquired The relatively infections ICU-acquired prevent to seeking trials of design future the for importance of in this population with mortality as endpoint.
ICU-acquired infections in critically ill patients 6 94 ICU-acquired infections in critically ill patients 95 6 Jun Jun Crit Care Care Crit
Am J Infect Control. Control. Infect J Am Oct 8 2014;312(14):1429-1437. Apr 2 2014;311(13):1308-1316. Apr 2003;31(4):1250-1256. Jan 1 2009;360(1):20-31. JAMA. Sep 2012;54(5):600-616. Levy MM, Fink MP, Marshall JC, et al. 2001 Marshall JC, et al. 2001 Levy MM, Fink MP, International SCCM/ESICM/ACCP/ATS/SIS Conference. Definitions Sepsis Med. Garner Emori TG, Horan TC, JS, Jarvis WR, Hughes JM. CDC definitions for nosocomial 1988. infections, 1988;16(3):128-140. Cohen J. The international sepsis Calandra T, on definitions forum consensus conference unit. Crit of infection in the intensive care Med. Jul 2005;33(7):1538-1548. Care Kaukonen KM, Bailey M, Suzuki S, Pilcher D, sepsis to severe Bellomo R. Mortality related and septic shock among critically ill patients in Australia and New Zealand, 2000-2012. JAMA. M, Cooper BS, Palomar-Martinez Wolkewitz M, et al. Multilevel competing risk models infection. nosocomial of risk the evaluate to Apr 8 2014;18(2):R64. Crit Care. Nguile-Makao M, Timsit Coeurjolly JF, Liquet B. Attributable risk estimation JF, multistate models: disability for adjusted application to nosocomial infections. Biom J. Schumacher M, Wangler M, Wolkewitz M, M, M, Wolkewitz Schumacher M, Wangler Beyersmann J. Attributable mortality due to nosocomial infections. A simple and useful application of multistate models. Methods Inf Med. 2007;46(5):595-600. RC. R: A language and environment Team for statistical computing. . R Foundation for from: Statistical Computing. 2013;Available http://www.r-project.org/ A, Bauer P, H, Hiesmayr JM, Savey Burgmann Metnitz B, Metnitz PG. Impact of nosocomial infections on clinical outcome and resource consumption in critically ill patients. Intensive Med. Sep 2010;36(9):1597-1601. Care events. Feasibility and validation. Am J Respirevents. Feasibility Med. Apr 15 2014;189(8):947-955. Crit Care et BS, Cooper JA, Kluytmans AM, Smet de the digestive tract al. Decontamination of in ICU patients. N Engl J and oropharynx Med. Oostdijk EA, Kesecioglu J, Schultz MJ, J, Schultz MJ, Oostdijk EA, Kesecioglu of decontamination of the et al. Effects antibiotic on tract intestinal and oropharynx in ICUs: a randomized clinical resistance trial. 15. 16. 17. 18. 19. 20. 21. 22. 23. 13. 14. Dec 2013;13(12):862-874. Trends Mol Med. Apr 2014;20(4):195- Trends May 2014;2(5):380-386. Angus DC, Linde-Zwirble WT, Lidicker Lidicker WT, Angus DC, Linde-Zwirble J, Pinsky MR. J, Clermont G, Carcillo sepsis in the United Epidemiology of severe outcome, and States: analysis of incidence, Med. Jul Crit Care associated costs of care. 2001;29(7):1303-1310. Sepsis- D. Payen G, Monneret RS, Hotchkiss cellular from induced immunosuppression: Nat Rev dysfunctions to immunotherapy. Immunol. Vincent JL, Marshall JC, Namendys-Silva VincentMarshall JC, Namendys-Silva JL, worldwide Assessment of the SA, et al. critical illness: the Intensive Care of burden audit. Lancet Respir Over Nations (ICON) Med. Leentjens J, Kox M, van der Hoeven JG, Immunotherapy for Netea MG, Pickkers P. of sepsis: from the adjunctive treatment to immunostimulation. immunosuppression for a paradigm change? Am J Respir Crit Time Med. Jun 15 2013;187(12):1287-1293. Care Ayala RS, Hotchkiss J, Unsinger NA, Hutchins A. The new normal: immunomodulatory agents against sepsis immune suppression. Mol Med. Apr 2014;20(4):224-233. Trends K, Chang KC, et al. Boomer JS, To in patients who die of Immunosuppression JAMA. failure. sepsis and multiple organ Dec 21 2011;306(23):2594-2605. Marshall JC. Why have clinical trials in sepsis failed? 203. Vincent JL, Masur H, Opal SM, Dellinger RP, Angus DC. The next generation of sepsis clinical trial designs: what is next after the demise of C?*. Crit human activated protein recombinant Med. Jul 2014;42(7):1714-1721. Care Cavaillon JM, Eisen D, Annane D. Is boosting sepsis appropriate? immune system in the 2014;18(1):216. Crit Care. Vincent J-L. Nosocomial infections in units. The Lancet. adult intensive-care 2003;361(9374):2068-2077. Bos LD, et PM, Ong DS, Klein Klouwenberg of Centers for agreement al. Interobserver criteria for and Prevention Disease Control classifying infections in critically ill patients. Med. Oct 2013;41(10):2373-2378. Crit Care MS, Ong PM, van Mourik Klein Klouwenberg of a novel implementation DS, et al. Electronic surveillance paradigm for ventilator-associated 2. 3. s Reference 1. 4. 5. 6. 7. 8. 9. 10. 11. 12. Crit Care. 2011;15(4):R183. Crit Care. intensive care units in Southern Europe, Turkey Turkey in Southern units Europe, intensive care multicenter point prospective and Iran--a J Infect. Feb 2014;68(2):131- study. prevalence 140. The al. et RA, Claus M, Sossdorf GP, Otto characterized by an late phase of sepsis is death and burden microbiological increased rate. C, Savey A, et al. Lambert ML, Suetens Clinical outcomes of health-care-associated in resistance infections and antimicrobial intensive- European patients admitted to Lancet Infect Dis. a cohort study. units: care Jan 2011;11(1):30-38. R, et al. S, Allard Januel JM, Harbarth mortality due to Estimating attributable in intensive nosocomial infections acquired Hosp Epidemiol. units. Infect Control care Apr 2010;31(4):388-394. 29. 30. 31. Feb 2014;26(1):7-11. Dereli N, Ozayar E, Degerli S, Sahin S, Koc Koc S, Sahin S, Degerli E, Ozayar N, Dereli of nosocomial evaluation Three-year F. of the ICU. Braz J Anesthesiol. infection rates Jan 2013;63(1):73-78. Francais A, et al.Landelle C, Lepape A, septic shock amongNosocomial infection after unit patients. Infect Control intensive care Hosp Epidemiol. Nov 2008;29(11):1054-1065. H, et al. C, Burchardi Alberti C, Brun-Buisson and infection in ICU Epidemiology of sepsis an international multicentre patients from Med. Feb Intensive Care cohort study. 2002;28(2):108-121. Hadzic S, AhmetagicCustovic A, Smajlovic J, N, Hadzagic H. Epidemiological S, Tihic nosocomial infectionssurveillance of bacterial unit. Mater intensive care in the surgical Sociomed. Erdem H, Inan A, Altindis S, et al. Surveillance, Erdem and management of infections in control 24. 25. 26. 27. 28.
ICU-acquired infections in critically ill patients 6 96 ICU-acquired infections in critically ill patients 97 6 P .22 .93 .18 .50 .37 .65 .36 .37 .55 .45 .26 .65 .04 .10 .61 .42 .75 .26 .88 .56 .82 .01 .19 .44 .15 .71 .63 .63 1.00 1.00 .002 .001 1.00 1.00 1487 1630 4 (0.3%) 4 (0.3%) 2 (0.1%) 7 (0.4%) 7 (0.4%) 15 (1.0%) 18 (1.2%) 10 (0.7%) 18 (1.2%) 19 (1.3%) 22 (1.5%) 49 (3.3%) 48 (3.2%) 80 (5.4%) 22 (1.5%) 30 (2.0%) 26 (1.7%) 72 (4.8%) 57 (3.8%) 77 (5.2%) 69 (4.2%) 15 (0.9%) 12 (0.7%) 142 (9.5%) 107 (6.6%) 815 (54.8%) 419 (28.2%) 349 (23.5%) 244 (16.4%) 226 (15.2%) 670 (41.1%) 366 (22.5%) 201 (12.3%) 176 (10.8%) No ICU-AI (n=3,117) - - - - - 232 291 4 (1.7%) 1 (0.4%) 1 (0.4%) 3 (1.3%) 5 (2.2%) 5 (2.2%) 3 (1.3%) 2 (0.9%) 5 (2.2%) 1 (0.3%) 2 (0.7%) 2 (0.7%) 12 (5.2%) 10 (4.3%) 11 (4.7%) 10 (4.3%) 13 (5.6%) 19 (6.5%) 15 (5.2%) 21 (7.2%) 66 (28.4%) 45 (19.4%) 44 (19.0%) 41 (17.7%) 38 (16.4%) 67 (23.0%) 39 (13.4%) 117 (50.4%) 125 (43.0%) ICU-AI (n=523) 1719 1921 4 (0.2%) 4 (0.2%) 9 (0.5%) 2 (0.1%) 9 (0.5%) 19 (1.1%) 19 (1.1%) 11 (0.6%) 21 (1.2%) 31 (1.8%) 27 (1.6%) 59 (3.4%) 59 (3.4%) 85 (4.9%) 22 (1.3%) 33 (1.9%) 28 (1.6%) 82 (4.8%) 62 (3.6%) 90 (5.2%) 84 (4.4%) 15 (0.8%) 13 (0.7%) 126 (6.6%) 932 (54.2%) 485 (28.2%) 394 (22.9%) 288 (16.8%) 267 (15.5%) 180 (10.5%) 795 (41.4%) 433 (22.5%) 222 (11.6%) 215 (11.2%) All (n=3,640) y data CAP HAP Lung abscess Pharyngitis VAP Sinusitis Abdominal sepsis infection Gastrointestinal Bacteremia CRBSI Endocarditis Mediastinitis Myocarditis Brain abscess Primary meningitis Secondary meningitis Spinal Abscess Admission diagnoses and incidence of ICU-acquired infections and incidence of ICU-acquired able S1. Admission diagnoses Sepsis Pulmonary tract T Abdominal tract Cardiovascular Neurological Urinary tract Skin Other Non-infectious disease Cardiovascular Neurologic Respiratory Trauma Gastrointestinal Transplant Genitourinary Metabolic Hematological Musculo-skeletal HAP infection; stream blood related catheter - CRBSI pneumonia; acquired community - CAP - post- unknown; Post-OP UK - infection; ICU-acquired pneumonia; ICU-AI - hospital acquired pneumonia. - ventilator acquired operative; VAP mentar Supple tables Supplementary P .01 .01 .06 .23 .02 .16 .36 .04 .40 .86 .01 .81 .12 .17 .18 .20 .38 .01 .06 .13 .17 .73 .20 .84 .002 1.00 .004 .002 <.001 <.001 8 18 366 420 0 (0%) 0 (0%) 8 (1.9%) 9 (2.1%) 4 (1.0%) 8 (1.9%) 4 (1.0%) 2 (0.5%) 4 (1.0%) 5 (1.2%) 4 (1.0%) 5 (1.2%) 1 (0.2%) 1 (0.2%) 2 (0.5%) 4 (1.0%) 3 (0.7%) 1 (0.2%) 29 (6.9%) 26 (6.2%) 30 (7.1%) 25 (6.0%) 15 (3.6%) 18 (4.3%) 11 (2.6%) 13 (3.1%) 22 (5.2%) 14 (3.2%) 124 (29.5%) 122 (29.0%) 150 (35.7%) Non-infectious 4 7 334 397 0 (0%) Sepsis 1 (0.3%) 5 (1.3%) 9 (2.3%) 4 (1.0%) 1 (0.3%) 9 (2.3%) 6 (1.5%) 3 (0.8%) 1 (0.3%) 1 (0.3%) 5 (1.3%) 7 (1.8%) 7 (1.8%) 9 (2.3%) 4 (1.0%) 4 (1.0%) 20 (5.0%) 21 (5.3%) 16 (4.0%) 30 (7.6%) 13 (3.3%) 32 (8.1%) 13 (3.3%) 32 (8.1%) 15 (3.8%) 49 (12.3%) 40 (10.1%) 89 (22.4%) 93 (23.4%) 151 (38.0%) 15 700 818 22 () 5 (0.6%) 8 (1.0%) 7 (0.9%) 6 (0.7%) 5 (0.6%) 5 (0.6%) 7 (0.8%) 7 (0.9%) 7 (0.9%) 7 (0.9%) 5 (0.6%) 78 (9.5%) 66 (8.1%) 50 (6.1%) 29 (3.5%) 25 (3.1%) 55 (6.7%) 28 (3.4%) 23 (2.8%) 17 (2.1%) 15 (1.8%) 14 (1.7%) 13 (1.6%) 54 (6.6%) 14 (1.7%) 35 (4.3%) 13 (1.6%) 29 (3.5%) 275 (33.6%) 211 (25.8%) 243 (29.7%) All admissions herpes simplex virus Causative pathogens of ICU-acquired infections of ICU-acquired able S2. Causative pathogens T infections ICU-acquired Assigned pathogens Gram-positive bacteria Staphylococcus epidermidis (CNS) Enterococcus faecium Enterococcus Staphylococcus aureus Enterococcus faecalis Enterococcus Enterococcus species Enterococcus Streptococcus pneumoniae Streptococcus Other Gram-positive bacteria Gram-negative bacteria Pseudomonas aeruginosa Escherichia coli Serratia marcescens Klebsiella pneumoniae Enterobacter cloacae Enterobacter Haemophilus influenzae Stenothrophomas maltophilia Stenothrophomas Bacteroides species Bacteroides Morganella species Morganella Citrobacter species Citrobacter Gram negative bacilli Proteus mirabilis Proteus Other viruses Yeast/Fungi Cytomegalovirus reactivation Primary Herpes simplex reactivation Other Gram-positive bacteria Virus Candida albicans Candida glabrata Aspergillus fumigatus Aspergillus Other yeasts or fungi Unknown or other
ICU-acquired infections in critically ill patients 6 98
7 The attributable mortality of delirium in critically ill patients: a prospective cohort study
Peter M.C. Klein Klouwenberg Irene J. Zaal Cristian Spitoni David S.Y. Ong Arendina W. van der Kooi Marc J.M. Bonten Arjen J.C. Slooter Olaf L. Cremer
British Medical Journal Overall, delirium prolongs admission in the intensive care unit but does unit but does admission in the intensive care Overall, delirium prolongs 1112 consecutive adults admitted to an intensive care unit for a minimum unit for a minimum admitted to an intensive care 1112 consecutive adults Trained observers evaluated delirium daily using a validated protocol. protocol. delirium daily using a validated observers evaluated Trained Among 1112 evaluated patients, 558 (50.2%) developed at least one one least at developed (50.2%) patients, 558 1112 evaluated Among not cause death in critically ill patients. Future studies should focus on episodes of studies should focus on episodes of not cause death in critically ill patients. Future persistent delirium and its long term sequelae rather than on acute mortality. episode of delirium, with a median duration of 3 days (interquartile range 2-7 days). 2-7 days). range (interquartile days 3 of duration median a with delirium, of episode with 40/554 patients with delirium compared Crude mortality was 94/558 (17%) in Delirium was significantly associated (7%) in patients without delirium (P<0.001). (odds ratio 1.77, 95% analysis regression with mortality in the multivariable logistic ratio survival analysis (subdistribution hazard confidence interval 1.15 to 2.72) and the association disappeared to 3.09). However, 2.08, 95% confidence interval 1.40 structural model in the marginal after adjustment for time varying confounders ratio 1.19, 95% confidence interval 0.75 to 1.89). Using this (subdistribution hazard only 7.2% (95% confidence interval -7.5% to 19.5%) of deaths in the approach, in excess mortality absolute an with delirium, to attributable were unit care intensive interval -0.9% to 2.3%) by day 30. In patients with delirium of 0.9% (95% confidence remained delirium that persisted for two days or more post hoc analyses, however, interval 1.2% to 2.8%) absolute mortality associated with a 2.0% (95% confidence competing risk analysis showed that delirium of any duration Furthermore, increase. the intensive care from reduced rate of discharge was associated with a significantly ratio 0.65, 95% confidence interval 0.55 to 0.76). unit (cause specific hazard Conclusions: Logistic regression and competing risks survival analyses were used to adjust for for adjust to used were analyses survival risks competing and regression Logistic for confounding structural model analysis to adjust a marginal baseline variables and the onset of delirium. by evolution of disease severity before unit. care Mortality during admission to an intensive Main outcome measures: Results: Abstract determine in critically ill caused by delirium mortality the attributable To Objective: patients. cohort study. Design: Prospective 2013. in the Netherlands, January 2011 to July unit care Setting: 32 mixed bed intensive Participants: of 24 hours. Exposures:
Attributable mortality of delirium in the ICU 7 102 Attributable mortality of delirium in the ICU 103 7 , several others several others , 6-14 . All relevant information relevant All . 21 . These inconsistencies have been explained by have been explained by . These inconsistencies . Although most studies have identified delirium have identified delirium . Although most studies 15-17 1-5 . To aid in the interpretation of our findings, we compared the results of theresults the we compared of our findings, aid in the interpretation . To . Deficiencies in modelling methodology and residual confounding may, may, residual confounding modelling methodology and . Deficiencies in 18 19,20 We estimated the proportion of deaths that can be attributed to delirium in a large to delirium in a large of deaths that can be attributed the proportion estimated We found no association with mortality found no association Delirium to this study used a validated flowchart to classify the mental team dedicated A research care intensive from discharge until daily patients of status Study population to hours 24 least at for admitted adults consecutive evaluated prospectively We the Utrecht, unit of the University Medical Centre the 32 mixed bed intensive care acute with patients excluded We 2013. June and 2011 January between Netherlands, assessments of delirium disease at baseline, those in whom neurological or premorbid or from and those transferred could not be performed owing to a language barrier, for an gave approval board unit. The local ethical review to another intensive care 10-056/12-421) whereby number board review opt-out consent method (institutional that was notified of the study by a brochure participants and family members were unit with an attached opt-out card. at admission to the intensive care provided Methods uction Introd patients of 30-60% in occurring illness, critical of complication common a is Delirium unit care admitted to an intensive differences in case mix, the tools used for the assessment of delirium, and the study study for the assessment of delirium, and the in case mix, the tools used differences design was available to the study team, including the 12 hourly confusion assessment methodconfusion assessment hourly 12 including the team, study the available to was categorised patients as comatose, unit conducted by nurses. We for the intensive care we assessed the level sedated, awake and delirious, or awake and non-delirious. Firstly, maximumof consciousness using the Richmond agitation-sedation scale. Patients with 24 hour observation period could not be assessed of -5 or -4 during the entire scores as an independent predictor of death in the intensive care unit intensive care the of death in an independent predictor as marginal structural model analysis with those of standard statistical regression methods. regression statistical model analysis with those of standard structural marginal however, provide an alternative explanation. In particular, none of the previous studies studies none of the previous an alternative explanation. In particular, provide however, or for the start of delirium, before disease progression adjusted for adequately have in the observation of mortality that may preclude as discharge) competing events (such a marker of is merely unclear whether delirium remains unit. It therefore intensive care unit. to mortality in the intensive care or causally linked poor prognosis structural model analysis from cohort of critically ill patients by performing a marginal the from bias that results Such analysis can overcome the discipline of causal inference. traditional sources as more well as severity until the onset of delirium disease of evolution of bias . 40 . A sedated state was. A sedated 22 . These are all time fixed variables, representing representing all time fixed variables, . These are 23 . Several physiological and laboratory variables (temperature, sodium, sodium, . Several physiological and laboratory variables (temperature, 1,6,7,15,17,24-38 No data were missing for baseline variables, daily mental status classifications, or missing for baseline variables, daily mental status classifications, or No data were defined as propofol continuously administered at a rate of >1 mg/kg/h and/or midazolama rate of >1 mg/kg/h at continuously administered propofol defined as or at any time in thetime of assessment either at the >50 mg/d or equivalent at a dose of Richmond of -5 or -4 on the All other patients with scores assessment. 48 hours before remaining patients We assessed the as comatose. classified were agitation-sedation scale unit as well as intensive care confusion assessment method for the for delirium using the were team. These patients notes and nursing charts by the research inspection of medical confusion assessment method when they tested positive on the classified as delirious the level was a description of fluctuation in when there unit and/or in the intensive care because Furthermore, agitation, disorientation, or hallucinations. of consciousness, delirium duringof treatment the for used exclusively were quetiapine and haloperidol the day of initiation of either also classified patients as delirious on the study period, we (AS) was consulted, who cast the decisive of these drugs. In case of doubt, a neurologist a sensitivity of 0.75 (95% had This procedure vote for classification of mental status. of 0.85 (95% confidence interval 0.68 to 0.94),confidence interval 0.47 to 0.92), specificity enable our (Fleiss’ κ 0.94) (unpublished data). To agreement and an excellent inter-rater aspatients sedated reclassifying by status mental the dichotomised we analysis primary sedation) as delirious (see Supplementarynon-delirious and comatose patients (without of the of the results for the patients was unaware S1). The clinical team responsible Figure team. delirium assessments made by the study urea, and haemoglobin) were transformed to account for their non-linear relation with with transformed to account for their non-linear relation and haemoglobin) were urea, health the acute physiology and chronic from using the cut-offs delirium or mortality, prospectively data collected this study to dedicated model 39. Observers IV evaluation checked for data integrity cohort study and regularly as part of a large the risk of delirium at baseline. However, because the risk of delirium onset is likely to to because the risk of delirium onset is likely the risk of delirium at baseline. However, unit depending on the evolution intensive care vary over the course of admission to an we also incorporated time dependent variables in our primary of disease severity, failure of the sequential organ 1). These included daily measurements analysis (Figure status, use mechanical ventilation temperature, sepsis status, core assessment score, acidosis, and haematocrit of sedative and analgesic drugs, and plasma sodium, urea, levels Covariables and outcome for covariables that we chose a priori based In all multivariable models we adjusted consideration and mortality after careful on their expected associations with delirium These covariables included age, sex, history of dementia, history of the literature. health index, acute physiology and chronic of alcohol misuse, Charlson comorbidity of sepsis on and presence status, admission type, readmission evaluation IV score, unit admission to the intensive care for delirium and were classified as either comatose or sedated comatose or sedated classified as either and were for delirium the outcome. However, for daily observations of several laboratory and physiological physiological and laboratory several of observations daily for However, outcome. the
Attributable mortality of delirium in the ICU 7 104 Attributable mortality of delirium in the ICU 105 7
. Mortality during during . Mortality B A 41 15 Day on ICU Day
10 5 Onset Onset of delirium 0
. Furthermore, adjustment for time dependent covariables using using for time dependent covariables adjustment . Furthermore, disease of Severity 18 . In this type of analysis, however, informative censoring of the survival time . In this type of analysis, however, Evolution of disease severity prior to onset of delirium in two hypothetical patients. Evolution of disease severity prior to onset of delirium in two hypothetical patients. 42 Figure 1: Figure hypothetical patients admitted to two in changes in the severity of disease the shows This figure Both patients have similar disease severities at admission, but the unit (ICU). the intensive care in develops As delirium preferentially condition of patient A worsens and of patient B improves. ill patients confounding occurs when disease severity after baseline is not adjusted severely more and survival analysis adjusts for baseline variables at t=0 only. for in the analysis. Logistic regression for changes in disease severity until the onset of delirium (dark structural model adjusts A marginal (light grey). but not thereafter grey), Statistical analysis in our cohort, obtain first estimates of the association between delirium and mortality To we performed a multivariable literature, with previous our results and to be able to compare baseline confounders. To analysis, adjusting for a priori selected logistic regression at we assumed that patients who develop delirium are comply with existing literature, unit, even if delirium risk for the duration of their stay in the intensive care increased by using bias can be overcome The resulting develops only several days after admission. as a time dependent analysis and with inclusion of delirium hazards a Cox proportional variable variables, 3.1% of data were missing overall (range 0-6.9% for individual variables). for individual variables). overall (range 0-6.9% missing 3.1% of data were variables, for each observation day each before longitudinal data the availability of Because of covariables for missing imputation performedpatient, we a trend marginal structural models requires daily information and about severity of disease structural models requires marginal the this precludes Practically, for the duration of observation. therapeutic interventions unit. the intensive care from lie beyond discharge use of outcomes that admission to an intensive care unit was the primary outcome of interest in all analyses. in all analyses. unit was the primary outcome of interest care admission to an intensive study we the present known to exist, in Although long term sequelae of delirium are authors have claimed on short termdeliberately focused previous outcomes because unit even after correction in the intensive care mortality rate increased up to a threefold for confounders should additionally be taken into account by considering discharge as a competing risk by considering discharge should additionally be taken into account in unit alive are the intensive care from discharged because patients who are for mortality, . A 43,44 . We adjusted adjusted . We 47 . For instance, severely agitated patients . For instance, severely 46 . A marginal structural model analysis deals with these . A marginal 25 . It enables assessment of what the mortality in the intensive care unit in the intensive care . It enables assessment of what the mortality , bias occurs when such changes in disease severity are not adjusted for not adjusted severity are disease in changes such when occurs bias , 25 19,20 To accomplish such a counterfactual analysis, we performed two steps. Firstly, we we accomplish such a counterfactual analysis, we performed two steps. Firstly, To Supplementary Figure S1). performed several post hoc sensitivity analyses (see Supplementary Figure We . The subdistribution hazard ratio is therefore a summary measure of all separate cause measure a summary is therefore ratio hazard . The subdistribution would have been in a hypothetical population in which all patients remained delirium-free, delirium-free, in which all patients remained would have been in a hypothetical population called a counterfactual analysis. and is therefore unit, using a of acquiring delirium in the intensive care modelled the daily probability baseline and daily patient analysis that included both multivariable logistic regression we calculated stabilised daily probabilities, characteristics. Based on these estimated the represent weights) that probability (so called inversed patient specific weights time for adjustment Because for each patient. acquiring delirium of risk cumulative we used in bias, result delirium may the start of after varying variables measured delirium on each day day to predict the preceding lagged values from for lagged values of the sequential organ failure assessment scores two days before two days before assessment scores failure for lagged values of the sequential organ of delirium the onset within 24 hours before measured to acknowledge that the scores we performed an may have been influenced by an insidious onset of delirium. Secondly, (death analysis with competing endpoints weighted Cox regression inverse probability death. of risk cumulative and hazard daily the both estimated and alive) discharge and the population attributable we computed of the results, aid in the interpretation To delirium. of patients who have died from fraction, which indicates the percentage these reclassified we non-delirious, sedated patients as categorising of instead Firstly, 45 a different health state from patients who remain admitted beyond that time point beyond that time admitted remain patients who state from health a different during the analysis. Secondly, bias might occur when a time dependent covariable is notis covariable time dependent when a occur might bias Secondly, during the analysis. subsequent delirium, and when delirium status only a risk factor for death but also predicts the risk factor time point predicts at a previous specific hazards and can be used to calculate the cumulative incidence of the outcome to calculate the cumulative incidence and can be used specific hazards adjust for baseline study). Although the methods can (that is, death in this of interest of delirium onset and informative censoring caused varying nature confounders, the time of these methods is that neither unit, a limitation the intensive care from by discharge of severity the Firstly, confounding. of sources important potentially other, for adjust can of representative unit may not be of admission to the intensive care disease on the day which typically occurs later on during stay inthe health state at the time of delirium onset, develops in patients who are 1). As delirium preferentially unit (Figure the intensive care ill severely more competing risks analysis provides two measures of association: the cause specific hazard specific hazard association: the cause of two measures risks analysis provides competing (both intensiveof delirium on outcome effects the direct in this case estimates ratio, which ratio, which describes the subdistribution hazard and death), and the unit discharge care delirium died from delirium given that the patient has not dying from instantaneous risk of with delirium may eventually be treated with sedatives, whereas sedative use itself is a with sedatives, whereas with delirium may eventually be treated known risk factor for delirium limitations by adjusting for the changes in disease severity before delirium onset, while in disease severity before limitations by adjusting for the changes bias preventing
Attributable mortality of delirium in the ICU 7 106 Attributable mortality of delirium in the ICU 107 7 ). The average ). The average Figure 2 Figure , and we used bootstrapping , and we used bootstrapping ICU 49 - 797 Exclusion: - 783 Admitted <24hrs -162 Transferred from or to other 399 neurological 194 stroke 74 elective neurosurgery 28 traumatic brain injury 19 seizure 84 other neurological disease 195 cardiac arrest 203 other for the marginal structural model analysis. P analysis. P structural model for the marginal 48 2854 1112 2692 1909 Admitted s Flowchart of patient inclusion. “Other neurological disease” includes patients with disease” includes patients with Flowchart of patient inclusion. “Other neurological lt su All analyses were performed using SAS 9.2 (Cary, NC) and R 2.14 (www.r-project. performed using SAS 9.2 (Cary, All analyses were e resulting from the marginal structural model analyses structural model analyses the marginal from resulting values less than 0.05 were considered to be statistically significant. We used robust robust We used be statistically significant. to considered were values less than 0.05 ratios for the hazard sandwich) to calculate confidence intervals estimators (Huber encephalitis, encephalopathy, coma, or hydrocephalus. “Other” includes patients with premorbid patients with premorbid “Other” includes coma, or hydrocephalus. encephalitis, encephalopathy, in whom delirium assessments could not be made due to for conditions or patients neurological mental retardation. instance a language barrier or severe Figure 2: Figure intensive care unit of whom 1112 met the inclusion criteria ( intensive care R admitted to our critically ill patients were During the 2.5 year study period, 2854 by the underlying condition, we performed subgroup analyses in patients with sepsis in patients with sepsis analyses we performedby the underlying condition, subgroup health evaluation IV score. physiology and chronic and stratified by acute only, used the R-package “IPW” We org). patients based on the first available valid assessment for delirium after the cessation of of after the cessation for delirium valid assessment on the first available patients based of definition rigorous we applied a more imputation. Secondly, using backward sedation, been classified as only when they had as being delirious considering patients delirium by modification to assess possible effect two consecutive days. Thirdly, delirious on at least to estimate the confidence intervals for the attributable mortality. intervals for the attributable to estimate the confidence value 0.54 0.26 0.02 0.29 <.001 <.001 <.001 <.001 <.001 <.001 <.001 <.001 <.001 P shows able 1 shows 33 (6%) 46 (9%) 9 (5-18) (N=558) 91 (16%) 94 (17%) 266 (48%) 142 (25%) 150 (27%) 356 (64%) 211 (38%) 170 (30%) 131 (23%) 534 (96%) 306 (55%) 510 (91%) 64 (54-74) 79 (62-97) 7.1 (1.0-11.4) ver delirium Ever delirium 3 (2-5) 12 (2%) 40 (7%) (N=554) 83 (15%) 71 (13%) 253 (46%) 165 (30%) 136 (24%) 316 (57%) 213 (38%) 165 (30%) 105 (19%) 535 (97%) 190 (34%) 470 (85%) 61 (49-69) 63 (48-81) 5.4 (0.0-10.2) ever delirium Never delirium a b medical scheduled surgery surgery emergency general surgery surgery and cardiothoracic cardiology internal medicine other Defined as prior ICU admission during current hospitalization. Defined as prior ICU admission during current able 1: Patient characteristics by delirium status Defined as alcohol consumption of > 40 gram alcohol/day. a b T Characteristic Age (years) Abbreviations: APACHE, Acute Physiology and Chronic Health Evaluation; ICU, intensive care unit. Health Evaluation; ICU, intensive care Acute Physiology and Chronic APACHE, Abbreviations: Admission type Male gender Charlson comorbidity index Prior ICU admission Medical specialty Caucasian race Sepsis at admission Current alcohol abuse Current Mechanical ventilation at admission APACHE IV score APACHE Length of stay (days) Continuous variables are expressed as medians (inter-quartile range); categorical variables as range); categorical variables as as medians (inter-quartile expressed Continuous variables are absolute numbers (%). ICU case fatality the patients’ characteristics by delirium status. Patients with delirium had significantly delirium had significantly by delirium status. Patients with the patients’ characteristics likely to be male. more and were older, admission, were severity of disease on greater with 40/554 compared was 94/558 (17%) in patients with delirium The crude mortality delirium (P<0.001). (7%) in those without length of stay was 8.9 days, amounting to a total of 9867 observation days. Delirium days. Delirium of 9867 observation amounting to a total stay was 8.9 days, length of dichotomous to obtain a However, of these days. on 2524 (26%) was observed 537 reclassified analyses, we additionally use in the regression variable for exposure to days without delirium. with delirium and 1371 sedation days coma days to days time median The delirium. of episode one least (50%) patients had at 558 this, After and the median range 1.0-4.0 days) was 2.0 days (interquartile to onset of delirium T range 2.0-7.0) days. episode was 3.0 (interquartile duration of a delirium
Attributable mortality of delirium in the ICU 7 108 Attributable mortality of delirium in the ICU 109 7 h g e f
Marginal Marginal 1.19 (0.75-1.89) structural model 3.14 (2.32-5.04)
3.14 (2.32-5.04) 2.08 (1.40-3.09) Competing risks survival regression survival regression
. ogistic Logistic 19 regression regression 2.60 (1.76-3.85) 1.77 (1.15-2.72)
a
b c d Crude Adjusted and discharge Evolution of disease prior to delirium onset Baseline covariables onset of delirium Time-varying Competing risks of death The inversed probability weight estimates were: mean 0.974 (range 0.127-8.51) and median mean 0.974 (range 0.127-8.51) and median estimates were: weight The inversed probability The logistic regression analysis gives an odds ratio whereas the survival analysis and marginal the survival analysis and marginal analysis gives an odds ratio whereas The logistic regression Logistic regression and survival analysis can also be used to correct for evolution of disease of disease for evolution and survival analysis can also be used to correct Logistic regression The crude subdistribution hazard ratio of the marginal structural model was calculated assuming structural model was calculated assuming ratio of the marginal The crude subdistribution hazard Effect estimates for the association between delirium and ICU mortality using various and ICU mortality using various for the association between delirium estimates able 2: Effect The adjusted cause-specific hazard ratios of the marginal structural model analysis were 0.38 0.38 structural model analysis were ratios of the marginal The adjusted cause-specific hazard The multivariable analysis was adjusted for baseline variables (age, gender, Charlson co-morbidity variables (age, gender, The multivariable analysis was adjusted for baseline shows the results of the regression models. Delirium was significantly associated associated Delirium was significantly models. regression of the results able 2 shows the Delirium was included as a time-dependent variable in the competing risks survival regression variable in the competing risks survival regression Delirium was included as a time-dependent The adjusted cause-specific hazard ratios of the competing risks survival regression were 0.64 0.64 were regression ratios of the competing risks survival The adjusted cause-specific hazard (0.22-0.65) and 0.65 (0.55-0.76) for mortality and discharge, respectively. (0.22-0.65) and 0.65 (0.55-0.76) for mortality and discharge, 0.894 (IQR 0.731-1.072). the weights to be equal to 1 and is therefore equal to the estimation of the competing risks equal to the estimation of the competing risks the weights to be equal to 1 and is therefore survival analysis. a Effect estimate Effect h e g b c d T statistical approaches f index, Acute Physiology and Chronic Health Evaluation IV score, admission type, and sepsis on Health Evaluation IV score, index, Acute Physiology and Chronic adjusted to time-varying variables: model was furthermore structural admission). The marginal concentration, sodium, urea sepsis status, temperature, Assessment score, Failure Sequential Organ acidosis, haematocrit, mechanical ventilation and sedative and analgesic medications. Adjustment for: (0.39-1.03) and 0.53 (0.46-0.61) for mortality and discharge, respectively. (0.39-1.03) and 0.53 (0.46-0.61) for mortality and discharge, structural model provide a subdistribution hazard ratio. ratio. a subdistribution hazard structural model provide and marginal structural models. and marginal severity, however over-adjustment and collider-stratification bias might occur. The marginal The marginal bias might occur. and collider-stratification however over-adjustment severity, these biases structural model prevents egression analyses Regression T 95% confidence odds ratio 2.60, analysis (crude by logistic regression with mortality interval 1.15 to 2.72). adjusted odds ratio 1.77, 95% confidence interval 1.76 to 3.85; direct no had delirium however, analysis, survival specific cause dependent, time In confidence 95% 0.64, ratio hazard specific (cause death of risk daily the on effect of being discharged in a lower daily probability but did result interval 0.39 to 1.03) ratio 0.53, 95% confidence interval hazard unit (cause specific the intensive care from for unit remained the risk of dying in the intensive care 0.46 to 0.61). Consequently, 30 20
Day on ICU on Day 10 observed ICU mortalityICUobserved delirium without mortality ICU estimated provides the results of the post hoc analyses using analyses using of the post hoc the results able S1 provides 0
8 6 4 2 0
14 12 10 ICU mortality (%) mortality ICU
Supplementary T Cumulative incidence of observed and estimated ICU mortality. This figure represents represents This figure 3: Cumulative incidence of observed and estimated ICU mortality. Figure in the the expected mortality in the whole cohort estimated by the cumulative incidence function adjusted for informative risks censoring using a competing We of delirium. absence and presence are structural method. The exact procedures analysis and for evolution of disease using a marginal explained in the methods section. longer in patients with delirium, resulting in a combined hazard of death for patients death for patients of hazard in a combined resulting patients with delirium, longer in ratio 2.08, 95% hazard (subdistribution increased that was significantly with delirium of for the evolution once we adjusted 3.09). In contrast, interval 1.40 to confidence analysis, structural model of delirium in the marginal the onset disease severity before (subdistribution unit care associated with death in the intensive delirium was no longer 3 shows the effect interval 0.75 to 1.89). Figure ratio 1.19, 95% confidence hazard By day 30, the population cumulative risk of death over time. of delirium on the unit was 7.2% intensive care of mortality due to delirium in the attributable fraction case fatality to an absolute -7.5% to 19.5%), corresponding (95% confidence interval interval -0.9% to 2.3%). of 0.9% (95% confidence imputation alternative analysis using backward for delirium. The sensitivity definitions by the use of sedation obscured were for patients in whom delirium assessments the sensitivity analysis using a delirium definition yielded similar estimations. However, consecutive days yielded a the derangement to persist for two or more that required association with mortality than did our primary analysis (subdistribution overall stronger 1.67, 95% confidence interval 1.13 to 2.47; 30 day absolute risk difference ratio hazard 1.2% to 2.8%). In this case, cause specific in mortality 2.0%, 95% confidence interval a prolonged mortality was mediated through analysis also showed that increased risk daily the on effect direct a than by rather unit care intensive in the stay of length of patients who subgroup found in the modification were of death. No signs of effect
Attributable mortality of delirium in the ICU 7 110 Attributable mortality of delirium in the ICU 111 7 . However, results results . However, 52 . able S2). 15,50,54,55 . However, as we added complexity to our as we added complexity to our . However, 6-8, 10, 11, 14-16, 42, 50, 51 . Furthermore, delirium was not associated with mortality even in a prespecified with mortality even in a prespecified delirium was not associated . Furthermore, 53 of post hoc sensitivity analyses suggest that patients who develop an episode of that patients who develop an episode of of post hoc sensitivity analyses suggest risk an overall increased exposed to than two days are delirium that persists for more unit. Although this finding needs further confirmation,it of death in the intensive care distinction between rapidly suggesting a fundamental reports gives support to recent persistent delirium in the intensive care delirium and severe, sedation related reversible, unit subgroup of patients with sepsis. Although sepsis is often associated with delirium, it is of patients with sepsis. Although sepsis is often associated with delirium, it is subgroup entity of delirium. encephalopathy” is a different unknown whether this so called “septic indicate that sepsis associated delirium is similar to other forms of delirium Our results indicated that all cause specific analyses case, however, In any mortality. regarding overall in increase any that rather but rates fatality case affect directly not does delirium intensive the in stay of length protracted more a through mediated be to seems mortality unit, exposing patients for longer to a fixed daily risk of dying (for example, due to care sedation, or mechanical ventilation, and nosocomial infections, drawbacks of prolonged unit). Indeed, the average length other “general” complications in the intensive care with 8.8 for patients with a of stay of patients without delirium was 4.0 days compared The effect short episode of delirium and 16.5 days for patients with a persistent episode. typically of delirium on the length of stay is plausible because patients with delirium are (hampering early mobilisation), may have an less likely to interact with their environment of catheters or tubes), incidence of complications (for example, self removal increased drugs with sedative effects and often receive Comparison with other studies only our findings confirmWhen we adjusted for baseline variables of the estimates with delirium that were case fatality rate associated increased a twofold to threefold studies previous in reported We estimated the mortality due to delirium in critically ill patients in intensive care care intensive in patients ill critically in delirium to due mortality the estimated We disease severity until the account bias caused by time varying while taking into the this approach, Using by the competing risk of discharge. onset of delirium and at 7.2% by unit was estimated mortality in the intensive care population attributable 0.9% if than more by no absolute case fatality can be reduced day 30, implying that in all patients. delirium able to completely prevent we were presented with sepsis, nor in relation to the severity of illness at the time of admission time of admission of illness at the to the severity nor in relation with sepsis, presented T (see Supplementary unit care to the intensive Discussion analyses and adjusted for changes in disease severity before the onset of delirium, we the onset of delirium, we severity before analyses and adjusted for changes in disease (of any duration) and mortality in the intensive found no association between delirium of clinical trials, which also meta-analysis which is consistent with a recent unit, care with short termshowed that delirium was not associated mortality . 7,42 . The marginal . The marginal . This dissimilarity . This dissimilarity 47,56 59 . When this happens, any mortality . When this happens, any mortality 57,58 . To avoid immortal time bias (that is, bias due to the time varying nature varying nature time bias (that is, bias due to the time avoid immortal . To 6,8,10 We acknowledge some limitations of our study design. Firstly, management management some limitations of our study design. Firstly, acknowledge We In addition to these methodological concerns, an important problem also remains also remains concerns,addition to these methodological In problem important an due to this sepsis event may then be attributed falsely to delirium, since a marginal attributed falsely to delirium, since a marginal due to this sepsis event may then be adjusts for the evolution of disease severity structural model analysis (deliberately) association of delirium with mortality might up until the onset of delirium. The true structural model, we by using a marginal Finally, be even lower than we report. therefore unit that can be expected to an intensive care estimated the mortality during admission competing risks ordinary population. Since for the entire if delirium would be prevented used in model, the population a conditional rather than a marginal is survival regression such unweighted analyses from structural model differs a marginal structural model analysis that we used is not affected by these problems. is not affected structural model analysis that we used the limit may which centres, between vary may and pragmatic largely is delirium of a using performed studies centres in particular, In generalisability of our findings. ours may yield disparate of delirium from or treatment strategy for prevention different of our initial case mix and the results both our Yet, of attributable mortality. estimates Secondly, literature. in line with previous analyses were logistic and survival regression rule out the possibility that unobserved studies, we cannot as is true for all observational number large even after accounting for a relatively confounding might have occurred, of an impending complication, delirium can be the first symptom of covariables. Thirdly, unit in the intensive care such as sepsis acquired may partly explain the observed differences in effect estimates yielded by the two by the two estimates yielded in effect may partly explain the observed differences of these models useful to comparative reporting we considered However, approaches. in many commonly used analyses. present illustrate the bias that is inherently the with deals domain) this in study any matter that for (or study our way the about assessment classification of unobserved days due to coma or sedation. Because the patients is impossible and the statistical methods that of delirium in unresponsive patient recategorised dichotomous classification of patients, we a we used required However, we considered it crucial to also incorporate the evolution of disease severity severity disease of evolution the incorporate also to crucial it considered we However, critically ill patients delirium episode into our analyses as the development of a before are which of both dysfunction, organ of reversal or deterioration rapid either show may over time. Although standard impact on the risk of delirium onset likely to considerably in the elimination their use may result models can be used for this purpose, regression also is confounder varying time a if mortality on delirium of effects potential any of an intermediate delirium, as well as leads up to factor in the causal pathway that when disease severity on a given day is bias that occurs to the collider-stratification in time point of delirium at a previous influenced by the presence Strengths and limitations of this study and limitations Strengths only in a few but with mortality, association of delirium have tackled the Many authors been performed multivariable analyses studies have at for severity of disease to adjust baseline of delirium), others have incorporated the time of delirium onset in their analyses their in onset delirium of time the incorporated have others delirium), of
Attributable mortality of delirium in the ICU 7 112 Attributable mortality of delirium in the ICU 113 7 , and , and 65 . However, in in . However, 60,61 , generates major , generates major 63 . These studies might, however, have been prone have been prone . These studies might, however, 42 . Nevertheless, the absolute risk increase of 2.0% after the onset of of 2.0% after the onset of . Nevertheless, the absolute risk increase 5,66-71 . Furthermore, we assumed that most patients with true delirium would would delirium true with most patients that assumed we Furthermore, . 53 . Delirium is distressing to both patients and their relatives to both patients and their relatives . Delirium is distressing 62 , may have severe long term such as cognitive impairment consequences , may have severe 64 costs contrast with these authors, we categorised patients who were unresponsive owing owing unresponsive patients who were categorised we with these authors, contrast sedative a stopping of hours 48 within were who those (or sedation continuous to otherwise would introduce because we believed that doing infusion) as non-delirious hours of sedated for at least the first who often remain patients bias against surgical stay their be detected after cessation of sedation after all, which would result in only a minor in only a minor after all, which would result be detected after cessation of sedation of risking a major error to the timing of delirium onset (rather than respect with error sedation for an receiving in fact they were misclassifying patients as delirious when we dealt with any possible In our sensitivity analyses but legitimate reason). unrelated by patients sedated in onset delirium of timing to respect with error misclassification similar. results were and the approach, backward using a first valid observation carried could cause long term mortality Clinical implications between delirium episodes of short duration Although we did not find an association an important remains unit, delirium mortality in the intensive care and increased whenever or treated in critically ill patients and should be prevented clinical syndrome possible days on which delirium could not be observed due to coma or sedation, or both. No or both. No due to coma or sedation, not be observed delirium could days on which previous and in fact many exist for such reallocation accepted methods universally chose to categorise these days. We how they dealt with not describe exactly authors did of ≤3 on the Richmond agitation-sedation (score comatose patients who remained because we after cessation of sedation as delirious, than 48 hours scale) for more that forms a continuum failure” “brain severe a state of believe that coma represents studies is similar to that of previous approach with delirium. This to the same bias—that is, failing to adjust for the evolution of disease severity before for the evolution of disease severity before to the same bias—that is, failing to adjust of episodes that indicated analyses sensitivity importantly, Most delirium. of onset the risk of death in the delirium may still be associated with an increased persistent severe, unit by day 30, despite our overall finding of no association with delirium intensive care incompletely for this association are of any duration. The possible causal mechanisms of delirium causing as a result understood, but may include autonomic dysregulation causing increased immunomodulatory effects failure, hypotension and subsequent organ production increased causing responses stress excessive and infections, to susceptibility of corticosteroid persistent delirium would translate into a number needed to treat of more than 50, even than 50, even of more persistent delirium would translate into a number needed to treat of intervention. this complication by some sort prevent able to effectively if we were since this finding was based on post hoc analyses, these observations Furthermore, need to be confirmed of critically ill patients. in other cohorts Intensive Intensive Crit Care Med 2004;32:2254-9. Crit Care Intensive Care Med 2013;39:481-8. Intensive Care care unit environment may affect the course of may affect unit environment care delirium. Riker RR, Shehabi Y, Bokesch PM, PM, Bokesch Riker RR, Shehabi Y, et Koura F, Ceraso D, Wisemandle W, al. Dexmedetomidine vs midazolam for sedation of critically ill patients: a randomized trial. JAMA 2009;301:489-99. Gottfried SB, Ouimet S, Kavanagh BP, Incidence, risk factors and Y. Skrobik delirium. ICU of consequences Med 2007;33:66-73. Care T, Speroff B, Truman A, Shintani EW, Ely et al. Delirium FE Jr, SM, Harrell Gordon mechanically of mortality in as a predictor unit. ventilated patients in the intensive care JAMA 2004;291:1753-62. CH, Lin HC, Huang Wang Lin SM, Liu CY, et al. The impact of delirium CD, Huang PY, on the survival of mechanically ventilated patients. 5. 6. 7. 8. Intensive Care Care Intensive
Van den Boogaard M, Pickkers P, Slooter Slooter M, Pickkers P, den Boogaard Van PE, van der Voort AJ, Kuiper MA, Spronk PH, et al. Development and validation of in DELIRium of (PREdiction PRE-DELIRIC model ICu patients) delirium prediction patients: observational for intensive care BMJ 2012;344:e420. study. multicentre L, Kadiman S, Alias A, Ismail Chan Shehabi Y, MA, et al. Sedation depth and long- WN, Tan term ventilated mechanically in mortality longitudinal critically ill adults: a prospective study. cohort multicentre Med 2013;39:910-8. Videbech I, P, Svenningsen H, Egerod Tonnesen M, Christensen D, Frydenberg EK. Fluctuations in sedation levels may contribute to delirium in ICU patients. Acta Anaesthesiol Scand 2013;57:288-93. Peelen LM, van Eijk MM, Zaal IJ, Spruyt CF, Wientjes R, Schneider MM, et al. Intensive References 1. 2. 3. 4. This work was supported by the Centre for Translational Molecular Medicine (http:// for Translational This work was supported by the Centre funding from research MARS (grant 04I-201). MB has received project www.ctmm.nl), (NWO Vici 918.76.611). The of Scientific Research the Netherlands Organization in study design; nor in the collection, analysis, study sponsors did not have a role and nor in the decision to of data; nor in the writing of the report; and interpretation submit the article for publication. We would like to thank Rolf Groenwold, Jos Frencken and Margaret Nicholls for their for their Nicholls and Margaret Jos Frencken would like to thank Rolf Groenwold, We invaluable statistical assistance. excellent advice and Sources of support ments Acknowledge usions Concl critically mortality in delirium associated the first study to estimate knowledge, this is our To for variations in correction counterfactual analysis that incorporates ill patients using a the absolute attributable of delirium. Using this approach, the onset disease severity before unit (of anyshort term in the intensive care associated with a delirium episode mortality studies should focus on suggested. Future previously than lower duration) was much delirium and its long termepisodes of persistent rather than on acute mortality. sequelae
Attributable mortality of delirium in the ICU 7 114 Attributable mortality of delirium in the ICU 115 7
2011;15(Suppl 1):P335. Crit Care 2009;13:R77. Crit Care 2008;17:567-75. Am J Crit Care Lifetime Data Anal 2010;16:45-70. Lifetime Data Bekaert M, Vansteelandt S, Mertens K. S, Mertens K. Vansteelandt Bekaert M, in confounding time-varying for Adjusting analysis of a competing the subdistribution risk. Dijk D, Slooter AJ. Zaal IJ, Peelen L, Van of an eight- Development and validation on the CAM-ICU: step flowchart based tool to highly adaptable and quick a of delirium in ICU determine the presence patients. Grap MJ, Brophy Sessler CN, Gosnell MS, Keane KA, et al. The GM, O’Neal PV, scale: validity Richmond agitation-sedation in adult intensive care and reliability Med unit patients. Am J Respir Crit Care 2002;166:1338-44. Zaal IJ, Slooter AJ. Delirium in critically ill pathophysiology, patients: epidemiology, diagnosis and management. Drugs 2012;72:1457-71. Rompaey B, Elseviers MM, Schuurmans Van S, MJ, Shortridge-Baggett LM, Truijen Bossaert L. Risk factors for delirium in cohort patients: a prospective intensive care study. Shintani A, Peterson J, Pun Pandharipande P, Wilkinson GR, Dittus RS, et al. Lorazepam BT, is an independent risk factor for transitioning unit patients. to delirium in intensive care Anesthesiology 2006;104:21-6. PE, van Eijk MM, van den Boogaard Spronk M, van der Jagt M, Kuiper MA, Smit EGM. 2010, NVIC Richtlijn op de Intensive Care 2010. O’Neill PJ, Cotton BA, Pun BT, Agarwal V, and et al. Prevalence Thompson J, Haney S, risk factors for development of delirium in unit patients. J Burnburn intensive care Res 2010;31:706-15. Care Pisani MA, Murphy TE, Araujo KL, Slattum Ness PH, Inouye SK. Benzodiazepine Van P, and opioid use and the duration of intensive population. unit delirium in an older care Med 2009;37:177-83. Crit Care Ness Pisani MA, Murphy TE, Araujo KL, Van PH. Factors associated with persistent admission unit delirium after intensive care in an older medical patient population. J 2010;25:540 e1-7. Crit Care Lin PJ, Chen MC, Liu CY. YF, Chang YL, Tsai and risk factors for postoperative Prevalence care intensive delirium in a cardiovascular unit. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30.
2008;23:372-9. Care Crit J Epidemiology 2000;11:561-70. Salluh JI, Soares M, Teles JM, Ceraso JM, Ceraso M, Teles Salluh JI, Soares Delirium al. et VS, Nava N, Raimondi D, (DECCA): an epidemiology in critical care 2010;14:R210. Crit Care international study. M, CD, Grandi C, Salluh J, Soares Tomasi Comparison al. et S, Cascaes VR, Giombelli of CAM-ICU and ICDSC for the detection of delirium in critically ill patients focusing clinical outcomes. J Crit Care on relevant Pisani MA, Kong SY, Kasl SV, Murphy TE, Murphy TE, SV, Kasl Kong SY, Pisani MA, Days of delirium are Ness PH. Araujo KL, Van in an older with 1-year mortality associated Am J Respir unit population. intensive care Med 2009;180:1092-7. Crit Care L, van Schoonhoven M, den Boogaard Van Pickkers T, der Hoeven JG, van Achterberg and short-term Incidence consequences P. ill patients: a of delirium in critically Int observational cohort study. prospective J Nurs Stud 2012;49:775-83. M, Peters SA, van der den Boogaard Van Pickkers P, PC, Leffers Hoeven JG, Dagnelie al. The impact of delirium on the et P, of in-hospital mortality in intensive prediction 2010;14:R146. Crit Care patients. care Lin HC, Wang Lin SM, Huang CD, Liu CY, et al. Risk factors for the CH, Huang PY, development of early-onset delirium and the subsequent clinical outcome in mechanically patients. ventilated 2012;27:212-7. N, Dumont M, Dial S, Dubois MJ, Bergeron unit: Delirium in an intensive care Y. Skrobik Med a study of risk factors. Intensive Care 2001;27:1297-304. H, K, Kurosawa Takezawa Y, Iwasaki Kishi Y, unit patients Endo S. Delirium in critical care Gen room. an emergency admitted through Hosp Psychiatry 1995;17:371-9. Pun Peterson JF, Shintani A, Thomason JW, unit Intensive care Jackson JC, Ely EW. BT, of delirium is an independent predictor analysis longer hospital stay: a prospective of 261 non-ventilated patients. Crit Care 2005;9:R375-81. Zhang Z, Pan L, Ni H. Impact of delirium on clinical outcome in critically ill patients: a meta-analysis. Gen Hosp Psychiatry 2013;35:105-11. Hernan MA, Brumback B, Robins JM. Marginal structural models to estimate the causal effect of zidovudine on the survival of HIV-positive men. 13. 14. 9. 10. 11. 12. 15. 16. 17. 18. 19.
Chin Med J (Engl) 2010;123:993-9. Am Stat 2006;60:299-302. J Am Stat Assoc 1999;94:496-509. Shintani AK, Girard TD, Eden SK, Arbogast TD, Eden SK, Arbogast Girard Shintani AK, time bias Immortal KGM, Ely EW. PG, Moons time- application of research: in critical care for observational regression varying Cox Med 2009;37:2939- cohort studies. Crit Care 45. H, R, Grundmann M, Vonberg Wolkewitz S, Bärwolff P, Gastmeier J, Beyersmann the development of et al. Risk factors for and mortality on nosocomial pneumonia competing of application units: care intensive 2008;12:R44. risks models. Crit Care M, Cooper BS, Bonten MJ, Wolkewitz Barnett AG, Schumacher M. Interpreting of comparing risks in the presence and 2014;349:g5060. competing events. BMJ hazards Gray RJ. A proportional Fine JP, model for the subdistribution of a competing risk. Robins JM, Hernan MA, Brumback B. structural models and causal Marginal Epidemiology in epidemiology. inference 2000;11:550-60. S, Vansteelandt JF, Bekaert M, Timsit M, A, Garrouste-Orgeas Vesin Depuydt P, et al. Attributable mortality of ventilator- using reappraisal a pneumonia: associated Med Care Respir Crit J causal analysis. Am 2011;184:1133-9. an R RB. IPW: WM, Geskus der Wal Van weighting. package for inverse probability 2011;43:1-23. J Stat Software DA. On the so-called “huber Freedman standard “robust and estimator” sandwich errors”. Balas MC, Deutschman CS, Sullivan-Marx Balas MC, Deutschman CS, Sullivan-Marx Richmond EM, Strumpf NE, Alston RP, TS. Delirium in older patients in surgical units. J Nurs Scholarsh intensive care 2007;39:147-54. DX, Chen KS, Gu XE. Shi CM, Wang Incidence and risk factors of delirium in critically ill patients after non-cardiac surgery. Al-Qadheeb NS, Balk EM, Fraser GL, Skrobik Al-Qadheeb NS, Balk EM, Fraser GL, Skrobik Randomized ICU et al. JP, Riker RR, Kress Y, trials do not demonstrate an association delirium between interventions that reduce duration and short-term mortality: a Crit and meta-analysis. systematic review Med 2014;42(6):1442-54. Care A, Hall JB, Pohlman Patel SB, Poston JT, sedation- Rapidly reversible, JP. Kress 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53.
Psychosomatics 2006;47:56-61. Med 2006;34:1297-310. Crit Care Woods JC, Mion LC, Connor JT, Viray F, Jahan Viray F, JC, Mion LC, Connor JT, Woods agitation among L, Huber C, et al. Severe unit patients: ventilated medical intensive care and outcomes. characteristics frequency, Med 2004;30:1066-72. Intensive Care Aldemir M, Ozen S, Kara IH, Sir A, Bac B. in the delirium for factors Predisposing unit. Crit Care intensive care surgical Bos LD, de PM, Ong DS, Klein Klouwenberg Huson MA, et Beer FM, van Hooijdonk RT, of Centers for agreement al. Interobserver criteria for and Prevention Disease Control classifying infections in critically ill patients. Med 2013;41:2373-8. Crit Care of missing Imputation Engels JM, Diehr P. longitudinal data: a comparison of methods. J Clin Epidemiol 2003;56:968-76. Pandharipande P, Cotton BA, Shintani A, BA, Shintani A, Cotton P, Pandharipande et al. Jr, Morris JA J, Pun BT, Thompson factors for development and risk Prevalence intensive and trauma in surgical of delirium 2008;65:34-41. patients. J Trauma unit care Bergbom CW, Axell AI, Malmros Granberg unit DB. Intensive care IL, Lundberg with is associated syndrome/delirium of duration and therapy drug anemia, Acta Anaesthesiol ventilation treatment. Scand 2002;46:726-31. Ely EW, Zhang Y, McNicoll L, Pisani MA, SK. Delirium in the Siegel MD, Inouye and clinical unit: occurrence intensive care Soc Geriatr patients. J Am course in older 2003;51:591-8. D, Carroll J, Schillerstrom JS, Seaman metabolism oxidative Impaired TM. Brown delirium: a study of 101 ICU precipitates patients. 2001;5:265-70. Araujo Ness PH, Pisani MA, Murphy TE, Van KL, Inouye SK. Characteristics associated medical in a in older patients with delirium Intern Med unit. Arch intensive care 2007;167:1629-34. Jaber S, Chanques G, Altairac C, Sebbane M, et al. A prospective C, Perrigault PF, Vergne ICU: study of agitation in a medical-surgical incidence, risk factors, and outcomes. Chest 2005;128:2749-57. Zimmerman JE, Kramer AA, McNair DS, Malila FM. Acute Physiology and Chronic hospital IV: Health Evaluation (APACHE) critically ill today’s for assessment mortality patients. 35. 36. 40. 41. 31. 32. 33. 34. 37. 38. 39.
Attributable mortality of delirium in the ICU 7 116 Attributable mortality of delirium in the ICU 117 7
N Engl J Med 2013;369:1306-16. Ely EW. The Modifying the Impact of ICU- The Modifying the Ely EW. Dysfunction-USA Neurological Associated National Study: ClinicalTrials.gov. (MIND-USA) 2013:NCT01211522. Library of Medicine, A. The Gibson C, Tremblay W, Breitbart and delirium recall delirium experience: in hospitalized distress delirium-related their spouses/ patients with cancer, and their nurses. Psychosomatics caregivers, 2002;43:183-94. B, van der Mast Strijbos MJ, Steunenberg RC, Inouye SK, Schuurmans MJ. Design Hospital Elder Life and methods of the (HELP), a multicomponent Program delirium intervention to prevent targeted and efficacy in hospitalized older patients: in Dutch health care. cost-effectiveness BMC Geriatr 2013;13:78. TD, Jackson JC, Girard Pandharipande PP, et al. Morandi A, Thompson JL, Pun BT, Long-term cognitive impairment after critical illness. Trzepacz PT. Update on the Update on the PT. Trzepacz of delirium. Dement neuropathogenesis 1999;10:330-4. Geriatr Cogn Disord common neural a final Is there PT. Trzepacz pathway in delirium? Focus on acetylcholine and dopamine. Semin Clin Neuropsychiatry 2000;5:132-48. ER, Inouye Fong TG, Marcantonio Hshieh TT, in hypothesis deficiency Cholinergic SK. evidence. J delirium: a synthesis of current A Biol Sci Med Sci 2008;63:764- Gerontol 72. Steiner LA. Postoperative delirium. Part 1: pathophysiology and risk factors. Eur J Anaesthesiol 2011;28:628-36. van de Beek D, Eikelenboom Gool WA, Van Systemic infection and delirium: when P. cytokines and acetylcholine collide. Lancet 2010;375:773-5. de Miller T, KJ, Maclullich AM, Ferguson Rooij SE, Cunningham C. Unravelling the focus on a delirium: of pathophysiology J responses. of aberrant stress the role Psychosom Res 2008;65:229-38. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71.
Crit Care Nat Rev Neurol Anesth Analg 2011;112:1212-7. Pandharipande PP, Pun BT, Herr DL, Maze M, Herr DL, Maze M, Pun BT, Pandharipande PP, of sedation TD, Miller RR, et al. Effect Girard with dexmedetomidine vs lorazepam on acute brain dysfunction in mechanically randomized MENDS the ventilated patients: trial. JAMA 2007;298:2644-53. controlled 2008;371:126-34. MC, Pohlman Schweickert WD, Pohlman al. et CL, Esbrook AJ, Pawlik C, Nigos AS, therapy Early physical and occupational critically ill in mechanically ventilated, trial. controlled patients: a randomised Lancet 2009;373:1874-82. S. Quantifying biases in Greenland causal models: classical confounding vs bias. Epidemiology collider-stratification 2003;14:300-6. GB. Sepsis-associated Gofton TE, Young encephalopathy. 2012;8:557-66. RC, Baskett RJ. Martin BJ, Buth KJ, Arora of sepsis in post- Delirium as a predictor artery bypass grafting patients: coronary study. cohort retrospective a 2010;14:R171. KA, AM, Glynn RJ, Chan Walker Kurth T, K, et al. Results of Gaziano JM, Berger propensity multivariable logistic regression, adjustment, and matching, propensity weighting under propensity-based Am J conditions of nonuniform effect. Epidemiol 2006;163:262-70. Review Hughes CG, Pandharipande PP. of perioperative and articles: the effects unit sedation on brain organ intensive care dysfunction. related delirium versus persistent delirium persistent delirium delirium versus related Crit unit. Am J Respir care in the intensive Med 2014;189:658-65. Care Fuchs BD, Thomason JP, TD, Kress Girard et al. WD, Pun BT, Schweickert JW, sedation and safety of a paired Efficacy for protocol and ventilator weaning patients in intensive mechanically ventilated Controlled and Breathing (Awakening care trial. Lancet trial): a randomised controlled 61. 55. 56. 57. 58. 59. 60. 54. Observed analysis Primary imputation Backward definition Two-day Observed analysis Primary imputation Backward definition Two-day Observed analysis Primary imputation Backward definition Two-day Observed analysis Primary imputation Backward definition Two-day Observed analysis Primary imputation Backward definition Two-day Observed analysis Primary imputation Backward definition Two-day D7 D6 D5 D4 Coma Y DATA D3 Sedation D2 Delirium D1 No delirium No Reclassification of sedation and coma days in the various sensitivity analyses. This figure in the various sensitivity analyses. This figure S1: Reclassification of sedation and coma days Figure patients using various of sedation and coma days in six hypothetical shows the reclassification the original scoring. The dotted line denotes the definitions of delirium. ‘Observed’ denotes Day 1 to Day 7. that particular analysis. D1 – D7 = ICU observation exclusion of the patient from MENTAR SUPPLE figures Supplementary
Attributable mortality of delirium in the ICU 7 118 Attributable mortality of delirium in the ICU 119 7 1.77 (1.15-2.72) 2.08 (1.40-3.09) 1.19 (0.75-1.89) 1.77 (1.15-2.72) 1.74 (1.16-2.61) 0.86 (0.56-1.33) 2.02 (1.34-3.03) 2.15 (1.50-3.09) 1.67 (1.13-2.47) a b 2 Competing risks survival regression 2 Competing risks survival structural model 3 Marginal 1 Logistic regression 2 Competing risks survival regression structural model 3 Marginal 2 Competing risks survival regression structural model 3 Marginal 1 Logistic regression 1 Logistic regression Estimates of delirium-associated intensive care unit mortality using various definitions mortality using various definitions unit intensive care able S1: Estimates of delirium-associated Patients were classified as being exposed when they had at least two consecutive days of positive days of positive classified as being exposed when they had at least two consecutive Patients were Sedated patients were reclassified based on the first available valid delirium assessment following first available valid delirium assessment following based on the reclassified Sedated patients were CAM-ICU screening. These patients were compared to all other patients. Patients with one day to all other patients. Patients with one day compared These patients were CAM-ICU screening. sedation or excluded. Patients having delirium before or death were discharge of delirium before classified as delirious. coma and awakening with delirium were the cessation of sedation, using backward imputation. imputation. the cessation of sedation, using backward b a Sensitivity analyses imputation of sedation days (n=1112) Backward T of delirium Primary analysis Whole cohort (n=1112) Observed delirium for at least 2 days (n=1095) Observed delirium for at least 2 days (n=1095) Charlson co-morbidity index, acute adjusted for baseline variables (age, gender, All models were admission). admission type, and sepsis on IV score, health evaluation (APACHE) physiology and chronic to time-varying variables: sequential organ adjusted structural model was furthermore The marginal concentration, acidosis, urea sodium, status, temperature, sepsis score, assessment (SOFA) failure and analgesic medications. The logistic regression hematocrit, mechanical ventilation and sedative returnmodel structural marginal and analysis survival the and (OR) ratio odds the returns analysis variable in the ratios. Delirium was included as a time-dependent the subdistribution (SHR) hazard models. structural and marginal competing risks survival regression Supplementary tables Supplementary 3.27 (1.62-6.62) 3.64 (1.91-6.94) 1.44 (0.68-3.03) 2.00 (0.98-4.05) 2.37 (1.24-4.52) 1.38 (0.67-2.86) <80 (n=698) APACHE o sepsis at admission (n=616) No sepsis at 1.44 (0.83-2.48) 1.44 (0.83-2.48) 1.95 (1.31-2.90) 1.52 (0.95-2.42) 1.32 (0.81-2.17) 1.78 (1.10-2.87) 1.51 (0.91-2.51) > 80 (n=414) APACHE Sepsis at admission (n=496) Sepsis at admission Estimates of delirium-associated intensive care unit mortality in subgroups unit mortality in intensive care delirium-associated able S2: Estimates of 1 Logistic regression T model Statistical 1 Logistic regression regression 2 Competing risks survival structural model 3 Marginal regression 2 Competing risks survival structural model 3 Marginal index, acute Charlson co-morbidity gender, adjusted for baseline variables (age, All models were type, and sepsis on admission). admission IV score, (APACHE) health evaluation physiology and chronic variables: sequential organ adjusted to time-varying structural model was furthermore The marginal concentration, acidosis, urea sodium, status, temperature, sepsis score, assessment (SOFA) failure and analgesic medications. The logistic regression hematocrit, mechanical ventilation and sedative returnmodel structural marginal and analysis survival the and (OR) ratio odds the returns analysis variable in the ratios. Delirium was included as a time-dependent the subdistribution (SHR) hazard models. structural and marginal competing risks survival regression
Attributable mortality of delirium in the ICU 7 120
8 Incidence and outcomes of new-onset atrial fibrillation in critically ill patients with sepsis: a cohort study
Peter M.C. Klein Klouwenberg Sanne Kuipers Jos F. Frencken David S.Y. Ong Linda M. Peelen Marcus J. Schultz Marc J. Bonten Olaf L. Cremer Critically ill patients with sepsis are at increased risk of developing cardiac cardiac developing of risk increased at are sepsis with patients ill Critically New-onset AF is a common complication in critically ill patients with in critically ill patients with New-onset AF is a common complication To determine in a cohort of critically the incidence and outcomes of AF To We studied new-onset AF in a cohort of critically ill patients admitted for for a cohort of critically ill patients admitted studied new-onset AF in We Among 1,782 patients with sepsis, a total of 1,078 AF episodes occurred in in occurred AF episodes a total of 1,078 patients with sepsis, Among 1,782 ABSTRACT Introduction: dysrhythmias, most commonly atrial fibrillation (AF), due to the systemic inflammatory to the systemic inflammatory commonly atrial fibrillation (AF), due dysrhythmias, most hormones, and autonomic dysfunction stress levels of circulating increased response, at patients with sepsis are It is uncertain which critically ill that accompanies sepsis. with increased and whether it is associated new-onset AF, highest risk for developing morbidity and mortality. Objectives: ill patients with sepsis. Methods: between January units (ICU) in the Netherlands intensive care sepsis to two tertiary surgery of AF and those following cardiac 2011 and June 2013. Patients with a history survival time-dependent competing risks multivariable used We excluded. were analyses to determine the association of new-onset AF with mortality. Results: risk of new-onset AF was 10% (95% CI 8-12), 410 (23%) individuals. The cumulative sepsis and in patients with sepsis, severe 22% (95% CI 18-25), 40% (95% CI 36-44) episodes was 2 (IQR 1-3) per The median number of AF septic shock, respectively. hours. New-onset AF was associated with patient, with a median duration of 4 (2-10) ratio (CSHR) 0.71; 95%CI 0.62- rate (cause-specific hazard discharge both a reduced death rate (CSHR 1.66; 95% CI 1.28-2.14), accounting for a 0.82) and an increased ratio 2.42; hazard overall mortality risk (subdistribution increased than 2-fold more as a competing risk. discharge 95% CI 1.88-3.12) when considering ICU Conclusion: sepsis and is associated with excess mortality. sepsis and is associated with excess mortality.
Atrial fibrillation in patients with sepsis 8 124 Atrial fibrillation in patients with sepsis 125 8 . 9 , all of which are plausible risk factors for the factors for the plausible risk , all of which are 4 and effective in critically ill patients with sepsis, in critically ill patients with sepsis, and effective 8 . However, the incidence of AF has not been studied well well studied been not has AF of incidence the However, . . The local institutional review boards of both participating of both participating boards . The local institutional review 9 1-3 , although their reports suffered from highly variable patient highly variable patient from suffered , although their reports . Indeed, some authors have reported very high rates of AF in in AF rates of very high reported have authors Indeed, some . 6,7 5 . Affected sites of infection and likelihood of the infection (ascending from from (ascending infection the of likelihood and infection of sites Affected . 10 In the present study, we aimed to gain a better understanding of the incidence and and incidence the of understanding better a gain to aimed we study, present the In Sepsis is characterized by a systemic release of pro-inflammatory cytokines, high high cytokines, of pro-inflammatory by a systemic release Sepsis is characterized particularly in those who are predicted to be at highest risk. predicted particularly in those who are to the ICU with sepsis ill patients who present outcomes of new-onset AF in critically for admission. as the primary reason hospitals (UMC Utrecht and AMC Amsterdam) approved an opt-out method of of method an opt-out approved Amsterdam) and AMC hospitals (UMC Utrecht were participants and family members informed whereby consent (IRB number 10-056), at ICU admission. In that was provided by a brochure notified of the MARS project that declared of the UMC Utrecht board analysis, the review addition, for the current Involving Humans ACT of the Netherlands does not apply (IRB the Medical Research analysis, we included consecutive adult patients who were 14-227). For the present of admitted to the ICU between January 2011 and June 2013 in whom a diagnosis to established sepsis was made within the first two days after admission according criteria in critically ill patients who present to the ICU with sepsis, despite the fact that these these that fact the despite sepsis, with ICU the to present who patients ill critically in of their condition. to developing AF due to the nature prone patients may be particularly hormones, volume autonomic dysfunction, intravascular stress levels of circulating compromise shifts and cardiovascular development of AF of development inclusion criteria and poor methodological quality, prohibiting a unified interpretation a unified interpretation prohibiting quality, poor methodological inclusion criteria and data). Prior studies also showed an association of findings (S. Kuipers, unpublished with sepsis, albeit a mortality in patients of AF and increased between the occurrence unknown whether new-onset it remains for confounding. Therefore lack of correction it whether or disease, of severity for marker a merely is sepsis with patients in AF anti-arrhythmic if AF truly causes excess mortality, truly impacts outcome. However, might be both feasible prophylaxis The present analysis was incorporated in the Molecular Diagnosis and Risk Stratification analysis was incorporated in the Molecular Diagnosis and Risk Stratification The present intensive in the mixed study cohort prospective ongoing an project, (MARS) Sepsis of (ClinicalTrials.gov centers in the Netherlands (ICU) of two tertiary referral units care identifier NCT01905033) S METHODS Study population UCTION INTROD unit care intensive the in problem common (AF) is a fibrillation atrial New-onset in 4-9% in medical patients to 32-50% varying from incidences (ICU), with reported patients postcardiotomy none, possible, probable to definite) were classified as described in detail previously in detail described as classified were to definite) probable possible, none, patients with sepsis patients with sepsis . 11 . All patients were continuously monitored monitored continuously . All patients were 9 . A competing risks analysis provides two two . A competing risks analysis provides 12 . The SHR is, therefore, a summary measure of all of all a summary measure . The SHR is, therefore, 13 . We included all ICU admissions to assess mortality; included all ICU admissions to assess mortality; . We 6,7,14-19 Hourly recordings of the observed heart rhythm were available for 92% of the available for 92% of the of the observed heart rhythm were Hourly recordings ariables and outcomes with a 3-lead ECG bedside monitor. New-onset AF was defined as (1) the observation of New-onset AF was defined monitor. with a 3-lead ECG bedside side nurse or clinician, or (2)during two consecutive hours by the bed AF or atrial flutter forpharmacological of AF or atrial flutter plus either at least one observation treatment of All episodes digoxin) or electrical cardioversion. sotalol, magnesium, AF (amiodarone, manually checked by the two primary authors (PK, SK). The diagnoses possible AF were consensus definitions of the based on the sepsis and septic shock were of sepsis, severe Medicine the Society of Critical Care American College of Chest Physicians and Statistical analysis analysis. survival time-dependent a using mortality ICU on AF of effect the assessed We In such an analysis informative accounted for censoring of the survival time should be because patients who are as a competing risk for mortality, by considering discharge to patients who state compared health in a different the ICU alive are from discharged beyond that time point admitted remain V collected the data for this prospectively project team dedicated to the MARS A research for data integrity screened study and regularly We excluded patients having a history of chronic or paroxysmal AF, being supported by by being supported AF, or paroxysmal of chronic having a history excluded patients We arrest or cardiac cardiotomy following recent pacemaker, assist device or a ventricular < 24 hours. Both of stay in the ICU having a total length as well as those (within 7 days), during the study period. methods care units used standardized intensive care measures of association: the cause-specific hazard ratio (CSHR), which estimates the which estimates the ratio (CSHR), of association: the cause-specific hazard measures subdistribution death), and the and discharge ICU (both outcome on AF of effects direct AF given that (SHR), which describes the instantaneous risk of dying from ratio hazard AF the subject has not died from the onset of the first AF episode was regarded the moment that a patient was at risk. regarded the onset of the first AF episode was admission time. 12 cases (0.7%) were overruled and were labeled as having AF in our AF in our having labeled as and were overruled (0.7%) were admission time. 12 cases cases (12%) the start time of AF was adjusted post-hoc analysis and in another 52 no were IQR 0-2). There (median shift 1 hour; to the start of treatment according nor outcomes. missing data in any of the baseline variables separate cause-specific hazards and can be used to calculate the cumulative incidence incidence the cumulative calculate to used be and can hazards cause-specific separate adjusted for confounders (i.e., death in this study). We of the outcome of interest chosen a priori based on their expected associations with AF and mortality that were and on clinical expertise. These included consideration of the literature after careful diabetes malignancy, immunocompromise, compromise, cardiovascular age, gender, acute index, mass body compromise, renal compromise, respiratory mellitus, sepsis surgery, recent score, IV (APACHE) evaluation health chronic and physiology on infection, type of infection (community vs. hospital severity at admission, source and pathogen acquired),
Atrial fibrillation in patients with sepsis 8 126 Atrial fibrillation in patients with sepsis 127 8 hours 31 dysrhythmia after cardiac arrest cardiac after 31 dysrhythmia 22 use of assist device 60 died within 24 309 history of AF surgery cardiothoracic 165 hours 24 within discharged 245 2087 1782 with sepsis 2614 admitted All analyses were performed using SAS 9.2 (Cary, NC) and R version 2.14 (www.r- NC) and performed SAS 9.2 (Cary, using were All analyses displays the baseline patient characteristics stratified by their new-onset AF by their new-onset AF patient characteristics stratified able 1 displays the baseline Figure 1: Flowchart Figure Description of AF episodes episode recurrent was 5 hours (IQR 2-11) and of a The median length of a first episode by 26 the median heart rate increased 4 hours (IQR 2-10). During first AF episodes, with episodes it increased during recurrent bpm to 132 bpm (IQR 116-152), whereas by 6 decreased The mean arterial pressure 21 bpm to 118 beats/min (IQR 105-136). Hg after first mm to 70 -1-13) (IQR 6 mm Hg and by Hg mm to 66 0-15) (IQR Hg mm the start or an with 170 patients (41%) requiring episodes, respectively, and recurrent S S RESULT A total of excluded. were sepsis at admission of whom 832 patients A total of 2,614 had 1). (Figure a total of 16,826 days of observation contributing 1,782 patients remained Incidence of AF the study period. in 418 (23%) patients during episodes occurred A total of 1,087 AF T 22% 8-12), (CI) interval 10% (95% confidence AF was of risk status. The cumulative sepsis and septic patients with sepsis, severe (95% CI 18-25), 40% (95% CI 36-44) in p < 0.001). In patients with a (Cochrane-Armitage for trend, test shock, respectively 10-20), CI (95% 15% were incidences the infection, of likelihood definite or probable episodes AF of the (95% CI 35-46). The median number and 41% (95% CI 20-31) 26% range (IQR) 1-3; range 1-26). The first episode per patient was two (interquartile a median of 36 hours (IQR 11-68) in the ICU. after occurred ). P-values less than 0.05 were considered to be statistically significant. significant. to be statistically considered less than 0.05 were ). P-values project.org alue 0.02 0.03 0.01 0.39 0.18 0.22 0.04 0.40 0.47 V 0.002 0.006 0.002 0.002 <0.001 <0.001 <0.001 <0.001 <0.001 P 120 (9%) 771(57%) 243 (18%) 335 (25%) 210 (15%) 230 (17%) 141 (10%) 275 (20%) 272 (20%) 179 (13%) 214 (16%) 164 (12%) 294 (22%) 2.5 (0-8.1) 57 (45, 67) 74 (58, 94) 1,173 (86%) 1,047 (77%) No AF (n=1,364) 87 (21%) 63 (15%) 59 (14%) 53 (13%) 65 (16%) 84 (20%) 57 (14%) 4.9 (0-10) 105 (25%) 125 (30%) 263 (63%) 381 (91%) 121 (29%) 110 (26%) 121 (29%) 317 (76%) 66 (59, 73) AF (n=418) 89 (72, 108) e a b d c Diabetes mellitus Renal insufficiency Renal insufficiency Age, years Gender, male Gender, Race, white Malignancy Immunocompromise Immunocompromise ACE-inhibitors Beta-blockers Calcium entry blockers Respiratory insufficiency Respiratory insufficiency Body mass index > 30 Corticosteroids Charlson comorbidity index Charlson comorbidity Admission type, medical Lipid lowering drugs Cardiovascular disease Cardiovascular APACHE IV score APACHE Cardiovascular disease was defined as cerebrovascular disease or chronic cardiovascular cardiovascular disease or chronic disease was defined as cerebrovascular Cardiovascular Respiratory insufficiency was defined as chronic obstructive pulmonary disease or chronic obstructive pulmonary disease or chronic was defined as chronic Respiratory insufficiency Immunocompromise was defined as having acquired immune-deficiency syndrome, the use of the use immune-deficiency syndrome, was defined as having acquired Immunocompromise Renal insufficiency was defined as chronic renal insufficiency (creatinine > 177 µmol/L) or chronic chronic or µmol/L) > 177 (creatinine renal insufficiency as chronic was defined Renal insufficiency able 1. Patient characteristics Malignancy included both metastatic and hematologic malignancies. Malignancy included both metastatic and hematologic malignancies. corticosteroids in high doses (equivalent to prednisolone of > 75 mg/day for at least one week), of > 75 mg/day for at least one week), to prednisolone in high doses (equivalent corticosteroids hematologic use of antineoplastic drugs, recent drugs, current use of immunosuppressive current or documented humoral or cellular deficiency. malignancy, insufficiency (New York Heart Association class 4), chronic congestive heart failure (ejection (ejection congestive heart failure York Heart Association class 4), chronic (New insufficiency intermittens,disease (claudicatio peripheral vascular or 30%) < fraction percutaneous patients with insufficiency). transluminal angioplasty or bypass for arterial c d e a b T Patient characteristics dialysis. dialysis. respiratory insufficiency with functional disabilities (chronic mechanical ventilation, oxygen use at mechanical ventilation, oxygen use at with functional disabilities (chronic insufficiency respiratory pulmonary hypertension). home, or severe evaluation; health chronic and acute physiologic - atrial fibrillation; APACHE - AF Abbreviations: unit. ICU - intensive care Chronic medication Chronic Co-morbidities Admission characteristics Data are expressed as number (%) or median (inter-quartile range). as number (%) or median (inter-quartile expressed Data are
Atrial fibrillation in patients with sepsis 8 128 Atrial fibrillation in patients with sepsis 129 8 . 6,18,19 . While the duration of . While the duration of 21,22 . The overall incidence of AF in our our in AF of incidence overall The . 20 and supports the hypothesis that inflammation occurring occurring and supports the hypothesis that inflammation 7 or did not continuously monitor patients for arrhythmias or did not continuously monitor patients 6,18,19 New-onset AF was directly responsible for excess mortality in critically ill patients in critically ill patients for excess mortality responsible New-onset AF was directly is: what causes the additional morbidity and An important question that remains The increase in incidence of AF in patients with increasing severity of sepsis confirmssepsis of severity increasing with patients in AF of incidence in increase The during sepsis may act as a major trigger for AF trigger for major a as act may sepsis during the observed AF episodes was relatively short, this could be disastrous in a patient in a patient short, this could be disastrous the observed AF episodes was relatively with sepsis. AF was also associated with a more protracted length of stay in the ICU, length of stay in the ICU, protracted a more with sepsis. AF was also associated with in of dying resulting hazard patients longer to a daily increased exposing thereby evaluating the association between AF an even higher mortality rate. Prior studies of AF, imbalances, the time varying nature and death failed to adjust for baseline cohort study, death). This large and the competing events for AF (i.e. ICU discharge, overcome evaluating 1782 critically ill adults over > 15,000 ICU days, is the first to published the strongest these important methodological issues and thus represents evidence to date about AF in critically ill patients with sepsis. for mortality of new-onset AF in critically ill patients with sepsis? Potential explanations due to supply and demand infarction that AF induces a myocardial are effect a direct output, diminishes cardiac heart failure, mismatch, triggers or worsens pre-existing complications such as stroke and leads to thromboembolic DISCUSSION ill patients with sepsis and in critically that new-onset AF occurs frequently found We AF was independently New-onset with sepsis severity. that the incidence increased ICU mortality. length of stay and increased associated with both an increased of prior work the results The crude length of ICU stay and ICU case fatality rates were significantly higher in higher in significantly were ICU stay and ICU case fatality rates The crude length of 4.1 (2.2-8.3) days, p <0.001; AF (length of stay 7.5 (3.9-14.5) vs patients developing associated with an increased p <0.001). AF was significantly mortality 29% vs 14%, of ICU discharge into accounting the competing event of ICU death when taking hazard ratios (CSHR) hazard 1.88-3.12). Analysis of the cause-specific (SHR 2.42; 95% CI of dying (HR 1.66; 95% on the hazard effect direct that AF had a significant revealed of being discharged in a lower daily probability AF resulted CI 1.28-2.14). In addition, of dying in exposing patients longer to a daily risk of AF, the ICU after the onset from risk of dying in the ICU after AF is increased the ICU (HR 0.71, 95% CI 0.62-0.82). The stay in the ICU. and a prolonged on mortality effect of both a direct the result therefore increase in dose of noradrenalin. Half of the first episodes were treated with amiodarone, with amiodarone, treated were Half of the first episodes in dose of noradrenalin. increase sotalol. 11% with digoxin or and electrical cardioversion, magnesium, 5% with 47% with mortality and length of stay Association with cohort, however, was higher than most previous studies. This might be explained by the This might be explained by the studies. was higher than most previous cohort, however, or AF (e.g. by using diagnosed sepsis fact that many of these studies retrospectively ICD-codes) . Yet trials trials . Yet 22,24 . Side-effects . Side-effects 23 , guidelines for medical , guidelines for medical 28-30 , such a strategy warrants further , such a strategy warrants further 8 and in the light of recent evidence on evidence on and in the light of recent 31 . Whatever the exact etiology of the excess mortality may be, the mortality may be, the excess of the Whatever the exact etiology . 25-27 Although the efficacy of pharmacologic prophylaxis and treatment of atrial of atrial and treatment pharmacologic of prophylaxis Although the efficacy This study has several limitations. Firstly, although we used prospective data, the data, the although we used prospective This study has several limitations. Firstly, of anti-arrhythmic drugs, mostly amiodarone, may also be a cause of the additional a cause of the additional may also be amiodarone, drugs, mostly of anti-arrhythmic and mortality of acute lung injury, including the development morbidity, ICU-patients with sepsis are missing so far. It is therefore important to examine important to examine It is therefore missing so far. sepsis are ICU-patients with with associated outcomes adverse of risk the diminish might that strategies preventive measure preventative effective an is use beta-blocker example, For sepsis. during AF population surgery in the cardiothoracic Sources of support Molecular Medicine (http:// This work was supported by the Center for Translational funding from research MARS (grant 04I-201). MB has received project www.ctmm.nl), (NWO Vici 918.76.611). of Scientific Research the Netherlands Organization CONCLUSION a common complication in critically ill patients This study shows that new-onset AF is possible association between severity of illness with sepsis. Our findings support the ICU mortality. AF is associated with increased of AF. and the occurrence who is already critically ill. In addition there is evidence that the negative hemodynamic hemodynamic that the negative is evidence ill. In addition there critically who is already of AF with resolution immediately might not disappear consequences the safety of beta-blocker use in septic patients the safety of beta-blocker use in septic importance of prevention of new-onset AF in critically ill sepsis patients is clear. is clear. new-onset AF in critically ill sepsis patients of importance of prevention evaluated well has been surgery fibrillation after cardiac investigation. When evaluating pharmacological prophylactic strategies it will be key key be will it strategies pharmacologicalevaluating When investigation. prophylactic risk of developing AF. at greatest patients who are those to target clinicians Because bedside nurses and retrospectively. identified AF episodes were deemed clinically insignificant, the incidence of episodes of AF that are may not record we excluded patients with although Secondly, AF in sepsis could be underestimated. admission ICU before AF undiagnosed with patients that history of AF it is possible a this study was performed in two tertiary ICUs which limits have been included. Thirdly, although the patient case-mix is similar to other studies in this the generalizability, is always a there of the study, because of the observational character Finally, area. confounding. possibility for unmeasured investigating the role of amiodarone as a prophylactic treatment for AF following for AF following treatment as a prophylactic of amiodarone investigating the role in adverse events or ICU significant differences have not shown thoracic surgery length of stay
Atrial fibrillation in patients with sepsis 8 130 Atrial fibrillation in patients with sepsis 131 8 Dec 16 Dec 16 J Intensive Care Med. Med. Care Intensive J J Am Statist Assoc. Jun 1999;94:496-509. Wolkewitz M, Vonberg R, Grundmann H, et al. R, Grundmann M, Vonberg Wolkewitz the development of nosocomialRisk factors for on intensive care pneumonia and mortality risks models.units: application of competing Apr 2008;12(2):R44. Critical Care. Hazards Proportional A RJ. Gray JP, Fine of a CompetingModel for the Subdistribution Risk. Gajic O, Afessa B. Salman S, Bajwa A, atrial fibrillation in critically ill Paroxysmal sepsis. with patients May-Jun 2008;23(3):178-183. MA, Heckbert SR, Greiner AJ, Walkey among Medicare et al. Atrial fibrillation with sepsis: beneficiaries hospitalized incidence and risk factors. Am Heart J. Jun 2013;165(6):949-955 e943. J, Nguyen HD, Ewer Gibbs HR, Swafford fibrillation atrial Postoperative MK. Ali MS, risks preoperative in cancer surgery: Aug Oncol. and clinical outcome. J Surg 1992;50(4):224-227. et al. R, Rich MW, Hidalgo JD, Krone after Supraventricular tachyarrhythmias hematopoietic stem cell transplantation: incidence, risk factors and outcomes. Bone Oct 2004;34(7):615-619. Transplant. Marrow GL, Morris PE. Incidence and Prognosis Wells of Atrial Fibrillation in Patients With Sepsis. 2(6):293-297. Res. 2011;Volume Card L, Jarbrink Christian SA, Schorr C, Ferchau Clinical DR. Gerber JE, Parrillo ME, characteristics and outcomes of septic patients with new-onset atrial fibrillation. J Dec 2008;23(4):532-536. Crit Care. RJ, Martin DO, Apperson-Hansen Aviles C, et al. Inflammation as a risk factor for atrial fibrillation. Circulation. 2003;108(24):3006-3010. Clark DM, Plumb VJ, Epstein AE, Kay an irregular of GN. Hemodynamic effects sequence of ventricular cycle lengths during Oct atrial fibrillation. J Am Coll Cardiol. 1997;30(4):1039-1045. C, Soave M. Atrial Volpe F, Cavaliere units. Curr fibrillation in intensive care 2006;17:367-374. Anaesth Crit Care. O’Neill PG, Puleo PR, Bolli R, Rokey R. Return of atrial mechanical function following electrical conversion of atrial dysrhythmias. Am Heart J. Aug 1990;120(2):353-359. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. Aug Aug Oct 23 2013;310(16):1683- Oct JAMA. Nov 23 2011;306(20):2248-2254. Jun 1992;101(6):1644-1655. Crit Care. 2010;14(3):R108. Crit Care. Jul 1 2008;178(1):20-25. Artucio H, Pereira M. Cardiac arrhythmias arrhythmias M. Cardiac Artucio H, Pereira epidemiologic study. in critically ill patients: Med. Dec 1990;18(12):1383-1388. Crit Care C. The new Brathwaite D, Weissman following major onset of atrial arrhythmias is associated surgery noncardiothoracic Chest. mortality. with increased Annane D, Sebille V, Duboc D, et al. Incidence al. Incidence Duboc D, et V, Annane D, Sebille in arrhythmias of sustained and prognosis critically ill patients. Am J Respir Crit Care Med. 1998;114(2):462-468. sepsis Severe T. Angus DC, van der Poll Aug 29 and septic shock. N Engl J Med. 2013;369(9):840-851. is Atrial fibrillation Y. Launey Seguin P, not just an artefact in the ICU. Crit Care. 2010;14(4):182. AJ, Wiener Curtis RS, Ghobrial JM, Walkey and mortality LH, Benjamin EJ. Incident stroke associated with new-onset atrial fibrillation sepsis. severe with hospitalized patients in JAMA. Meierhenrich R, Steinhilber E, Eggermann impact of C, et al. Incidence and prognostic new-onset atrial fibrillation in patients with observational septic shock: a prospective study. M, et al. A, Ertmer C, Westphal Morelli with esmolol of heart rate control Effect on hemodynamic and clinical outcomes in randomizedpatients with septic shock: a trial. clinical 1691. PM, Ong DS, Bos LD, et Klein Klouwenberg of Centers for agreement al. Interobserver criteria for and Prevention Disease Control classifying infections in critically ill patients. Med. Oct 2013;41(10):2373-2378. Crit Care Bone RC, Balk RA, Cerra FB, et al. Definitions and guidelines failure for sepsis and organ for the use of innovative therapies in sepsis. The ACCP/SCCM Consensus Conference Committee. American College of Chest Medicine. Physicians/Society of Critical Care Chest. Levy MM, Rhodes A, et al. Dellinger RP, Surviving Sepsis Campaign: international sepsis guidelines for management of severe Med. and septic shock, 2012. Intensive Care Feb 2013;39(2):165-228. 2. 3. s Reference 1. 4. 5. 6. 7. 8. 9. 10. 11. Martinez EA, Bass EB, Zimetbaum P. P. Bass EB, Zimetbaum Martinez EA, rhythm: American of Pharmacologic control thefor Physicians guidelines College of Chest of postoperative and management prevention Chest. surgery. cardiac atrial fibrillation after Aug 2005;128(2 Suppl):48S-55S. Krishnan K, Kim MH, AT, Aasbo JD, Lawrence reduces prophylaxis RG. Amiodarone Trohman morbidity and length of major cardiovascular a meta-analysis. surgery: stay after cardiac Ann Intern Med. Sep 6 2005;143(5):327-336. al. et E, Crystal AM, KA, Yusuf Arsenault post-operative Interventions for preventing patients undergoing atrial fibrillation in Rev. Cochrane Database Syst heart surgery. 2013;1:CD003611. DC, Kilborn MJ, Keech AC. Burgess of post-operative Interventions for prevention atrial fibrillation and its complications after a meta-analysis. Eur Heart J. surgery: cardiac Dec 2006;27(23):2846-2857. 28. 29. 30. 31. Skroubis G, Galiatsou E, Metafratzi E, Metafratzi G, Galiatsou Skroubis Nakos G. A, Kitsakos A, Z, Karahaliou in toxicity lung acute Amiodarone-induced Scand. Acta Anaesthesiol an ICU setting. Apr 2005;49(4):569-571. DS, et al. HA, Wall JE, Wroblewski Tisdale amiodarone A randomized trial evaluating of atrial fibrillation after for prevention Sep Ann Thorac Surg. pulmonary resection. 894-885. 2009;88(3):886-893; discussion DS, et al. A HA, Wall JE, Wroblewski Tisdale study of amiodarone randomized, controlled of atrial fibrillation after for prevention J Thorac transthoracic esophagectomy. Jul 2010;140(1):45-51. Surg. Cardiovasc Giving al. et M, Qamar A, Soodan J, Kerstein perioperatively IV and oral amiodarone of postoperative atrial for the prevention coronary fibrillation in patients undergoing Chest. the GAP study. artery bypass surgery: Sep 2004;126(3):716-724. 24. 25. 26. 27.
Atrial fibrillation in patients with sepsis 8 132
9 Predicting evolution of disease severity in critically ill patients with severe sepsis or septic shock
Peter M.C. Klein Klouwenberg Cristian Spitoni David S.Y. Ong Jos F. Frencken Lonneke A. van Vught Tom van der Poll Marc J. Bonten Olaf L. Cremer Patients admitted with sepsis to the intensive care unit (ICU) typically (ICU) typically unit care to the intensive with sepsis Patients admitted We developed a prognostic model for daily predictions of treatment of treatment daily predictions for model a prognostic developed We We studied 1,399 ICU admissions for sepsis in 1,264 patients, contributing 1,399 ICU admissions for sepsis in 1,264 patients, contributing studied We response in critically ill patients admitted for sepsis. This model may help clinicians critically ill patients admitted for sepsis. This model may help clinicians in response and admission, to evaluate the added to make informed decisions about treatment value of new biomarkers, and can be used for in silico modeling of new prognostic the design of new clinical trials. therapies for sepsis to improve present with new organ dysfunction, and it is difficult to predict whether resolution resolution whether predict to difficult is it and dysfunction, organ new with present a developed We of abnormalities resuscitation. following occur will progression or estimating daily patients admitted with sepsis, by model for individual prediction one week after failure organ to death or multiple of disease progression probabilities parameters. available in the ICU, using routinely studied we prospectively period, Over a three-year Material and methods: in the Netherlands with a admitted with sepsis to two tertiary ICUs consecutive adults disease stages of increasing classified patients into three hours. We length of stay > 24 organ rapidly reversible failure, at risk for developing organ severity at admission: calculated the absolute probabilities We failure. dysfunction, or persistent multi-organ after one week in the ICU failure and multi-organ to transition to death, discharge model, that was updated daily using both using a continuous time multi-state Markov the Prediction-Infection-Response- factors from baseline and time-varying prognostic (PIRO) model. dysfunction Organ Results: day 7. At the time of admission, 100 (7%) a total of 8,664 observation days until and 133 (10%) had dysfunction, at risk, 1,166 (83%) had organ classified patients were dysfunction evolved to a 245 (21%) patients admitted with organ failure. multi-organ alive within discharged or were stage or died and 788 (68%) improved severe more 43 (32%) failure admitted with multi-organ the first seven days. Among the patients alive within the first seven days. discharged or were died and 78 (59%) improved evolution predict available parameters we could moderately efficiently Using routinely lactate levels being status and increased of disease with age, immunocompromised failure. to multi-organ significantly associated with progression Conclusion: Abstract Introduction:
Predicting treatment response in patients with sepsis 9 136 Predicting treatment response in patients with sepsis 137 9 , which , which 2 . Although the . Although the 3 Moreover, the model the model Moreover, . The Institutional Review Board approved an opt-out an opt-out approved . The Institutional Review Board 4 . Since the early 1990s, the severity of sepsis is usually classified into into the severity of sepsis is usually classified . Since the early 1990s,
1 In patients admitted to the ICU with recent onset of organ dysfunction or shock, dysfunction or shock, onset of organ recent In patients admitted to the ICU with are associated with clinically relevant outcomes, in particular mortality outcomes, relevant associated with clinically are Methods Study design and population (MARS) This work was part of the Molecular Diagnosis and Risk Stratification of Sepsis cohort study that was performed in the mixed ICUs of prospective a large project, centers in the Netherlands between 2011 and 2013 (ClinicalTrials. two tertiary referral gov identifier NCT01905033) uction Introd injures infection severe to a response body’s that arises when the syndrome Sepsis is a its own tissues classification of disease stage by the presence of organ dysfunction or hypotension dysfunction or hypotension of organ stage by the presence classification of disease with community-acquired risk stratification of patients presenting seems suitable for for appropriate this categorization may be less department, sepsis to the emergency management of signs and with nosocomial sepsis. Because active hospitalized patients magnitude of the underlying as complications ensue, the true symptoms takes place nosocomial sepsis, i.e. without Furthermore, may be obscured. disease process not to cause mortality and will typically or shock, is unlikely failure associated organ unit (ICU). In fact, most for admittance to an intensive care be the principal reason of these patients will have alternative admission, with pathologies that warrant ICU a co-diagnosis. nosocomial sepsis being merely abnormalities is the main of these —potentially still reversible— averting progression bedside - at the for clinicians to predict it is very difficult However, goal of critical care. and who will interventions, favorably to critical care - which patients will respond or death despite resuscitative failure multi-organ deteriorate to develop prolonged with developed a model to calculate, for an individual patient therefore, We, efforts. of multi-organ and the development of death, discharge sepsis, the probabilities the first week in ICU. For this, we used a framework that assumes the during failure dysfunction at ICU admission and in which organ of potentially reversible presence or death. Our failure at risk for the development of multiple organ patients are method uses information to make daily of the clinical course of individual patients the impact of of outcome. As such, it can be used to predict updated predictions but failure, on intermediate outcomes or organ novel interventions with known effects on clinical outcomes such as death or persistent multiple-organ with unknown effects . of new clinical trials the design can assist in therefore and failure value of new biomarkers. may help in the evaluation of added prognostic three stages of increasing severity: sepsis, severe sepsis, and septic shock severe severity: sepsis, stages of increasing three consent method (IRB number 10-056C). For the current study, we analyzed all adults adults all analyzed we study, current the For 10-056C). number method (IRB consent able 2). . We estimated the estimated the . We 8 with laboratory and clinical clinical with laboratory and 6 , i.e., a documented or suspected or suspected , i.e., a documented 5 and included both baseline and daily (time-varying) variables: predisposing variables: predisposing and included both baseline and daily (time-varying) ). For instance, we defined cardiovascular dysfunction not only by the dysfunction not only by the cardiovascular able 1). For instance, we defined 5,7 infection combined with a disorder in at least one of either general, inflammatory, inflammatory, in at least one of either general, with a disorder infection combined or tissue perfusion parameters. dysfunction hemodynamic, organ lassification of organ dysfunction Classification of organ failure have associated organ admitted to the ICU for sepsis will Since almost all patients sepsis is not defined precisely, for severe failure the level of organ or shock and since the (dys)function, by extending to their level of organ according we classified patients criteria score Assessment (SOFA) Failure Sequential Organ ‘transient’ states (at risk, dysfunction, and multi-organ transitions between the three alive and death) in this model. A multi- and the ‘absorbing’ states (discharge failure) moves state model with piecewise constant intensities describes how a patient between a series of disease states, and enables the calculation of the probabilities transitions to these movements. The Markov model assumes that any future related may occur state of the patient. Carry-over effects dependent only on the current are failure, ‘incubating’ organ by already affected variables are when values of predictor with sepsis as their main reason for ICU admission between January 2011 and and January 2011 ICU admission between for as their main reason with sepsis to the was defined according 24 hours. Sepsis than more 2013, with a stay of December 2001 International Definitions Conference Sepsis Statistical analysis of death, calculated, for an individual patient with sepsis, absolute probabilities We ICU admission, after one week of failure and the development of multi-organ discharge 1) using a continuous-time Markov multi-state model (Figure Predictors from the Prediction-Infection-Response-Organ dysfunction (PIRO) classification the Prediction-Infection-Response-Organ from Predictors used were rognostic variables Prognostic variables (T elevated lactate levels and positive fluid also by infusion, but need for norepinephrine the duration of symptoms in our definitions in balances. In addition, we incorporated instance, For malfunction. organ the of reversibility potential the for allow to order failure, few hours would indicate risk of organ oliguria or hypotension lasting only a to be a marker of or hypotension that lasted for > 1 day was regarded oliguria whereas we classified patients as being at risk, having Subsequently, failure. established organ (T failure organ dysfunction, or having persistent multiple organ reversible characteristics (i.e. age, gender, immunodeficiency, cardiovascular disease, respiratory disease, cardiovascular immunodeficiency, characteristics (i.e. age, gender, use of corticosteroids), mellitus, and current diabetes renal insufficiency, insufficiency, site of infection, and causative pathogen),infection characteristics (i.e. time of acquisition, white blood cell count, lactate, temperature, protein, (i.e. C-reactive variables response of prediction. the time dysfunction at of organ rate), and the level heart rate, and respiratory
Predicting treatment response in patients with sepsis 9 138 Predicting treatment response in patients with sepsis 139 9 Death Failure displays the main patient able 3 displays the main patient Death Discharge Dysfunction and SAS 9.2 (Cary, NC). P values <0.05 were considered to be to be considered NC). P values <0.05 were and SAS 9.2 (Cary, 9 At risk Discharge
Performance of the model was assessed using the Pearson’s goodness of fit test which Performance of the model was assessed using the Pearson’s 2013, Team performed version 3.0.2 (R Core using R studio All analyses were characteristics at admission, stratified by the severity of organ failure at admission. at admission. failure characteristics at admission, stratified by the severity of organ as at admission increased IV score by the APACHE The severity of illness as measured was also associated with failure Organ pronounced. dysfunction became more organ length of stay and case fatality rate in the ICU. In contrast, age, gender, an increased co-morbidities, and the admission type (medical vs. surgical) of chronic the presence between the groups. did not significantly differ s Result Study population ICU admissions for sepsis in 1,264 patients, During the study period, we studied 1,399 2 shows the classification of day 7. Figure yielding 8,664 observation days until as at the time of ICU admission failure categories of organ the three patients across 1,166 patients (83%) had potentially well as their outcome status. At ICU admission, persistent multiple more dysfunction, 133 patients (10%) had organ reversible Overall failure. at risk to develop organ patients (7%) were and 100 failure, organ ICU mortality rate at day 7 was 14% (n=199). T and thus become part of the outcome rather than being a true prognostic factor. factor. rather than being a true prognostic and thus become part of the outcome for every other day (days 1, 3, 5, and 7). only modeled transitions were Therefore, evolution of disease. Here, of how well the model is in capturing the measure a provides test. using the Chi square of the upper bound of the p-values approximation an we report Vienna, Austria) Transitions of the cascade. The arrows represent forward or backward progression from a a from progression or backward forward represent cascade. The arrows of the 1: Transitions Figure advancing to a of state. The probabilities a stage to an absorbing or from severity stage to another calculated by stage or to an absorbing state are to a less severe stage or regressing advanced more out of a total of 3,855 transitions model with piecewise constant intensities. 43 the multi-state Markov discharge. or risk’ ‘at to directly ‘failure’ from or death or ‘failure’ to directly risk’ ‘at from were (1%) modeled by adding the intermediate transitions. state of ‘dysfunction’ for these These transitions were statistically significant.
Impaired enteral feeding enteral Impaired function gut Normal Gastro-intestinal Persistent food intolerance food Persistent
w x
, or deficient protein synthesis protein deficient or , or abnormal protein synthesis protein abnormal or transaminitis
s v u
Mild hyperbilirubinemia Mild function liver Normal Liver , , transaminitis mild or , severe severe , hyperbilirubinemia Severe
r q t
Mild thrombocytopenia Mild hemostasis Normal Coagulation or abnormal coagulation abnormal or Severe thrombocytopenia Severe
o n p
Oliguria GFR preserved with diuresis Adequate Renal or GFR decrease > 50% > decrease GFR or 75% > decrease GFR or oliguria/anuria Persistent
k j m l
Mild hypoxemia Mild hypoxemia without breathing Spontaneous Respiratory Persistent severe hypoxemia severe Persistent
h i
, or positive fluid balance fluid positive or , vasopressors
e f
Hemodynamic stability without support without stability Hemodynamic Cardiovascular , use of inotropes/ of use , hypotension Arterial Refractory shock Refractory
d g
or use of continuous sedation continuous of use or Persistent coma Persistent Awake and non-delirious and Awake CNS Delirium
a a c b
isk R ysfunction D Failure
Classification of new-onset organ failure organ new-onset of Classification able 1: 1: able T
Predicting treatment response in patients with sepsis 9 140
Predicting treatment response in patients with sepsis 141 9
diarrhea, intra-abdominal hypertension or abdominal compartment syndrome for > 24 hrs 24 > for syndrome compartment abdominal or hypertension intra-abdominal diarrhea,
Persistent food intolerance was defined as the inability to provide enteral feeding due to high gastric aspirate volume, vomiting, bowel distension, severe severe distension, bowel vomiting, volume, aspirate gastric high to due feeding enteral provide to inability the as defined was intolerance food Persistent
x
Impaired enteral feeding was defined as a daily caloric intake < 50% of calculated needs calculated of 50% < intake caloric daily a as defined was feeding enteral Impaired
w w
Deficient protein synthesis was defined as a plasma albumen concentration < 15 g/L 15 < concentration albumen plasma a as defined was synthesis protein Deficient
v v
Severe transaminitis was defined as aspartate transferase or alanine transaminase blood levels > 1,000 U/L 1,000 > levels blood transaminase alanine or transferase aspartate as defined was transaminitis Severe
u
Severe hyperbilirubinemia was defined as plasma total bilirubin > 100 mmol/L 100 > bilirubin total plasma as defined was hyperbilirubinemia Severe
t
Mild transaminitis was defined as aspartate transferase or alanine transaminase blood levels > 500 U/L 500 > levels blood transaminase alanine or transferase aspartate as defined was transaminitis Mild
s
Abnormal protein synthesis was defined as a plasma albumen concentration < 20 g/L g/L 20 < concentration albumen plasma a as defined was synthesis protein Abnormal
r r
Mild hyperbilirubinemia was defined as plasma total bilirubin > 30 mmol/L 30 > bilirubin total plasma as defined was hyperbilirubinemia Mild
q
Severe thrombocytopenia was defined as a platelet count < 50,000/uL < count platelet a as defined was thrombocytopenia Severe
p
Abnormal coagulation was defined as International normalized ratio > 1.5 or activated partial thromboplastin time > 60 sec 60 > time thromboplastin partial activated or 1.5 > ratio normalized International as defined was coagulation Abnormal
o
Mild thrombocytopenia was defined as a platelet count < 100,000/uL < count platelet a as defined was thrombocytopenia Mild
n
µmol/L with an acute rise of > 44 µmol/L, or use of renal replacement therapy replacement renal of use or µmol/L, 44 > of rise acute an with µmol/L
Glomerular filtration rate decrease > 75% was defined as a > 3-fold increase in serum creatinine from baseline, a single creatinine concentration > 350 350 > concentration creatinine single a baseline, from creatinine serum in increase 3-fold > a as defined was 75% > decrease rate filtration Glomerular
m
Persistent oliguria/anuria was defined as urine output < 0.3 ml/kg/h for > 24 hrs, or < 200 ml per day per ml 200 < or hrs, 24 > for ml/kg/h 0.3 < output urine as defined was oliguria/anuria Persistent
l
Glomerular filtration rate decrease > 50% was defined as a > 1.5-fold increase in serum creatinine from baseline baseline from creatinine serum in increase 1.5-fold > a as defined was 50% > decrease rate filtration Glomerular
k
Oliguria was defined as urine output < 0.5 ml/kg/h for > 6 hrs, or < 500 ml per day per ml 500 < or hrs, 6 > for ml/kg/h 0.5 < output urine as defined was Oliguria
j
ventilation with positive end-expiratory pressure > 8 cm H2O cm 8 > pressure end-expiratory positive with ventilation
Persistent severe arterial hypoxemia was defined as the ratio of partial pressure arterial oxygen and fraction of inspired oxygen < 200 despite mechanical mechanical despite 200 < oxygen inspired of fraction and oxygen arterial pressure partial of ratio the as defined was hypoxemia arterial severe Persistent
i i
300 and positive end-expiratory pressure > 5 cm H2O cm 5 > pressure end-expiratory positive and 300
Mild arterial hypoxemia was defined as the use of mechanical ventilation with the ratio of partial pressure arterial oxygen and fraction of inspired oxygen < < oxygen inspired of fraction and oxygen arterial pressure partial of ratio the with ventilation mechanical of use the as defined was hypoxemia arterial Mild
h
> 3 mmol/L 3 > lactatemia or ml/24h 2,000 > 12 hrs, with concurrent positive fluid balances fluid positive concurrent with hrs, 12
Refractory shock was defined as the use of high-dose vasopressors high-dose of use the as defined was shock Refractory (including norepinephrine at > 100 ng/kg/min or arginine vasopressin at any dose) for > > for dose) any at vasopressin arginine or ng/kg/min 100 > at norepinephrine (including
g
A positive balance was defined as cumulative fluid intake minus output > 2,000 ml/24h 2,000 > output minus intake fluid cumulative as defined was balance positive A
f
Use of inotropes and vasopressors includes (but is not limited to) the continuous infusion of dobutamine, milrinone, and norepinephrine at any dose dose any at norepinephrine and milrinone, dobutamine, of infusion continuous the to) limited not is (but includes vasopressors and inotropes of Use
e
Arterial hypotension was defined as a systolic blood pressure < 90 mm Hg for > 2 hrs 2 > for Hg mm 90 < pressure blood systolic a as defined was hypotension Arterial
d
sedation scale ≤ -4 or Glasgow coma scale ≤ 8) for > 24 hrs 24 > for 8) ≤ scale coma Glasgow or -4 ≤ scale sedation
Persistent coma was defined as unresponsiveness to verbal commands, both with or without the use of continuous intravenous sedation (Richmond agitation agitation (Richmond sedation intravenous continuous of use the without or with both commands, verbal to unresponsiveness as defined was coma Persistent
c
Use of continuous sedation includes (but is not limited to) the infusion of propofol and midazolam at any dose any at midazolam and propofol of infusion the to) limited not is (but includes sedation continuous of Use
b
Delirium was defined as a positive confusion assessment method for the intensive care unit (CAM-ICU) score on ≥ 1 observation observation 1 ≥ on score (CAM-ICU) unit care intensive the for method assessment confusion positive a as defined was Delirium
a
failure level for which the patient classifies is assigned. is classifies patient the which for level failure
are considered to be ‘at risk’ for the development of organ failure. In cases where definitions for organ failure are not mutually exclusive the highest organ organ highest the exclusive mutually not are failure organ for definitions where cases In failure. organ of development the for risk’ ‘at be to considered are ‘Failure’ denotes long-term or irreversible loss of organ function; ‘dysfunction’ denotes readily reversible or only partial loss of organ failure. All other patients patients other All failure. organ of loss partial only or reversible readily denotes ‘dysfunction’ function; organ of loss irreversible or long-term denotes ‘Failure’ n=146 n=136 n=143 -sepsis) hrs hrs Failure Multi-organ failure Multi-organ Severe malfunctions in ≥ 3 malfunctions in ≥ 3 Severe systems organ n=133 (10%) Discharged <24 hrs Death <24 Alternative (non diagnosis <24 Exclusions • • • n=1,399 n=1,824 Dysfunction n=1,166 (83%) rgan dysfunction Organ Mild malfunctions of limited systems duration in up to 3 organ malfunctions in up to 2 Severe systems organ Sepsis at admission At risk n=100 (7%) Classification of organ failure on the patient level failure of organ able 2: Classification provides the transition hazards as obtained by multivariable modeling. Age was as obtained by multivariable transition hazards the able 4 provides T Risk No significant organ dysfunction No significant organ Mild malfunctions of limited systems duration in up to 2 organ Figure 2: Flowchart of patients Figure rediction of treatment response of treatment Prediction T to death, and immunodeficiency was particularlyassociated with especially the transitions volution of organ failure Evolution of organ admission. function during the first week of of organ 3 depicts the evolution Figure prevalences 1, and subsequent day on most prevalent dysfunction was Acute organ mostly dysfunction occurred and respiratory system: cardiovascular varied by organ remained renal and central nervous system dysfunction during the first days in ICU, and in present was failure Cardiovascular of admission. first week the during stable more with and in 550 of 1,309 days (42%) failure (84%) with multi-organ 295 of 350 days (21%) admitted with dysfunction evolved dysfunction. In total, 245 of 1,166 patients (9%) developed multi- stage or died in the first week: 100 patients severe to a more patients (68%) admitted and 145 patients (12%) died. In contrast, 788 failure organ days. alive within the first seven discharged or were dysfunction improved with organ the within died (32%) 43 failure multi-organ with admitted patients 133 the Among alive discharged or were improved and 78 (59%) first seven days after admission, organ show that the classification of results therefore within the first seven days. Our failure. of organ and progression function works well for both improvement
Predicting treatment response in patients with sepsis 9 142 Predicting treatment response in patients with sepsis 143 9
Figure 4 Figure , with higher values , with higher values -14 Evolution of organ failure over time. Distribution of the severity of organ failure by organ by organ failure of the severity of organ over time. Distribution failure Evolution of organ The performance upper limit of the p-value of the full model as shown by the
levels had an absolute risk at day 7 for discharge alive of 48%, for death of 15%, and multi- levels had an absolute risk at day 7 for discharge with a urinary tract healthy female patient A 53-year old previously of 1%. failure organ of of discharge and lactate levels had a probability protein infection and normal C-reactive of 1%. failure 72%, of death of 6%, and of multi-organ (p=0.01) was significantly better than the crude model (p=10 depicts the predicted probabilities of organ failure and outcome in a hypothetical ‘naïve’hypothetical in a and outcome failure organ of probabilities predicted the depicts co-morbidities and with no systemic response) modeled patient (young male patient without male For instance, a 72-year old immunocompromised patients. representative and three protein C-reactive pneumonia and increased patient admitted with a community-acquired alive of 39%, for death of 33%, and lactate levels had an absolute risk at day 7 for discharge and lactate protein of 5%. The same patients with normal C-reactive failure and multi-organ as < 3 individual dysfunctioning organ systems. systems. as < 3 individual dysfunctioning organ to abdominal Compared failure. organ severe to more associated with progression Lactate but not associated with delayed discharge. infections, pulmonary infections were mortality. increased and disease of worsening predicted levels protein C-reactive Figure 3: Figure renal and Since central nervous system (CNS), system during the first nine days of admission. patients could not be admitted with this > 1 day, for had to be present failure abdominal organ and the absorbing states death failure of organ The last panel shows the level failure. type of organ was defined as ≥ 3 individual failing organ on the patient level. For this panel, failure and discharge systems, and at risk systems or ≥ 3 dysfunctioning organ systems, dysfunction as 1 or 2 failing organ indicating better performance). 0.19 0.49 0.33 0.21 0.59 0.22 0.51 0.01 0.90 0.98 0.32 0.036 0.002 0.015 0.027 <.001 <0.001 <0.001 <0.001 <0.001 <0.001 P-value 11 (8) 61 (46) 20 (15) 41 (31) 24 (18) 30 (23) 41 (31) 20 (15) 24 (18) 65 (49) 38 (29) 45 (34) 18 (14) 32 (24) 82 (62) 66 (50) 92 (69) 93 (70) 103 (77) 130 (98) 122 (92) 10 (4-18) 63 (52-71) 12 (11-15) 112 (95-130) 6.2 (4.1-10.7) 225 (123-293) N=133 Failure 67 (6) 7 (3-12) 8 (7-10) 298 (26) 217 (19) 318 (27) 316 (27) 192 (16) 171 (15) 853 (73) 514 (44) 676 (58) 125 (11) 276 (24) 355 (30) 132 (11) 403 (35) 709 (61) 247 (21) 683 (59) 839 (72) 939 (81) 1,122 (96) 63 (53-71) 84 (69-102) 2.7 (1.7-4.5) 189 (101-296) N=1,166 Dysfunction 5 (5) 6 (6) 6 (6) 5 (4-7) 26 (26) 28 (28) 33 (33) 23 (23) 15 (15) 78 (78) 42 (42) 68 (68) 21 (21) 21 (21) 24 (24) 49 (49) 61 (61) 10 (10) 53 (53) 65 (65) 86 (86) 75 (75) 5 (3-12) 63 (48-73) 70 (60-87) 2.0 (1.3-2.9) 118 (75-209) At risk N=100 a d b c e Pulmonary Abdomen Urinary tract Other or unknown Gram-positive bacteria Gram-negative bacteria and fungi Yeasts Other/unknown Diabetes mellitus disease Cardiovascular Temperature Leukocytes Respiratory rate Heart rate Renal insufficiency Renal insufficiency Respiratory insufficiency Immunodeficiency ariable ariable Predisposition, infection, response, and organ failure (PIRO) characteristics failure and organ response, infection, able 3: Predisposition, Admission type, medical Insult (hospital-acquired) Source system Site / organ Pathogen Response SIRS criteria Predisposition Age Male gender Male gender comorbidities Chronic T V Outcome ICU case fatality day ICU length of stay, C-reactive protein C-reactive Lactate dysfunction Organ at admission score SOFA IV score APACHE
Predicting treatment response in patients with sepsis 9 144 Predicting treatment response in patients with sepsis 145 9 . 5 /L or /L or 9 . The systemic inflammatory response syndrome (SIRS) that defines the (SIRS) that defines the response syndrome . The systemic inflammatory 10,11 With our approach we aimed to overcome some of the shortcomings of previous some of the shortcomings of previous With we aimed to overcome our approach In our cohort the rate of progression to worse outcome (i.e. failure or death) was or death) was failure to worse outcome (i.e. In our cohort the rate of progression Respiratory insufficiency was defined as chronic obstructive pulmonary disease or chronic disease or chronic obstructive pulmonary as chronic was defined Respiratory insufficiency Cardiovascular disease was defined as cerebrovascular disease or chronic cardiovascular cardiovascular or chronic disease disease was defined as cerebrovascular Cardiovascular Immunodeficiency was defined as having acquired immune-deficiency syndrome, the use of the use immune-deficiency syndrome, defined as having acquired Immunodeficiency was able 3: Continued Renal insufficiency was defined as chronic renal insufficiency (creatinine > 177 µmol/L) or chronic chronic or (creatinine > 177 µmol/L) renal insufficiency was defined as chronic Renal insufficiency Systemic inflammatory response syndrome criteria were defined as: temperature < 36.0 or > < 36.0 or defined as: temperature criteria were response syndrome Systemic inflammatory insufficiency (New York Heart Association class 4), chronic congestive heart failure (ejection (ejection heart failure congestive class 4), chronic York Heart Association (New insufficiency intermittens, disease (claudicatio peripheral vascular or 30%) < fraction percutaneous patients with or bypass for arterial insufficiency). transluminal angioplasty dialysis. dialysis. d e b c T median or numbers (percentage) Data are at admission. failure the severity of organ Data indicates range). (inter-quartile a > 10% immature (band) forms; heart rate > 90/min during at least one hour; respiratory rate > 20/ forms; (band) during at least one hour; respiratory heart rate > 90/min > 10% immature pCO2 < 32 mm Hg, or mechanical ventilation. min during at least one hour, Unit; SIRS Systemic Health Evaluation; ICU Intensive Care Acute Physiology and Chronic APACHE Inflammatory Response Syndrome; corticosteroids in high doses (equivalent to prednisolone of > 75 mg/day for at least one week), one week), of > 75 mg/day for at least in high doses (equivalent to prednisolone corticosteroids hematologic drugs recent use of antineoplastic, drugs, current use of immunosuppressive current humoral or cellular deficiency. or documented malignancy, respiratory insufficiency with functional disabilities (chronic mechanical ventilation, oxygen use at ventilation, oxygen use at mechanical disabilities (chronic with functional insufficiency respiratory pulmonary hypertension). home, or severe 38.0°C during at least two and one hour, respectively; white blood cell count < 4 or > 12*10 respectively; 38.0°C during at least two and one hour, We found that especially higher age, immunodeficiency, and lactate were predictive predictive and lactate were that especially higher age, immunodeficiency, found We infection or most likely causing pathogens. In of worsening, in contrast to the site of markers of disease severity such as SAPS contrast to prior studies, we did not include only known after the first 24 hours of admission (but since they are scores or APACHE not suitable therefore, purposes) and are, often calculated much later for research are or inclusion of patients into clinical trials. for ‘bedside’ prediction sepsis and septic shock. Staging the to severe studies determining the progression and septic sepsis, stages of sepsis, severe to infection into the three host response in system classification preferred the been has and acceptance wide gained has shock prior studies Discussion for use in critically ill changes in disease severity a model to predict developed We and calculates the risks of death, discharge, patients admitted for sepsis. Our model can be updated whenever and patients, individual sepsis for failure organ persistent new information becomes available. the probabilities thereafter: and decreased highest during the first days of admission To 14%, 7%, and 5% at days 3, 5, and 7. were or death to failure of progression model PIRO the from predictors evaluated we disease, of evolution the examine presence of sepsis is, however, often criticized of being overly sensitive and non-specific often criticized of being overly sensitive and non-specific of sepsis is, however, presence
to the previous measurement, the variable was set to 0. to set was variable the measurement, previous the to
variable was set to 0. C-reactive protein levels were dichotomized using a cutoff of 100 mmol/l. When levels decreased with more than 50 mmol/l compared compared mmol/l 50 than more with decreased levels When mmol/l. 100 of cutoff a using dichotomized were levels protein C-reactive 0. to set was variable
Lactate levels were dichotomized using a cutoff of 3 mmol/l. When levels decreased with more than 3 mmol/l compared to the previous measurement, the the measurement, previous the to compared mmol/l 3 than more with decreased levels When mmol/l. 3 of cutoff a using dichotomized were levels Lactate
2.65 (1.37-5.13) 2.65 (0.51-1.00) 0.71 (1.45-3.45) 2.24 (1.54-3.18) 2.21 (0.59-1.00) 0.76 (0.64-1.40) 0.95 (0.71-2.13) 1.23 (1.08-2.01) 1.47 Lactate
0.93 (0.52-1.66) 0.93 (0.58-1.14) 0.82 (0.57-1.37) 0.88 (0.93-1.93) 1.34 (0.70-1.09) 0.88 (0.62-1.32) 0.90 (0.49-1.14) 0.75 (0.71-1.25) 0.94 protein C-reactive
Other/unknown 1.29 (0.85-1.96) 1.29 0.97 (0.46-2.04) 0.97 0.35 (0.17-0.71) 0.35 1.35 (0.95-1.93) 1.35 1.36 (0.82-2.26) 1.36 1.70 (0.80-3.58) 1.70 1.51 (0.92-2.46) 1.51 0.88 (0.38-2.05) 0.88
Urinary tract Urinary 0.65 (0.33-1.30) 0.65 0.73 (0.25-2.12) 0.73 1.41 (0.76-2.64) 1.41 0.85 (0.49-1.46) 0.85 0.94 (0.40-2.20) 0.94 1.54 (0.61-3.93) 1.54 1.40 (0.71-2.78) 1.40 0.64 (0.11-3.61) 0.64
Pulmonary 0.66 (0.45-0.96) 0.66 1.36 (0.74-2.50) 1.36 0.46 (0.29-0.71) 0.46 1.48 (1.08-2.03) 1.48 0.82 (0.50-1.32) 0.82 1.46 (0.75-2.86) 1.46 1.29 (0.81-2.05) 1.29 1.24 (0.60-2.54) 1.24
Site of infection (abdomen = ref) = (abdomen infection of Site
1.21 (0.65-2.25) 1.21 (0.61-1.30) 0.89 (0.83-2.11) 1.32 (1.15-2.40) 1.66 (0.66-1.14) 0.87 (0.76-1.68) 1.13 (0.90-2.17) 1.40 (0.84-1.54) 1.13 Immunodeficiency
1.68 (0.90-3.12) 1.68 (0.55-1.10) 0.78 (0.46-1.07) 0.70 (0.76-1.56) 1.09 (0.69-1.09) 0.86 (0.88-2.16) 1.37 (0.50-1.11) 0.74 (0.62-1.06) 0.81 ref) = (male Gender
> 70 > 0.85 (0.60-1.20) 0.85 0.84 (0.55-1.29) 0.84 0.75 (0.50-1.13) 0.75 1.25 (0.96-1.63) 1.25 0.78 (0.49-1.23) 0.78 1.74 (1.05-2.87) 1.74 1.13 (0.72-1.77) 1.13 2.50 (1.22-5.13) 2.50
60-70 1.90 (1.38-2.61) 1.90 1.41 (0.87-2.28) 1.41 0.22 (0.05-0.90) 0.22 1.57 (1.17-2.11) 1.57 0.93 (0.61-1.40) 0.93 1.19 (0.69-2.04) 1.19 1.37 (0.92-2.05) 1.37 1.77 (0.92-3.42) 1.77
Age (< 60 = ref) = 60 (< Age
eath D ysfunction D eath D Failure risk t A ischarge D ysfunction D ischarge D ariable V
t risk to risk t A ysfunction to ysfunction D Failure to Failure
Transition hazards for selected variables selected for hazards Transition able 4: 4: able T
Predicting treatment response in patients with sepsis 9 146 Predicting treatment response in patients with sepsis 147 9 . In another study, . In another study, 10 . Using this approach, they excluded all patients they excluded all patients . Using this approach, 14 . SIRS can also be caused by other conditions than infection than infection . SIRS can also be caused by other conditions 12,13 Unfortunately, there is no universal way of defining organ failure, and especially and especially failure, is no universal way of defining organ there Unfortunately,
the same group included critically ill patients with sepsis and predicted the evolution the evolution included critically ill patients with sepsis and predicted the same group sepsis and septic shock to severe in critically ill patients sepsis and shock, which accounted for 55% of all patients. These admitted with severe to multi-organ to progress especially prone however, ill patients are, severely more Furthermore, mortality. part explain morbidity and death and for a large or failure likely have other (possibly will most failure, patients with sepsis, but without organ to leading factors and ICU the to admission required that conditions serious) more factors. In addition, non-sepsis related, will be driven by other, of disease progression most at risk a clinical perspective it is important to identify those patients who are from to treatment. most likely to respond also those patients that are but for progressing, nowadays focus on the level Most clinical trials and epidemiological studies therefore shock. sepsis or septic i.e. on patients with severe for inclusion, failure of organ of organ degree remains ambiguous. This failure the definition of the level of organ is an important factor in the light of the much needed stratification of patients failure failure Within sepsis, organ the concept of severe of progression. their risks regarding subjective or but also on various other more scores, has been defined based on SOFA Modeled incidences of organ failure, death and discharge in four representative patients. patients. in four representative and discharge death failure, organ 4 Modeled incidences of Figure 3 to 1 Situations function. organ of evolution modeled average the represents patient average The 1 depicts the transitions of in four hypothetical patients. Situation denote the modeled situations pneumonia, for a community-acquired male patient, admitted a 72-year old immunocompromised with but patient, same the represents 2 Situation lactate. and protein C-reactive increased with healthy a 53-year old previously and lactate. Situation 3 represents protein normal C-reactive female patient with a urinary tract infection. such as sterile pancreatitis, trauma or surgery. Predictors of worsening in patients in patients of worsening Predictors trauma or surgery. such as sterile pancreatitis, depending Furthermore, patients with infection. from without infection might differ of the definition of SIRS, almost all patients will have SIRS at ICU on the restrictiveness Indeed, in their power. in a suboptimal predictive admission, possibly also resulting presence the between association an observe not did colleagues and Alberti study in critically ill patients or number of SIRS criteria and outcome . Second, although we . Second, although we 17 . We included readily available available included readily . We 15 . However, the authors also included less severely included less severely the authors also . However, 16 . The current study therefore examined the transitionsthe examined therefore study current The . 11,14 The strength of this model is that it may help clinicians to stratify critically ill patients of this model is that it may help clinicians to stratify critically ill patients The strength Most studies only considered patient or infection characteristics on the first day of patient or infection characteristics on Most studies only considered from one stage of organ failure to another, and the factors associated with worsening or and the factors associated with to another, failure of organ one stage from in the first seven days of admission only. improvement may at initial presentation, of progression to their probability with sepsis according and early interventions in high-risk aggressive assist in decision making for initiation of this model can be used studies. Furthermore patients, or inclusion in experimental of in the design which may help treatments of (hypothetical) effects simulate the to value of new biomarkers. We prognostic new sepsis trials or to estimate the added First, this study was performed in our study. also acknowledge some limitations of general ICU practice. In particular, not reflect two centers in the Netherlands and may to ours may of sepsis compared strategy for treatment a very different using centres methods care units used standardized Yet, both intensive care results. find disparate guidelines and followed the Surviving Sepsis Campaign ill patients from non-ICU wards in whom predictions of treatment response might be verybe might response of treatment predictions in whom wards non-ICU from patients ill patients for the ill. Most prior studies followed severely to the more compared different such as as outcome. By doing this, outcomes ICU mortality admission and reported entire not be related might occurring much later during follow-up failure death or multi-organ In our cohort, most to the ICU anymore. of infection at presentation to the presence within the first week, which is (92%) occurred failure of the deaths (58%) or multi-organ studies previous with consistent and objective markers for organ failure. In addition, whenever possible, our classification our classification whenever possible, In addition, failure. markers for organ and objective causes non-sepsis related excluding (thereby failure organ the etiology of also included the duration of the dysfunction. also considered we of shock). Furthermore, thein changes daily capture to or covariables time-varying include to failed and infection, patients. rapidly in this population of critically ill which might change failure, levels of organ over progression covariables to estimate the risk of study included time-varying A recent with infection the first week in patients evaluated the goodness-of-fit of our model, it would be useful to validate our model model, it would be useful to validate our model evaluated the goodness-of-fit of our on an external database of critically ill patients with sepsis. Conclusion to according for sepsis admitted patients ill critically stratify to a model propose We and which can be used to identify high-risk patients who will likely response, treatment and early interventions. This model can also be possibly aggressive benefit most from which will help in the design of (hypothetical) treatments used to simulate the effects value of new biomarkers. of new sepsis trials or to estimate the added prognostic less frequently occurring symptoms (e.g. skin mottling) symptoms (e.g. occurring less frequently
Predicting treatment response in patients with sepsis 9 148 Predicting treatment response in patients with sepsis 149 9 Am J Respir Respir J Am Jan 1-7 2005;365(9453):63-78. Lancet. May 2012;38(5):811-819. Mar 1 2005;171(5):461-468. Leon AL, Hoyos NA, Barrera LI, et al. Clinical LI, et al. Clinical Leon AL, Hoyos NA, Barrera and septic sepsis, course of sepsis, severe from infected patients of a cohort shock in ten Colombian hospitals. BMC Infect Dis. 2013;13:345. Levy MM, Rhodes A, et al. Dellinger RP, Surviving Sepsis Campaign: international sepsis guidelines for management of severe Med. and septic shock, 2012. Intensive Care Feb 2013;39(2):165-228. for Statistical Computing. Available from: from: Available for Statistical Computing. . 2013. http://www.r-project.org/ C, Goodman SV, Alberti C, Brun-Buisson inflammatory et al. Influence of systemic and sepsis on outcome syndrome response patients. infected ill critically of Med. Jul 1 2003;168(1):77-84. Crit Care Rangel-Frausto MS, Pittet D, Hwang T, dynamicsThe RP. Wenzel RF, Woolson in sepsis: Markov of disease progression modeling describing the natural history and antisepsis agents. the likely impact of effective Clin Infect Dis. Jul 1998;27(1):185-190. Vincent JL. Dear SIRS, I’m sorry to say Med. Feb like you. Crit Care that I don’t 1997;25(2):372-374. PM, Ong DS, Bonten Klein Klouwenberg OL. Classification of sepsis, MJ, Cremer impact the shock: septic and sepsis severe and of minor variations in data capture definition of SIRS criteria. Intensive Care Med. Alberti C, Brun-Buisson C, Chevret S, et S, et Alberti C, Brun-Buisson C, Chevret response and al. Systemic inflammatory sepsis in critically to severe progression ill infected patients. Am J Respir Crit Care Med. Annane D, Bellissant E, Cavaillon JM. Septic shock. 16. 17. 10. 11. 12. 13. 14. 15.
Jan 11 1995;273(2):117-123. 2011;38(8). Jun 1992;101(6):1644-1655. JAMA. Jul 1996;22(7):707-710. Angus DC, van der Poll T. Severe sepsis sepsis Severe T. Angus DC, van der Poll Aug 29 and septic shock. N Engl J Med. 2013;369(9):840-851. FB, et al. Definitions Bone RC, Balk RA, Cerra and guidelines failure for sepsis and organ therapies in sepsis. for the use of innovative The ACCP/SCCM Consensus Conference Committee. American College of Chest Medicine. Physicians/Society of Critical Care Chest. Rangel-Frausto MS, Pittet D, Costigan The RP. Davis CS, Wenzel M, Hwang T, natural history of the systemic inflammatory (SIRS). A prospective syndrome response study. PM, Ong DS, Bos LD, et Klein Klouwenberg of Centers for agreement al. Interobserver criteria for and Prevention Disease Control classifying infections in critically ill patients. Med. Oct 2013;41(10):2373-2378. Crit Care Marshall JC, et al. 2001 Levy MM, Fink MP, International SCCM/ESICM/ACCP/ATS/SIS Intensive Sepsis Definitions Conference. Med. Apr 2003;29(4):530-538. Care J, et al. The SOFA R, Takala Vincent JL, Moreno Assessment) Failure Organ (Sepsis-related dysfunction/failure. to describe organ score on Sepsis- Group On behalf of the Working Society of the European Related Problems Medicine. Intensive Care of Intensive Care Med. Marshall JC, Ramsay G, Nelson Rubulotta F, D, Levy M, Williams insult/ M. Predisposition, dysfunction: and organ infection, response, sepsis. Crit severe for staging model new A Med. Apr 2009;37(4):1329-1335. Care Jackson CH. Multi-State Models for Panel Stat R. J for msm Package Data: The Software. R. R: A language and environment Team for statistical computing. R Foundation References 1. 2. 3. 4. 5. 6. 7. 8. 9. s of support Source (http:// Molecular Medicine the Center for Translational was supported by This work funding from research MB has received MARS (grant 04I-201). project www.ctmm.nl), (NWO Vici 918.76.611). of Scientific Research the Netherlands Organization
PART IV
DISCUSSION AND SUMMARY
10 General discussion
Peter M.C. Klein Klouwenberg
General discussion 155 10 . In 1991 the current definitions definitions . In 1991 the current 6 . Its aim is to generate tools that . Its aim is to generate tools that , although the mortality related to to related mortality the although , 1 4,5 . In the developed world the incidence of incidence of world the . In the developed 1,2 . 3 . However, despite the deceptive simplicity of this of this despite the deceptive simplicity . However, 8 . According to these definitions, the diagnosis of sepsis to these definitions, the diagnosis of sepsis . According 7 Patient populations with sepsis are heterogeneous with regard to the implicated to the implicated with regard heterogeneous with sepsis are Patient populations This thesis described some of the studies that were performed within the framework that were This thesis described some of the studies requires clinical suspicion of infection and the presence of a systemic inflammatory inflammatory of a systemic of infection and the presence clinical suspicion requires (SIRS) that is characterized by specific physiological alterations, syndrome response white blood cell count, heart rate and respiratory including aberrations in temperature, and septic shock as failure, organ sepsis was defined as sepsis-related rate. Severe has been very sepsis. This classification of severe shock in the presence refractory with useful in epidemiologic studies or clinical trials, and showed good correlation outcomes such as mortality pathogen, the underlying infectious disease and their genetic profile. Current tools tools Current profile. infectious disease and their genetic pathogen, the underlying mainly comprise of symptom to manage patients with sepsis available to the clinician delayed non-specific and techniques, which provide and culture classification systems outcome of the to improve informationabout the host and pathogen. In order in be approached patient population needs to heterogeneous sepsis, this highly number and in the large clinical practice those used in current from ways different stratified be to need patients individual particular, In trials. sepsis clinical (failed) of to their risk of according i.e. homogeneous subgroups, in clinically meaningful more the diagnosis complications. Furthermore, dying or development of specific sepsis and the definitions of sepsis should rapidly, of infection should be performed more the Molecular Diagnosis and Risk be described. For these reasons, accurately more study was started Stratification of Sepsis (MARS) data collection initiative and biobank units in 2010 in two Dutch tertiary intensive care provide rapid and accurate information rapid the insulting pathogen and the severity about provide of the critically ill patient with sepsis. and the stage of the immune response in diagnosis that exist described the difficulties of this consortium. In the first part, we sepsis. In the second part, we described some important in patients with presumed the for possible solutions ill patients and provided critically in sepsis of complications of patients with sepsis. prognostication improved sis of sepsis Challenges in the diagno Inflammatory Response Syndrome Establishment of the Systemic was an attempt to define in the late 1980s of the “sepsis syndrome” The introduction an at-risk population for inclusion in clinical studies Sepsis is a major and increasing cause of in-hospital morbidity and mortality. It is the It is the and mortality. in-hospital morbidity cause of major and increasing Sepsis is a worldwide cause of death second leading of sepsis, severe sepsis, and septic shock were established through the consensus the consensus established through sepsis, and septic shock were of sepsis, severe of investigators of a group sepsis has decreased since then since then sepsis has decreased sepsis showed an annual increase of 9% and a tripling of the number of deaths related related number of deaths and a tripling of the of 9% an annual increase sepsis showed 2000 until population 100,000 per 44 to sepsis to . An expanded . An expanded 9 . However, many authors many authors . However, 10,11 describes the results of our systematical evaluation of the usability usability of our systematical evaluation of the Chapter 3 describes the results . 11-16 ). We show that the measured incidences of SIRS in patients admitted to to in patients admitted incidences of SIRS measured show that the ). We Chapter 2 of CDC criteria to diagnose infections in general in critically ill patients. Although in general in critically ill patients. Although of CDC criteria to diagnose infections we found substantially was high, agreement overall the observed interobserver in particular ventilator- for some subtypes of infection, of concordance levels reduced the CDC infection definitions are In practice, however, associated pneumonia (VAP). in critically ill patients of infection used by clinicians to diagnose the source infrequently sepsis. Instead, they will often start antibiotic therapy empirically for with presumed disease, even when the ill patients with life threatening suspected infection in severely of a Chapter 4 describes the results to be low. plausibility of infection is considered in study investigating the accuracy of infection diagnosis made by bedside clinicians than 40% of all critically infection at admission. More critically ill patients with presumed possibly ill patients had a low likelihood of infection in post-hoc classification, and mortality Unexpectedly, therapeutic antibiotics unnecessarily in retrospect. received diagnosis. was higher in the population with a low likelihood of infection in post-hoc mortality rate behind this increased the reasons Our design was not suited to explore speculative remains in patients with a post-hoc diagnosis of “no infection”. It therefore of due to delayed recognition treatment whether this is caused by a delay in effective The establishment of the other, but equally important, part of the definition, the the but equally important, part of the definition, The establishment of the other, In clinical practice, the identification either. infection, is not straightforward of presence source, the localization of its presumed of infection as the primary cause of disease, to identify its likely causative micro- results of microbiological and the interpretation situation, Centers for this improve to be complicated. To been proven have organism developed in 1988 were (CDC) diagnostic criteria and Prevention Disease Control the ICU population and later specifically modified for Establishment of infection concept, the application of the SIRS criteria in diagnosing sepsis is not straightforward not straightforward sepsis is SIRS criteria in diagnosing application of the concept, the ( or on how restrictively 49 to 99%, depending from ICUs varied hugely two academic sepsis and the incidences of sepsis, severe Subsequently, liberally SIRS was measured. 4 to 9% for the most 6 to 27%, and from to 31%, from 22 from septic shock ranged 39 to 98% Yet, settings, respectively. liberal measurement the most versus restrictive of patients without a considerable number had SIRS, whereas of non-infected patients and sensitive, overly both are criteria SIRS the that showing infection, an had SIRS still soon after recognized were shortcomings in the sepsis definitions non-specific. These updated one decade later were and the definitions the conference list of signs and symptoms was introduced to better describe the clinical response to to response to better describe the clinical was introduced list of signs and symptoms ileus, and other markers capillary refill, of decreased the presence infection, including describe sepsis as a Although taken together these criteria better failure. of organ and interrater variability. subject to interpretation more clinical disease, they are have suggested that they are too complicated, subjective, and imprecise for practical for practical too complicated, subjective, and imprecise have suggested that they are use
General discussion 156 10 General discussion 157 10
. 12,15,17-19 . This algorithm 20-22 . In addition, patients with Gram-negative infections. In addition, patients with Gram-negative 23 . Currently available tools for prognostication comprise only symptom comprise only available tools for prognostication . Currently 24-26 describes a model that predicts, for an individual patient with for an individual patient with 9 describes a model that predicts, Chapter In patients admitted to the ICU with recent onset of organ dysfunction or shock, dysfunction or shock, onset of organ In patients admitted to the ICU with recent might have similar levels of endotoxin as patients with noninfectious acute disorders such patients with noninfectious acute disorders might have similar levels of endotoxin as or as the general population of critically ill patients at the time of as acute pancreatitis ICU admission describes the results of a study assessing the feasibility and validity of a novel assessing the feasibility and validity of of a study Chapter 5 describes the results in introduced that was recently events (VAE) for ventilator-associated surveillance system of the conventional diagnosis of VAP a replacement the United States as identifies ventilator-associated conditions (VAC), infection-related VAC (IVAC), and VAP. VAP. and VAC (IVAC), infection-related conditions (VAC), identifies ventilator-associated it to in two academic ICUs and compared implemented this algorithm electronically We algorithm detected at The VAE study. within the MARS registration VAP the prospective VAC signals were surveillance. patients identified by prospective most 32% of the VAP In addition, infections, but not necessarily VAP. most often caused by volume overload and varied considerably. associations of the events and mortality with sepsis The heterogeneity of the critically ill patient population with sepsis makes it extremely of the critically ill patient population with sepsis makes it extremely The heterogeneity medical decision-making. For instance, patients challenging to initiate appropriate patterns of might have very different meeting similar criteria for sepsis syndrome inflammatory mediators circulating s of patient stication Challenges in the progno response treatment Predicting Concerns with the reliability of the diagnosis of VAP have initiated the development of have initiated the of the diagnosis of VAP Concernsthe reliability with implementation amenable to electronic that are objective case-definitions more an alternativean the alternative,or simply that disease, these in diagnosis non-infectious, infectious diagnosis. rejected injurious than the even more patients is infection Surveillance of classification systems and conventional microbiological techniques. The first are only techniques. The first microbiological classification systems and conventional not for individual patients, populations and in large validated for outcome predictions diagnostic delays of at least 24 hours.and the second have low sensitivities and abnormalities is the main of these —potentially still reversible— averting progression - at the bedside for clinicians to predict it is very difficult However, goal of critical care. interventions, and who will favorably to critical care - which patients will respond resuscitative or death despite failure multi-organ deteriorate to develop prolonged efforts. sepsis, the probabilities of death, discharge and the development of persistent critical and the development of persistent critical of death, discharge sepsis, the probabilities routinely available parameters. This illness during the first week of admission, using their to with sepsis according stratify critically ill patients help clinicians to model may may assist in decision making for at initial presentation, of progression probability
. These inconsistencies have been . These inconsistencies have been . However, there is conflicting data is conflicting data there . However, 32-43 27-31 describes the incidence of atrial fibrillation in critically ill patients Chapter 8 describes the incidence of atrial fibrillation in critically ill patients . Deficiencies in modelling methodology and residual confounding may, residual confounding may, . Deficiencies in modelling methodology and 44 explained by differences in case-mix, tools used for delirium assessment, and study used for delirium assessment, and study tools in case-mix, differences explained by design Complications of sepsis of critical illness, especially in patients with Delirium is a common complication sepsis, occurring in 30-60% of patients Sepsis and the anti-inflammatory response Sepsis and the anti-inflammatory episode of sepsis, patients of patients nowadays survive the initial Although the majority enter a state of frequently the first days after ICU admission with sepsis who survive them vulnerable for secondary ICU- render that may immune suppression prolonged and that immune suppression it has been suggested infections. Moreover, acquired important denominators are occurring as a consequence thereof secondary infections immune stimulation has in the ICU. In light of these findings, of late sepsis mortality a paradigm shift from strategy for sepsis, representing been suggested as a treatment in many unsuccessful sepsis trails. Chapter 6 the use of anti-inflammatory agents infections in patients mortality of ICU-acquired describes the incidence and attributable admission at of sepsis presence the that show admission. We sepsis at with and without infections, which contradicts the ICU-acquired does not influence the incidence of prolonged in a more that sepsis in critically ill patients results hypothesis current than non-septic disease. Instead, the incidence of ICU- period of immunosuppression or shock at admission, failure in patients with organ was increased infections acquired as stay length of ICU increased resulting of disease and the the severity suggesting was infections mortality of ICU-acquired most influencing factors. The attributable to those at admission compared sepsis in patients without higher 13% by day 60, but included patients with analyses that was consistent in subgroup with sepsis. This result low attributable mortality suggests disease at admission. The relatively severe more to mortality. more and shock) contribute failure that other factors (e.g., organ initiation of aggressive and early interventions in high-risk patients, or inclusion in in patients, or inclusion in high-risk and early interventions of aggressive initiation the effects used to simulate this model can be studies. Furthermore experimental or to of new sepsis trials help in the design which may treatments of (hypothetical) value of new biomarkers. prognostic estimate the added about the mortality caused by delirium about the mortality caused by delirium with sepsis. Theoretically, the systemic release of pro-inflammatory cytokines, high cytokines, high of pro-inflammatory the systemic release with sepsis. Theoretically, however, provide an alternative we estimated the mortality explanation. In chapter 7 provide however, by due to delirium in critically ill patients while taking into account bias caused risk time-varying disease severity until the onset of delirium and by the competing the population attributable ICU mortality was Using this approach, of discharge. reduced estimated at 7.2% by day 30, implying that absolute case fatality can be in all patients. delirium able to completely prevent than 0.9% if we were by no more Likewise,
General discussion 158 10 General discussion 159 10 and 9,47 are all plausible risk all plausible risk are 45 . 47,48 . Among 1,786 patients with sepsis, total of 1,076 of 1,076 total sepsis, patients with . Among 1,786 46 . It is a notable effort to stratify patients by combining to stratify patients by combining . It is a notable effort 9,49-52 We need to perform prospective cohort studies to develop and validate methods cohort studies to develop and validate methods need to perform prospective We most at risk of developing to stratify critically ill patients with sepsis who are The PIRO model might serve as an initial template failure. complications or organ for such an initiative predisposition and other patient characteristics, infection characteristics, clinical and other patient characteristics, infection characteristics, clinical predisposition analogous to the TNM system failure, signs and symptoms, and the level of organ the PIRO model only includes generally available currently However, in cancer care. stratify patients making it less suitable characteristics and markers that insufficiently The current straightforward syndrome-based consensus definition of sepsis does consensus definition of sepsis does syndrome-based straightforward The current this complex disease and is not useful in clinical practice, nor not fully capture objective, efficient develop more should therefore value. We has good prognostic but on specific not based on clinical symptoms definitions, which are and reliable associated with adverse outcome. biochemical profiles of infection. The current methods to establish the presence improve should We of infection is solely based on clinical signs and identification of the source to be complicated and and has been proven results symptoms and microbiological detect pathogens and antimicrobial time-consuming. New assays that efficiently genes should be developed resistance factors for the development of AF the development factors for 3. spectives Future per anddiagnosis of infection the in challenges important explored we thesis, this In in of critically ill patients with sepsis. Because the clinical syndrome prognostication non-specific signs and symptoms, the severity ofpatients with sepsis is characterized by of site pathogen, the of virulence the including factors, many by influenced be may which evolution of the condition temporal comorbidity, the infection, host susceptibility, levels of circulating stress hormones, intravascular volume dysfunction, autonomic stress circulating levels of during sepsis that occur compromise cardiovascular shifts and AF episodes occurred in 410 (23%) individuals. The cumulative risk of new-onset AF of new-onset AF (23%) individuals. The cumulative risk in 410 AF episodes occurred severe sepsis, patients with (35-43%) in 39% 21% (18-24%), CI 8-13%), (95% was 10% for imbalances in baseline After correction shock, respectively. sepsis and septic reduced a both with associated remained AF new-onset severity, disease of markers 2-fold rate, accounting for an approximately death rate and an increased discharge risk. as a competing when considering ICU discharge mortality risk overall increased concurrent therapeutic interventions, it is not surprising that the current approach of the approach therapeutic interventions, it is not surprising that the current concurrent advances have been made in terms great management of sepsis is suboptimal. Although and the pathophysiological events that contribute to of understanding the host response unacceptably high. I in patients with sepsis, mortality rates still remain dysfunction organ success: the possibility of on these key issues to increase to focus propose would therefore 1. 2. or who may benefit or who may benefit . 47,53-55 59,60 http://www.who.int/whr/2004/ . 56-58 stratum in WHO regions estimates for estimates for stratum in WHO regions 2004; 2002. annex/topic/en/annex_2_en.pdf. Accessed 2014. October, Kaukonen KM, Bailey M, Suzuki S, Pilcher D, sepsis severe to Bellomo R. Mortality related 10.
Apr 17 2003;348(16):1546-1554. We need to organize international clinical trials using the knowledge from the host the host international clinical trials using the knowledge from organize need to We critically ill patients with sepsis. These trials should use entry criteria of response instead of clinical signs and symptoms common biochemical profile based on a optimally will that population patient a selecting of the likelihood increase to We should focus on both short-term and long-term the intervention. benefit from morbidity and mortality. long-termthe study to need sepsis, with associated mortality and morbidity We should management and interventions. We its complications, and current of therapy as this might alter in the continuously gauge the cost-effectiveness changing therapeutic environment. from interventional therapies. Specifically, host biomarkers that discriminate host biomarkers that discriminate Specifically, interventional therapies. from in the ICU and that and non-infectious causes of critical illness between infectious needed urgently therapy are the use of antibiotic The host susceptibility for infection and the predisposition to develop organ failure failure organ to develop for infection and the predisposition The host susceptibility of patients needs to be determined by genetic profiling for use in clinical practice. Only with the addition of more sophisticated markers it it sophisticated markers of more with the addition clinical practice. Only for use in model. a fully effective to develop into the prospect may have We is inadequate. in sepsis patients response of the host Our understanding discover suitable the aim to pathways with to disentangle the biochemical need who are These biomarkers should identify patients biomarkers for prognostication. at an early stage failure organ most at risk of developing WHO. The world health report, changing changing WHO. The world health report, history: Deaths by cause, sex and mortality Martin GS, Mannino DM, Eaton S, Moss M. M. Moss S, Eaton DM, Mannino GS, Martin The epidemiology of sepsis in the United 2000. N Engl J 1979 through States from Med. 6. 7. 9. References 8. Concluding remarks patient outcome should be sought both in renewed The key to make a change and improve and clinically applicable, as well as in a betterdefinitions of sepsis that should be accurate patients of critically ill group heterogeneous within the larger identification of subgroups adequate and distinct associations with patient outcome. This will require that have more to account for the various challenges that currently advanced methodological approaches as an important estimations. The work in this thesis can be considered accurate prohibit sepsis. outcomes of patients with keys to improve to and provides step forward 5. 4.
General discussion 160 10 General discussion 161 10 PLoS One. 2011;6(3):e18062. Oct 1999;54(10):867-873. Nov 2013;41(11):2467-2475. Allami MK, Jamil W, Fourie B, Ashton V, B, Ashton V, Fourie Jamil W, Allami MK, in incisional infection PJ. Superficial Gregg limb. Interobserver of the lower arthroplasty criteria. J diagnostic of the current reliability Sep 2005;87(9):1267-1271. Br. Bone Joint Surg variability in Klompas M. Interobserver pneumonia surveillance. ventilator-associated Apr 2010;38(3):237-239. Am J Infect Control. E, WilsonAshby E, Haddad FS, O’Donnell AP. site infection be measured How will surgical for all”? J Bone “high quality care to ensure Sep 2010;92(9):1294-1299. Br. Joint Surg et al. A, N, Ewig S, Torres Fabregas ventilator associated Clinical diagnosis of validation comparative revisited: pneumonia lung biopsies. using immediate post-mortem Thorax. PL, Cook Fakhry SM, Ferguson Michetti CP, R. Ventilator-associated FO, Gross A, Moore pneumonia rates at major trauma centers with a national benchmark: a multi- compared J Trauma institutional study of the AAST. May 2012;72(5):1165-1173. Surg. Acute Care Klompas M. Eight initiatives that misleadingly rates. pneumonia lower ventilator-associated Jun 2012;40(5):408-410. Am J Infect Control. al. et K, Kleinman Y, Khan M, Klompas Multicenter evaluation of a novel surveillance paradigm for complications of mechanical ventilation. National Healthcare Safety Network (NHSN). Safety Network (NHSN). National Healthcare Safety Network The National Healthcare device-associated module: ventilator- Atlanta, GA: associated event protocol. and Prevention; Centers for Disease Control July 2013. Magill SS, Klompas M, Balk R, et al. Developing to surveillance approach national new, a events*. Crit Care for ventilator-associated Med. Casey LC, Balk RA, Bone RC. Plasma cytokine Casey LC, Balk RA, Bone RC. Plasma cytokine survival with correlate levels endotoxin and Ann in patients with the sepsis syndrome. Intern Med. Oct 15 1993;119(8):771-778. D, Hiki N, et al. Buttenschoen K, Berger in antibodies antiendotoxin and Endotoxin Eur J Surg. patients with acute pancreatitis. Jun 2000;166(6):459-466. Marshall JC, Foster D, Vincent JL, et al. implications Diagnostic and prognostic of endotoxemia in critical illness: results J Infect Dis. Aug 1 of the MEDIC study. 2004;190(3):527-534. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. Jun Jun Crit Care Care Crit Am J Infect Control. Control. Infect J Am Jan 11 1995;273(2):117-123. Apr 2 2014;311(13):1308-1316. Jun 1992;101(6):1644-1655. JAMA. Apr 2003;31(4):1250-1256. May 2012;38(5):811-819. Levy MM, Fink MP, Marshall JC, et al. 2001 Marshall JC, et al. 2001 Levy MM, Fink MP, International SCCM/ESICM/ACCP/ATS/SIS Conference. Definitions Sepsis Med. Garner Emori TG, Horan TC, JS, Jarvis WR, Hughes JM. CDC definitions for nosocomial 1988. infections, 1988;16(3):128-140. Cohen J. The international sepsis Calandra T, definitions on forum consensus conference unit. Crit of infection in the intensive care Med. Jul 2005;33(7):1538-1548. Care Horan TC, Andrus M, Dudeck MA. CDC/ NHSN surveillance definition of health care- associated infection and criteria for specific setting. types of infections in the acute care Jun 2008;36(5):309-332. Am J Infect Control. BC, et al. Gibbons C, Reeves Wilson AP, performance wound infection as a Surgical of common definitions indicator: agreement of wound infection in 4773 patients. Bmj. Sep 25 2004;329(7468):720. Klein Klouwenberg PM, Ong DS, Bonten PM, Ong DS, Bonten Klein Klouwenberg OL. Classification of sepsis, MJ, Cremer impact the shock: septic and sepsis severe and data capture of minor variations in Intensive Care definition of SIRS criteria. Med. PM, Ong DS, Bos LD, et Klein Klouwenberg for of Centers agreement al. Interobserver for criteria and Prevention Disease Control critically ill patients. classifying infections in Med. Oct 2013;41(10):2373-2378. Crit Care Slotman Clemmer TP, Bone RC, Fisher CJ, Jr., Sepsis syndrome: GJ, Metz CA, Balk RA. Methylprednisolone a valid clinical entity. Med. Crit Care Group. Sepsis Study Severe May 1989;17(5):389-393. Bone RC, Balk RA, Cerra FB, et al. Definitions and guidelines failure for sepsis and organ for the use of innovative therapies in sepsis. The ACCP/SCCM Consensus Conference Committee. American College of Chest Medicine. Physicians/Society of Critical Care Chest. Rangel-Frausto MS, Pittet D, Costigan The RP. Davis CS, Wenzel M, Hwang T, natural history of the systemic inflammatory (SIRS). A prospective syndrome response study. and septic shock among critically ill patients ill patients among critically and septic shock and New Zealand, 2000-2012. in Australia JAMA. 16. 17. 18. 19. 20. 11. 12. 13. 14. 15. Mar- Gen Hosp Psychiatry. Psychiatry. Hosp Gen Sep 2008;23(3):372-379. van den Boogaard M, Peters SA, van der van den Boogaard theon delirium of impact The al. et JG, Hoeven mortality in intensive of in-hospital prediction Aug 2010;14(4):R146. patients. Crit Care. care factors et al. Risk CY, Lin SM, Huang CD, Liu early-onset delirium for the development of clinical outcome in and the subsequent patients. J Crit mechanically ventilated Care. Deliriumet al. JM, Teles M, Soares JI, Salluh (DECCA): an care epidemiology in critical 2010;14(6):R210. Crit Care. international study. C, Salluh J, et al. CD, Grandi Tomasi and ICDSC for the Comparison of CAM-ICU in critically ill patients detection of delirium clinical outcomes. J focusing on relevant Apr 2012;27(2):212-217. Crit Care. N, Dumont M, Dial S, Dubois MJ, Bergeron unit: Delirium in an intensive care Y. Skrobik Med. a study of risk factors. Intensive Care Aug 2001;27(8):1297-1304. H, K, Kurosawa Takezawa Y, Iwasaki Kishi Y, unit patients Endo S. Delirium in critical care Gen room. an emergency admitted through Sep 1995;17(5):371-379. Hosp Psychiatry. Pun Peterson JF, Shintani A, Thomason JW, unit Intensive care Jackson JC, Ely EW. BT, of delirium is an independent predictor analysis longer hospital stay: a prospective of 261 non-ventilated patients. Crit Care. Aug 2005;9(4):R375-381. Zhang Z, Pan L, Ni H. Impact of delirium on clinical outcome in critically ill patients: a meta-analysis. Apr 2013;35(2):105-111. sepsis Severe Angus DC, van der Poll T. and septic shock. N Engl J Med. Aug 29 2013;369(9):840-851. Atrial fibrillation is not just an Launey Y. Seguin P, 2010;14(4):182. artefact in the ICU. Crit Care. Mackenzie I. Sepsis: definition, Lever A, and diagnosis. Bmj. epidemiology, 2007;335(7625):879-883. Becker K, et al. A Hinder F, Bloos F, culture blood multicenter trial to compare in severe with polymerase chain reaction Med. Feb human sepsis. Intensive Care 2010;36(2):241-247. Hunziker S, P, Schuetz D, MD, Talmor Howell the principle: of Proof NI. Shapiro AE, Jones organ infection, response, predisposition, Med. sepsis staging system. Crit Care failure Feb 2011;39(2):322-327. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. Intensive Intensive Feb 2012;344:e420. Feb BMJ. Feb 4 2009;301(5):489-499. Apr 14 2004;291(14):1753-1762. Mar 2013;57(3):288-293. JAMA. Care Med. Jan 2007;33(1):66-73. Care Delirium al. et B, Truman A, Shintani EW, Ely of mortality in mechanically as a predictor unit. ventilated patients in the intensive care JAMA. impact The al. et CH, Wang CY, Liu SM, Lin mechanically of survival the on delirium of Nov Med. ventilated patients. Crit Care 2004;32(11):2254-2259. TE,Murphy SV, Kasl SY, Kong MA, Pisani Ness PH. Days of delirium are Araujo KL, Van associated with 1-year mortality in an older unit population. Am J Respir Crit intensive care Med. Dec 1 2009;180(11):1092-1097. Care L, van M, Schoonhoven van den Boogaard Pickkers T, der Hoeven JG, van Achterberg Incidence and short-term consequences P. of delirium in critically ill patients: A Int study. observational cohort prospective J Nurs Stud. Jul 2012;49(7):775-783. Ouimet S, Kavanagh BP, Gottfried SB, Gottfried SB, Ouimet S, Kavanagh BP, risk factors and Incidence, Y. Skrobik delirium. ICU of consequences Shehabi Y, Chan L, Kadiman S, et al. Sedation Chan L, Kadiman S, et al. Shehabi Y, long-termand depth mechanicallyin mortality adults: a prospective ventilated critically ill cohort study. longitudinal multicentre Med. May 2013;39(5):910-918. Intensive Care I, Videbech P, Svenningsen H, Egerod EK. M, Tonnesen Christensen D, Frydenberg Fluctuations in sedation levels may contribute to delirium in ICU patients. Acta Anaesthesiol Scand. Peelen LM, et al. Intensive Zaal IJ, Spruyt CF, the course may affect unit environment care Mar Med. of delirium. Intensive Care 2013;39(3):481-488. al. et PM, Bokesch Y, Shehabi RR, Riker Dexmedetomidine vs midazolam for sedation of critically ill patients: a randomized trial. Bates DW, Parsonnet J, Ketchum PA, et et PA, Parsonnet J, Ketchum Bates DW, assay for amebocyte lysate al. Limulus with of endotoxin in patients detection AMCC Sepsis Project sepsis syndrome. Sep Clin Infect Dis. Group. Working 1998;27(3):582-591. Slooter M, Pickkers P, van den Boogaard and validation of AJ, et al. Development in DELIRium of (PREdiction PRE-DELIRIC model prediction ICu patients) delirium patients: observational for intensive care study. multicentre 40. 41. 42. 43. 39. 35. 36. 37. 38. 33. 34.
General discussion 162 10 General discussion 163 10 Jan 1 2014;5(1):154-160. Feb 6 2010;375(9713):463-474. 6 Feb Crit Care. 2010;14(1):R15. Crit Care. Virulence. Lancet. Tang BM, McLean AS, Dawes IW, Huang Huang IW, Dawes AS, McLean BM, Tang of profiling Gene-expression SJ, Lin RC. in sepsis. mononuclear cells peripheral blood Med. Mar 2009;37(3):882-888. Crit Care of sepsis. Curr Marshall JC. Biomarkers Infect Dis Rep. Sep 2006;8(5):351-357. Pierrakos C, Vincent JL. Sepsis biomarkers: a review. Rapid diagnosis of Reinhart K. Bloos F, sepsis. et al. Use of F, Bouadma L, Luyt CE, Tubach to patients’ exposure to reduce procalcitonin (PRORATA units care intensive in antibiotics randomised controlled trial): a multicentre trial. FO, Chapman SJ, et al. Khor CC, Vannberg CISH and susceptibility to infectious diseases. N Engl J Med. Jun 3 2010;362(22):2092-2101. Laramie JM, Meyer DJ, et al. Chung TP, Molecular diagnostics in sepsis: from Nov bedside to bench. J Am Coll Surg. 2006;203(5):585-598. 61. 62. 63. 64. 65. 66. 67. Rubulotta F, Marshall JC, Ramsay G, Nelson G, Nelson Marshall JC, Ramsay Rubulotta F, D, Levy M, Williams insult/ M. Predisposition, dysfunction: organ and infection, response, sepsis. Crit severe staging for model new A Apr 2009;37(4):1329-1335. Med. Care JE, et al. S, Hollander Trzeciak NI, Shapiro multicenter derivation of A prospective, assess risk of organ a biomarker panel to death in emergency dysfunction, shock, and suspected sepsis. department patients with Med. Jan 2009;37(1):96-104. Crit Care et al. Marshall JC, Vincent JL, Fink MP, a markers, and mediators: toward Measures, of sepsis. A report staging system for clinical Sepsis Roundtable, Toronto, the Fifth Toronto 25-26, 2000. Crit Ontario, Canada, October Med. May 2003;31(5):1560-1567. Care BM, Eslick GD, Craig JC, McLean Tang for sepsis AS. Accuracy of procalcitonin diagnosis in critically ill patients: systematic and meta-analysis. Lancet Infect Dis. review Mar 2007;7(3):210-217. 57. 58. 59. 60.
Nederlandse samenvatting
Sepsis is de afweerreactie van het lichaam op een infectie. Als deze reactie zeer sterk is, kan schade ontstaan aan weefsels en kunnen organen belemmerd worden in hun normale functie. Sepsis is een belangrijke oorzaak van ziekenhuismorbiditeit en wereldwijd gezien is het de één na belangrijkste oorzaak van overlijden. Er is daarom dringend behoefte aan een effectieve behandeling van sepsis, maar tot op heden is die er niet. Een groot probleem bij de ontwikkeling van een effectief medicijn is het gebrek aan adequate diagnose en de mate van afweerreactie in de heterogene groep patiënten met sepsis. Patiënten met sepsis verschillen namelijk sterk van elkaar wat betreft hun chronische co-morbiditeit, hun genetische profiel, en de ernst van ziekte. Daarnaast kan de infectie overal in het lichaam voorkomen en door vele verschillende ziekteverwekkers veroorzaakt worden. Een belangrijke vraag is daarom bijvoorbeeld welke sepsis patiënten geneesmiddelen moeten krijgen die de afweer stimuleren, en welke patiënten juist middelen die de afweer remmen. Daarnaast is de huidige diagnose van sepsis aspecifiek en langzaam, waardoor een optimale behandeling vaak laat kan worden gestart. Om deze situatie te verbeteren is de Moleculaire Diagnose en Risico Stratificatie van Sepsis (MARS) biobank studie in 2010 opgezet in twee Nederlandse intensive care units. Het doel van dit project is het ontwikkelen van instrumenten die snelle en accurate informatie verschaffen over een individuele patiënt ten aanzien van welk micro-organisme de infectie veroorzaakt en wat de ernst en de fase van de immuunreactie van de patiënt zijn. Dit proefschrift beschrijft enkele van de studies die werden uitgevoerd in het kader van dit consortium. In het eerste deel worden de uitdagingen in de diagnose bij patiënten die opgenomen worden op de intensive care met vermoedelijke sepsis en in het tweede deel de prognose van ernstig zieke patiënten na enkele belangrijke complicaties van sepsis beschreven.
Uitdagingen in de diagnose van sepsis &
In 1991 werden de huidige definities van sepsis, ernstige sepsis en septische shock N ederlandse samenvatting opgesteld door consensus van een groep experts. Deze definities stellen dat de diagnose van sepsis de klinische verdenking van infectie plus de aanwezigheid van een systemische inflammatoire respons syndroom, het zogenaamde SIRS, vereist. SIRS wordt gekenmerkt door specifieke fysiologische veranderingen, met afwijkingen in lichaamstemperatuur, witte bloedcellen, hartslag en ademhaling. Ernstige sepsis wordt gedefinieerd als de situatie van verminderde werking van organen door sepsis, en septische shock als de situatie van extreem lage bloeddruk in aanwezigheid van ernstige sepsis. Deze definities worden nog steeds gebruikt, maar de toepassing van de SIRS criteria blijkt in de dagelijkse praktijk niet eenvoudig. We laten in hoofdstuk 2 zien dat de incidentie van SIRS bij intensive care patiënten enorm varieerde van ongeveer 50 tot 100%, afhankelijk van hoe SIRS werd gemeten. We hebben hierbij de effecten van geautomatiseerde versus handmatige methoden van data collectie en van continue registratie van gegevens
165 versus 1-uursmetingen vergeleken. Daarnaast hebben we de invloed van verschillen in het vereiste aantal SIRS criteria, van het gelijktijdig aanwezig zijn van SIRS criteria, en van de minimaal vereiste duur van fysiologische verstoringen onderzocht. Aangezien deze kleine variaties in definities en metingen grote invloed hadden op het optreden van SIRS, leidde dit ook tot grote variaties in de incidenties van sepsis, ernstige sepsis en septische shock van 22 tot 31%, 6 tot 27%, en 4 tot 9% voor respectievelijk de meest restrictieve versus de meest liberale definities. Daarnaast had 39 tot 98% van de niet- infectieuze patiënten SIRS, maar had een aanzienlijk aantal patiënten zonder SIRS toch een infectie. Hieruit kan geconcludeerd worden dat de SIRS criteria zowel subjectief als niet-specifiek zijn. Deze tekortkomingen in de sepsis definities werden spoedig erkend en een decennium later bijgewerkt. Een uitgebreide lijst van symptomen werd ingevoerd om de klinische respons op infectie beter te beschrijven. Hoewel deze criteria tezamen sepsis beter beschrijven, zijn ze subjectief en afhankelijk van interpretatie. Ook de vaststelling van de aanwezigheid van infectie is niet eenvoudig. Zowel de determinatie van infectie als de primaire oorzaak van de ziekte, en de lokalisatie en de vermoedelijke bron van infectie zijn gecompliceerd. De Centers for Disease Control and Prevention (CDC) heeft daarom diagnostische criteria voor infecties ontwikkeld. Vele onderzoekers hebben echter gesuggereerd dat deze criteria te ingewikkeld, subjectief en onnauwkeurig zijn. Hoofdstuk 3 beschrijft de resultaten van onze systematische evaluatie van de bruikbaarheid van deze criteria voor de diagnose van infecties bij ernstig zieke patiënten. Hoewel de inter-beoordelaarsbetrouwbaarheid over het algemeen hoog was, vonden we aanzienlijk lagere niveaus van concordantie voor sommige subtypen van infectie, vooral voor beademingsgeassocieerde pneumonie (VAP). In de klinische praktijk worden de CDC infectiedefinities echter zelden gebruikt, maar zullen ernstig zieke patiënten vaak empirische antibiotische therapie krijgen voor een mogelijke infectie, zelfs wanneer de waarschijnlijkheid van infectie laag is. & Hoofdstuk 4 beschrijft de resultaten van een studie die de nauwkeurigheid van een “real-time” infectiediagnose onderzocht in patiënten die verdacht werden van een
N ederlandse samenvatting infectie bij opname op de intensive care. Meer dan 40% van alle ernstig zieke patiënten bleek achteraf een lage waarschijnlijkheid op infectie te hebben en zij hadden dus mogelijk ten onrechte therapeutische antibiotica toegediend gekregen. Opvallend was dat een initiële incorrecte diagnose geassocieerd was met een hogere mortaliteit. Onze studie was niet geschikt om dit interessante resultaat te verklaren. Het blijft daarom speculatief of de hogere sterfte in patiënten met een lage waarschijnlijkheid van infectie wordt veroorzaakt door een vertraging in gerichte behandeling wegens vertraagde herkenning van een alternatieve ziekte, of dat de alternatieve, niet- infectieuze, diagnose in deze patiënten schadelijker is dan de infectieuze diagnose. In hoofdstuk 5 wordt een nieuw systeem voor de surveillance van beademingsgerelateerde complicaties geëvalueerd. Deze aanpak werd onlangs in de Verenigde Staten als een vervanging van de conventionele surveillance van VAP geïntroduceerd. Zoals onder andere is gebleken uit ons onderzoek (hoofdstuk 3) blijkt de
166 diagnose van VAP gecompliceerd en subjectief te zijn, met als gevolg dat de diagnose vaak niet eenduidig gesteld kan worden. De nieuwe methode tracht een objectievere surveillance te bereiken door plotselinge verslechteringen in beademingsinstellingen te registreren. Zo’n episode wordt een beademingsgerelateerde complicatie (VAC) genoemd. Wij hebben dit algoritme elektronisch geïmplementeerd en vergeleken met de huidige handmatige VAP registratie binnen het MARS-onderzoek. Het bleek dat het algoritme hooguit 32% van de VAP patiënten detecteerde en een post-hoc analyse liet verder zien dat de VAC signalen meestal het gevolg waren van volume overbelasting en infecties, maar niet noodzakelijkerwijs van VAP. Bovendien was de attributieve sterfte sterk verschillend tussen de verschillende episodes, wat de opvatting ondersteunt dat het nieuwe algoritme verschillende soorten van VAP episodes identificeert.
Uitdagingen in de prognostiek van patiënten met sepsis De heterogeniteit van de patiëntpopulatie met sepsis maakt het zeer uitdagend om passende medische besluitvorming te initiëren. Op dit moment zijn er slechts symptoom gebaseerde classificatiesystemen en conventionele microbiologische technieken om patiënten te stratificeren. Beiden hebben nadelen: de classificatiesystemen zijn gevalideerd in grote populaties maar niet voor individuele patiënten, en diagnostische technieken hebben een lage sensitiviteit en nemen tenminste 24 uur in beslag. In de volgende alinea’s worden een aantal specifieke problemen wat betreft de prognostiek van patiënten met sepsis besproken. De verbeterde overleving van de initiële episode van sepsis in intensive care patiënten leidt tot secundaire problemen, onder andere een hogere incidentie van lange-termijn complicaties. Zo gaat sepsis vaak gepaard met een langdurige immuunsuppressie, wat mogelijk weer kan leiden tot secundaire intensive care- gerelateerde infecties. Er is bovendien gesuggereerd dat immuunsuppressie en secundaire infecties belangrijke oorzaken zijn van late sepsis sterfte in de intensive care. & Recent is daarom immuunstimulatie door middel van geneesmiddelen voorgesteld als
een behandelingsoptie voor sepsis; dit is een paradigmaverschuiving ten opzichte N ederlandse samenvatting van nog niet zo lang geleden toen het gebruik van anti-inflammatoire middelen werd voorgesteld. Hoofdstuk 6 beschrijft de incidentie en de attributieve mortaliteit van intensive care-opgelopen infecties bij patiënten met en zonder sepsis bij opname op de intensive care. Wij concluderen dat de incidentie van intensive care-opgelopen infecties onafhankelijk is van de aan- of afwezigheid van sepsis bij opname, wat in tegenspraak is met de huidige hypothese dat sepsis bij intensive care patiënten in een langdurigere en meer uitgesproken periode van immuunsuppressie resulteert dan niet- infectieuze ziekte bij opname. De incidentie van intensive care-opgelopen infecties was echter hoger bij patiënten met orgaanfalen of shock, hetgeen impliceert dat de ernst van ziekte en de daaruit voortvloeiende verblijfsduur op de intensive care belangrijkere factoren zijn. De attributieve mortaliteit van intensive care-opgelopen infecties was 13% op dag 60, en steeds hoger in patiënten zonder sepsis vergeleken met diegenen
167 met sepsis bij opname. De relatief lage attributieve mortaliteit suggereert dat andere factoren (bijvoorbeeld orgaanfalen en shock) meer tot sterfte bijdragen. Delirium is een complicatie van ernstige ziekte en wordt gekenmerkt door plotseling optredende verwardheid, hallucinaties, onrust of agressie. Delirium komt vaak voor bij ernstig zieke patiënten met sepsis en wordt geassocieerd met negatieve uitkomsten voor de intensive care patiënt. In hoofdstuk 7 is de relatie tussen delirium en overlijden op de intensive care onderzocht. In de literatuur worden 2 tot 3 maal hogere sterfterisico’s genoemd in patiënten met delirium in vergelijking met patiënten die geen delirium ontwikkelden. In deze onderzoeken werd in de statistische analyse echter geen rekening gehouden met de “concurrerende” risico’s voor delirium. Patiënten die snel worden ontslagen, hebben geen mogelijkheid meer om delirium te ontwikkelen op de intensive care; we zeggen dan dat (statistisch gezien) ontslag “concurreert” met het krijgen van delirium of met sterfte. Daarnaast werd niet eerder rekening gehouden met het beloop van ziekte tot het moment dat delirium optreedt. Het negeren van beide factoren kan tot vertekende resultaten leiden. Wij hebben met deze factoren wel rekening gehouden in onze studie. Onze studie laat zien dat delirium niet leidt tot extra sterfte op de intensive care, maar wel tot een vertraagd ontslag van de intensive care. Hierdoor liggen patiënten met delirium langer op de intensive care en lopen zij langer het risico op complicaties. Delirium dat twee of meer dagen duurde had echter wel een significant nadelig effect op mortaliteit. Hoofdstuk 8 beschrijft de incidentie en mortaliteit van een andere complicatie van sepsis, atriumfibrilleren (boezemfibrilleren). Zoals eerder besproken leidt sepsis tot de systemische productie van pro-inflammatoire cytokinen en stresshormonen, autonome dysfunctie, intravasculaire volumeverschuivingen en cardiovasculaire problemen. Dit zijn allen mogelijke risicofactoren voor de ontwikkeling van atriumfibrilleren. Wij onderzochten de effecten van de ernst van ziekte op de incidentie van atriumfibrilleren & en vonden dat hoe ernstiger de ziekte, hoe hoger de incidentie van atriumfibrilleren, namelijk ongeveer 10% in patiënten met sepsis, 20% in patiënten met ernstige sepsis
N ederlandse samenvatting en 40% in patiënten met septische shock. Atriumfibrilleren was geassocieerd met zowel een langere intensive care opname als een verhoogd sterftecijfer, en resulteerde in een ongeveer 2-voudige verhoogde mortaliteit, indien ontslag van de intensive care werd meegenomen als concurrerend risico. Bij intensive care patiënten die worden opgenomen met sepsis en orgaanfalen, is het voorkomen van progressie van deze, potentieel nog omkeerbare, afwijkingen het belangrijkste doel. Het blijkt echter moeilijk voor clinici te voorspellen - aan het bed - welke patiënten zullen reageren op interventies, en welke zullen verslechteren met als gevolg multi-orgaanfalen of dood. Hoofdstuk 9 beschrijft een model dat, voor een individuele patiënt met sepsis, de waarschijnlijkheid van sterfte, ontslag, en de ontwikkeling van multi-orgaanfalen tijdens de eerste week van de opname voorspelt, met behulp van routinematig beschikbare parameters. Dit model kan clinici helpen om intensive care patiënten met sepsis bij opname, maar ook gedurende het
168 verblijf op de intensive care, te stratificeren wat betreft hun kans op progressie van ziekte. Bovendien kan dit model worden gebruikt bij simulatie van de effecten van (hypothetische) behandelingen voor klinische sepsis studies of om de additionele prognostische waarde van nieuwe biomarkers te evalueren.
Conclusie In dit proefschrift worden enkele belangrijke uitdagingen in de diagnostiek en prognostiek van intensive care patiënten met sepsis besproken. Om de uitkomsten van patiënten met sepsis te verbeteren, moet worden gestreefd naar zowel nauwkeurigere en klinisch beter toepasbare definities van sepsis, als ook een betere identificatie van subgroepen binnen de heterogene groep van intensive care patiënten met sepsis. Met dit proefschrift zijn hierin belangrijke stappen gezet.
& N ederlandse samenvatting
169
About the author List of publications This thesis Klein Klouwenberg PM, Ong DS, Bonten MJ, Cremer OL. Classification of sepsis, severe sepsis and septic shock: the impact of minor variations in data capture and definition of SIRS criteria. Intensive Care Med. 2012 May;38(5):811-9.
Klein Klouwenberg PM, Ong DS, Bos LD, de Beer FM, van Hooijdonk RT, Huson MA, Straat M, van Vught LA, Wieske L, Horn J, Schultz MJ, van der Poll T, Bonten MJ, Cremer OL. Interobserver agreement of Centers for Disease Control and Prevention criteria for classifying infections in critically ill patients. Crit Care Med. 2013 Oct;41(10):2373-8.
Klein Klouwenberg PM, van Mourik MS, Ong DS, Horn J, Schultz MJ, Cremer OL, Bonten MJ, MARS Consortium. Electronic implementation of a novel surveillance paradigm for ventilator-associated events. Feasibility and validation. Am J Respir Crit Care Med. 2014 Apr 15;189(8):947-55.
Klein Klouwenberg PM, Zaal IJ, Spitoni C, Ong DS, van der Kooi AW, Bonten MJ, Slooter AJ, Cremer OL. The attributable mortality of delirium in critically ill patients: a prospective cohort study. BMJ. 2014;349:g6652.
Other publications Kuipers S, Klein Klouwenberg PM, Cremer, OL. Incidence, risk factors and outcomes of new-onset atrial fibrillation in patients with sepsis, severe sepsis and septic shock: a systematic review. Crit Care. Accepted.
Ong DS, Klein Klouwenberg PM, Verduyn Lunel FM, Spitoni C, Frencken JF, Dekker HA, Schultz MJ, Bonten MJ, Cremer OL. Cytomegalovirus seroprevalence as a risk factor for poor outcome in Acute Respiratory Distress Syndrome. Crit Care Medicine. & Accepted. L ist of publications Ong DS, Faber TE, Klein Klouwenberg PM, Cremer OL, Christiaan Boerma E, Sietses M, van Loon AM, Bonten MJ, Bont LJ. Respiratory syncytial virus in critically ill adult patients with community-acquired respiratory failure: a prospective observational study. Clin Microbiol Infect. 2014 Aug;20(8):O505-7.
Ong DS, Klein Klouwenberg PM, Spitoni C, Bonten MJ, Cremer OL. Nebulised amphotericin B to eradicate Candida colonisation from the respiratory tract in critically ill patients receiving selective digestive decontamination: a cohort study. Crit Care. 2013 Oct 11;17(5):R233.
Klein Klouwenberg PM, Sasi P, Bashraheil M, Awuondo K, Bonten M, Berkley J, Marsh K, Borrmann S. Temporal association of acute hepatitis A and Plasmodium falciparum malaria in children. PLoS One. 2011;6(7):e21013
173 Djimdé AA, Tekete M, Abdulla S, Lyimo J, Bassat Q, Mandomando I, Lefèvre G, Klein Klouwenberg PM, Borrmann S; B2303 Study Group. Pharmacokinetic and pharmacodynamic characteristics of a new pediatric formulation of artemether- lumefantrine in African children with uncomplicated Plasmodium falciparum malaria. Antimicrob Agents Chemother. 2011 Sep;55(9):3994-9.
Klein Klouwenberg PM, Tan L, Werkman W, van Bleek G, Coenjaerts FE. The Role of Toll-like Receptors in Regulating the Immune Response against Respiratory Syncytial Virus. Crit Review Imm. 2009;29(6):531-50.
Klein Klouwenberg PM, Valkenburg M, Cate O ten, Bootsma A, Dijk M van. The use of virtual microscopy versus traditional light microscopy in histopathology education. Dutch Journal of Medical Education 2009;28(6):261–268.
Klein Klouwenberg PM, Bont L. Neonatal and infantile immune responses to encapsulated bacteria and conjugate vaccines. Clin Dev Immunol. 2008;2008:628963.
Grobusch MP, Lell B, Schwarz NG, Gabor J, Dornemann J, Potschke M, Oyakhirome S, Kiessling GC, Necek M, Langin MU, Klein Klouwenberg PM, Klopfer A, Naumann B, Altun H, Agnandji ST, Goesch J, Decker M, Salazar CL, Supan C, Kombila DU, Borchert L, Koster KB, Pongratz P, Adegnika AA, Glasenapp IV, Issifou S, Kremsner PG. Intermittent Preventive Treatment against Malaria in Infants in Gabon--A Randomized, Double-Blind, Placebo-Controlled Trial. J Infect Dis. 2007 Dec 1;196(11):1595-1602
Verbaan FJ, Klein Klouwenberg PM, Steenis JH, Snel CJ, Boerman O, Hennink WE, Storm G. Application of poly(2-(dimethylamino)ethyl methacrylate)-based polyplexes for gene transfer into human ovarian carcinoma cells. Int J Pharm. 2005 Nov 4;304(1-2):185-92
Schwarz NG, Oyakhirome S, Pötschke M, Gläser B, Klein Klouwenberg PM, Altun H, Adegnika AA, Issifou S, Kun JF, Kremsner PG, Grobusch MP. 5-Day non-observed & artesunate monotherapy for treating uncomplicated falciparum malaria in young Gabonese children. Am J Trop Med Hyg. 2005 Oct;73(4):705-9 L ist of publications Klein Klouwenberg PM, Oyakhirome S, Schwarz NG, Glaser B, Issifou S, Kiessling G, Klopfer A, Kremsner PG, Langin M, Lassmann B, Necek M, Potschke M, Ritz A, Grobusch MP. Malaria and asymptomatic parasitaemia in Gabonese infants under the age of 3 months. Acta Trop. 2005 Aug;95(2):81-5.
Ramharter M, Oyakhirome S, Klein Klouwenberg PM, Adegnika AA, Agnandji ST, Missinou MA, Matsiegui PB, Mordmuller B, Borrmann S, Kun JF, Lell B, Krishna S, Graninger W, Issifou S, Kremsner PG. Artesunate-clindamycin versus quinine- clindamycin in the treatment of Plasmodium falciparum malaria: a randomized controlled trial. Clin Infect Dis. 2005 Jun 15;40(12):1777-84
174