An Analysis of Maternal Mortality and Morbidity in Perinatal Health Surveillance Systems in Europe

DOI: 10.1111/j.1471-0528.2012.03330.x Epidemiology www.bjog.org

What about the mothers? An analysis of maternal mortality and morbidity in perinatal health surveillance systems in Europe

M-H Bouvier-Colle,a AD Mohangoo,b M Gissler,c Z Novak-Antolic,d C Vutuc,e K Szamotulska,f J Zeitlina for The Euro-Peristat Scientific Committee* a Institut National de la Sante´ et de la Recherche me´dicale-Unite´ 953 Recherche e´pide´miologique en sante´ pe´rinatale et sante´ des femmes et des enfants—INSERM, UMR S953 Epidemiological Research Unit on Perinatal Health and Women’s and Children’s Health, UPMC University Paris, Paris, France b Netherlands Organization for Applied Scientific Research TNO, Leiden, the Netherlands c National Institute for Health and Welfare, Helsinki, Finland and Nordic School of Public Health, Gothenburg, Sweden d Department of Obstetrics and Gynaecology, Division of Perinatology, University Medical Centre, Ljubljana, Slovenia e Leiter der Abteilung fu¨r Epidemiologie Zentrum fu¨r Public Health der Med. Univ. Wien, Austria f Department of Epidemiology National Research Institute of Mother and Child, Warsaw, Poland Correspondence: Dr M-H Bouvier-Colle, Institut National de la Sante´ et de la Recherche me´dicale-Unite´ 953 recherche e´pide´miologique en sante´ pe´rinatale et sante´ des femmes—82, av Denfert-Rochereau, 75014 Paris, France. Email [email protected] * Members of the Euro-Peristat Scientific Committee are in the Appendix.

Accepted 27 January 2012.

Objective To assess capacity to develop routine monitoring of Results In 22 countries that provided data, the maternal maternal health in the European Union using indicators of mortality ratio was 6.3 per 100 000 live births overall and maternal mortality and severe morbidity. ranged from 0 to 29.6. Under-ascertainment was evident from comparisons with studies that use enhanced identification of Design Analysis of aggregate data from routine statistical systems deaths. Furthermore, routine cause of death registration systems compiled by the EURO-PERISTAT project and comparison with in countries with specific systems for audit reported higher data from national enquiries. maternal mortality ratio than those in countries without audits. Setting Twenty-five countries in the European Union and Norway. For severe acute maternal morbidity, 16 countries provided data about at least one category of morbidity, and only three Population Women giving birth in participating countries in 2003 provided data for all categories. Reported values ranged widely and 2004. (from 0.2 to 1.6 women with eclampsia per 1000 women Methods Application of a common collection of data by selecting giving birth and from 0.2 to 1.0 hysterectomies per 1000 specific International Classification of Disease codes from the women). ‘Pregnancy, childbirth and the puerperium’ chapter. External Conclusions Currently available data on maternal mortality and validity was assessed by reviewing the results of national morbidity are insufficient for monitoring trends over time in confidential enquiries and linkage studies. Europe and for comparison between countries. Confidential Main outcome measures Maternal mortality ratio, with enquiries into maternal deaths are recommended. distribution of specific obstetric causes, and severe acute maternal Keywords European Union, hospital data, maternal mortality morbidity, which included: eclampsia, surgery and blood transfusion ratio, severe obstetric complications. for obstetric haemorrhage, and intensive-care unit admission.

Please cite this paper as: Bouvier-Colle M, Mohangoo A, Gissler M, Novak-Antolic Z, Vutuc C, Szamotulska K, Zeitlin J for The Euro-Peristat Scientiﬁc Committee. What about the mothers? An analysis of maternal mortality and morbidity in perinatal health surveillance systems in Europe. BJOG 2012;119:880–890. Introduction ers. For instance, antenatal care has focused more on the During the 1970s and the 1980s, maternal and child health prevention of health problems for the fetus or infant rather policies focused more attention on infants than on moth- than on the organisation of obstetric and intensive care for the mothers in case of severe maternal complications. At Re-use of this article is permitted in accordance with the Terms and Conditions set out at http://wileyonlinelibrary.com/onlineopen#Online the end of the 1980s, maternal mortality was labelled ‘a 1,2 Open_Terms neglected tragedy’. In 1987, a plea for safe motherhood

880 ª 2012 The Authors BJOG An International Journal of Obstetrics and Gynaecology ª 2012 RCOG Maternal health in European surveillance systems worldwide was launched. This led to a greater international in maternal mortality and morbidity in Europe and to focus on maternal health because of high maternal mortal- compare the MMR with information from other sources— ity ratios in low-income countries. In Europe, as in other notably confidential enquiries. high-income countries, specific surveys were carried out on maternal mortality and initiatives for severe maternal mor- Methods bidity were taken.3–7 An important question is therefore the extent to which these initiatives have improved the We used data from the EURO-PERISTAT project, the capacity to analyse and to develop public-health strategies methods for which are described below and elsewhere,8,9 as for maternal health in Europe. well as data from published reports of national enquiries or Five million women give birth each year in the European specific studies into maternal deaths. Union. Another million women have failed pregnancies with first-trimester losses. Overall in the European Union, Definitions between 335 and 1000 women are estimated to die during or because of pregnancy, delivery or the puerperium.8 The Maternal mortality EURO-PERISTAT group, a European collaboration estab- Internationally accepted definitions for indicators of mater- lished to develop an information system on perinatal health nal mortality and obstetric causes of death have been pub- in Europe, recommends the maternal mortality ratio lished by the World Health Organization.10 Maternal death (MMR) and the causes of maternal death as two principal is defined as the death of a woman while pregnant or indicators of maternal health.5 Maternal mortality is con- within 42 days of the termination of pregnancy irrespective sidered to be an important indicator of health system per- of the duration and site of the pregnancy for any cause formance4 even in high-income countries where maternal related to or aggravated by the pregnancy or its manage- deaths are very rare but are considered sentinel events that ment, but not from accidental or incidental causes. Mater- raise questions about the effectiveness and quality of care. nal deaths are subdivided into direct and indirect obstetric Vital registration and healthcare information systems exist causes of death (Chapter O, digits from O 00 to O 99) of in all member states, and provide an opportunity to pro- the 10th revision of the International Classification of duce direct estimates of MMR using a common classifica- Diseases (ICD-10), which defines the classification of pathol- tion of causes of death. ogies related to pregnancy, childbirth and the puerperium. In addition to maternal mortality, EURO-PERISTAT rec- Currently all national cause-of-death registries in the study ommends an indicator of severe maternal morbidity. The countries record deaths coded according to ICD-10. difficulties involved in establishing common definitions of For the EURO-PERISTAT project we compiled data maternal morbidity have been apparent for some time. In about all maternal deaths and deaths attributed to the main the 1990s, the European Concerted Action on Maternal causes: ectopic pregnancy or abortion, complications of Mortality and Severe Morbidity (the MOMS study) with hypertension, antepartum and postpartum haemorrhage collaborators from 15 countries studied three types of severe (PPH), uterine rupture, amniotic fluid embolism or throm- maternal morbidity complications (severe postpartum boembolism, sepsis or chorioamnionitis, other direct obstet- haemorrhage, eclampsia/pre-eclampsia and sepsis) for which ric causes, indirect obstetric causes and unknown causes. the participants drew up common definitions.6 Special epi- These causes were then aggregated into nine major cate- demiological surveys were carried out in the participating gories, haemorrhages and uterine ruptures, amniotic fluid countries to estimate the prevalence of these maternal mor- embolism, thromboembolism, complications of hyperten- bidities. These showed wide differences between the morbid- sion, ectopic pregnancies and abortions, anaesthetic compli- ity rates. The study concluded that these were probably the cations, other direct obstetric causes, indirect causes, and result of differences in the data survey procedures (prospec- unknown causes.4 tive or retrospective for example). After a review of the liter- As a consequence of the very small number of deaths ature and based on studies performed in Europe, the each year in most countries, we requested data covering at EURO-PERISTAT group proposed in 2004 a series of severe least 2 years (2003 and 2004). Small countries provided maternal conditions linked to pregnancy and childbirth, data for longer periods to provide a more reliable estimate which might be generated using data from routine systems for their MMR. For example, Luxembourg provided data (hospital discharge registers and medical birth registers). for 5 years. Two large countries, Germany and Italy, pro- This article reports on the results of data collection on vided data for only 1 year. maternal health indicators (mortality and morbidity) in 25 The MMR is defined as the number of all maternal European Union member states and Norway to produce deaths from direct and indirect obstetric causes per the European Perinatal Health Report. Our aim was to 100 000 live births. We did not calculate MMRs by cause analyse current capacity to monitor trends and differences of death because of the small number of deaths.

ª 2012 The Authors BJOG An International Journal of Obstetrics and Gynaecology ª 2012 RCOG 881 Bouvier-Colle et al.

We calculated the percentage of deaths in each group of before 2004, except for maternal mortality for which data causes, as the number of the deaths reported (numerator) for two or more years were requested. TNO Quality of Life by the total number of maternal deaths (denominator). in the Netherlands was responsible for developing the data collection instrument and overseeing the collection process. Severe acute maternal morbidity The second source of data was publications of national The EURO-PERISTAT working group conducted an exten- results from countries that studied maternal deaths using sive review of potential maternal morbidity indicators. We enhanced systems of registration by confidential enquiries identified four possible sources for morbidity data used in into maternal deaths and/or data linkage. We drew on the scientific literature: (i) hospital discharge registers and published reports from Austria,11 Denmark,12 Finland,13 databases, (ii) financial data about hospital care, (iii) France,14 Italy (five regions),15 the Netherlands,16 Norway,17 obstetric quality registers, and (iv) medical birth registers Slovenia18 and the UK.19 These sources are listed in Appen- and databases. Because the most frequently used sources of dix S1 (see Supporting Information). We also included data are hospital discharge registers and databases, we comparison data from a recent international study using selected indicators that could be generated using data usu- routinely collected medical causes of death data, which were ally included in these databases. Our hypothesis was that corrected for under-reporting.20 Data from these studies all women with severe acute maternal morbidity (SAMM) were used for external validation of national MMR. In would receive hospital care and so be included in hospital addition, for those countries that carry out routine surveil- databases. lance of maternal deaths using enhanced systems (Finland, The EURO-PERISTAT maternal morbidity indicators France, the Netherlands, Slovenia and the UK), we included both management-based criteria and clinical diag- compared national MMR with MMR from other countries noses. The choice of indicators was agreed upon in a meet- with similar infant mortality rates, but no enhanced system ing of the EURO-PERISTAT committee, based on results for ascertaining maternal deaths. from EURO-PERISTAT I5 and our literature review. Four indicators, had been suggested in the first phase of the pro- Results ject (eclampsia, surgery, blood transfusion, and intensive- care unit admission). Embolisation was added as a fifth Maternal mortality ratios indicator. Our initial intent was to present each indicator All countries contributed data on maternal deaths except separately and also to combine them in a composite indica- Cyprus, Ireland and Slovakia. Belgium, Denmark, Greece, tor of SAMM. Appendix S2 (see Supporting Information) Norway, Portugal and Sweden provided only the total gives the exact definitions for each indicator. The indicators number of deaths without information about their causes. were defined as the number of women experiencing the The total number of maternal deaths reported by country condition or procedure as a rate (per 1000) of all women and by year varied from 0 in Malta, in both years and in giving birth to one or more live or stillborn babies. Norway and Slovenia in 2004, to 55 in both France and the UK in 2003. Data collection The total number of deaths per country ranged from 0 Members of the EURO-PERISTAT Scientific Committee to108; the total number of live births ranged from <8000 were responsible for organising data collection in their to over 1.5 million. The highest ratio was reported in Esto- own country. They either compiled the data themselves nia, with 29.6 per 100 000 live births and the lowest was 0 from data published by national organisations or provided in Malta, as shown in Table 1. Austria, Belgium, France the names of people to whom the data collection instru- and Hungary had rates around the EU average (of 6.2 per ments should be sent. The aim of EURO-PERISTAT was 100 000 live births). Sweden and Greece reported low ratios to gather population-based data at a national level. If of 2.0 per 100 000. these were not available, regional data were accepted if they covered a geographically defined population. Only Causes of maternal deaths data from existing routine data sources—including vital The profile of causes varied substantially from country to registration systems, hospital administrative data, systems country (Table 1) because of differences in the proportions or regular surveys—were used. The Scientific Committee of deaths attributed to unknown causes: seven countries did member for each participating country was responsible for not use this category at all, whereas in other countries many selecting the most appropriate data source. Appendix S1 deaths had no cause stated. This problem was greatest in (see Supporting Information) gives the data source used the Netherlands and in Germany where, respectively, 19 in each country. and 47% had no reported cause. Nevertheless, the general Aggregated data were collected using an excel-based sys- European profile of stated direct obstetric causes of death tem. We asked for data for 2004 or the latest available year shows that all obstetric haemorrhages, including a majority

882 ª 2012 The Authors BJOG An International Journal of Obstetrics and Gynaecology ª 2012 RCOG Maternal health in European surveillance systems

Table 1. Live births, maternal deaths, maternal mortality ratios, repartition (%) of the maternal deaths according to the obstetric causes, by country in 2003/04

Countries* Live births (n) Maternal MMR per 100 000 Causes of maternal deaths (%)** deaths (n) live births H AFE OTE CHT OD AI UK

Austria1 155 912 10 6.4 0 10 10 20 10 50 0 Belgium2 156 167 7 4.5 Cyprus No data Czech Republic 191 349 19 9.9 11 16 21 0 26 21 5 Denmark2 129 466 12 9.3 Estonia3 27 028 8 29.6 0 20 0 20 60 0 0 Finland 114 018 9 7.9 11 11 0 11 44 22 0 France 1 529 280 107 7.0 18 14 14 14 28 8 4 Germany4 1 320 820 67 5.1 7 5 7 2 16 16 473 Greece5,2 104 355 2 1.9 Hungary 190 274 14 7.4 14 0 14 0 36 29 7 Ireland No data Italy4,6 539 066 17 3.2 18 6 6 6 53 6 63 Latvia 41 340 5 12.1 0 20 20 0 0 60 0 Lithuania 61 017 6 9.8 0 0 17 17 67 0 0 Luxembourg7 27 252 2 7.3 0 0 0 0 100 0 0 Malta 7923 0 0.0 The Netherlands 362 012 32 8.8 9 0 13 13 13 34 19 Norway2 113 409 4 3.5 Poland 707 203 31 4.4 39 13 3 6 39 0 0 Portugal2 221 945 17 7.7 Slovakia No data Slovenia7 34 907 4 11.5 50 0 25 0 0 25 0 Spain 896 472 41 4.6 0 0 0 25 75 0 0 Sweden2 200 316 4 2.0 United Kingdom 1 411 545 108 7.7 6 14 8 9 41 22 0 All countries 8 308 853 519 6.2 13 11 10 9 32 17 8 Countries with <10% of 6 626 021 420 6.3 unknown causes

AFE, amniotic ﬂuid embolism; AI, all indirect; CHT, complications of hypertension; H, obstetric haemorrhage; OD, other direct (other direct obstetric causes: chorioamnionitis/sepsis, abortion/ectopic pregnancy; anaesthetic; uterine rupture and others); OTE, Other thromboembolisms; UK, unknown. *:1In Austria indirect causes of maternal death are registered since 2004. 2Belgium, Denmark, Greece, Norway, Portugal and Sweden provided no data on maternal mortality by cause of death. 3Estonia provided data for the years 2004 and 2005, and Slovenia provided data for the years 2001 and 2002. 4Germany and Italy provided data on maternal mortality by cause of death for 1 year only, respectively 2004 and 2002. 5Greece provided data for 1 year, 2003. 6Italy provided data on maternal mortality by cause of death based on ICD-9 codes. 7Luxembourg provided data on maternal death for the years 2000–2004 and Slovenia for the years 2001 and 2002. **We are very concerned by the fact that calculating percentages with so small numbers sometimes is not pertinent, but we need to have the same presentation for all the members of the EU.

of PPH by atony, accounted for the highest proportion many to 25% in Spain), and amniotic ﬂuid embolism (11% of maternal deaths in participating countries (13%). In in the European Union, ranging from 5% in Germany to countries that reported it as a direct cause of maternal 20% in Latvia and Estonia). death, the proportion of haemorrhages ranged from 7% in In nine countries that provided data to the EURO-PERI- Germany to 67% in Slovenia. Three other direct causes each STAT project, we were able to check the completeness for accounted for around 9–11% of maternal deaths: thrombo- routinely collected data about maternal deaths by compar- embolisms (10% in the European Union, ranging from 3% ing ratios with other published studies on maternal deaths: in Poland to 33% in Slovenia), complications of hyperten- in Austria, Denmark, Finland, France, Italy, the Nether- sion (9% in the European Union, ranging from 2% in Ger- lands, Norway, Slovenia and the UK. Table 2 shows that

ª 2012 The Authors BJOG An International Journal of Obstetrics and Gynaecology ª 2012 RCOG 883 Bouvier-Colle et al.

Table 2. Maternal mortality data according to different sources, numbers, ratios per 100 000 live births and percentage of underestimation, in France, Finland, Italy, Netherlands and UK around 2000–04

Countries (a) Conﬁdential (b) Civil (c) Under-estimation* MMRs according to different sources Years enquiries n registration (%) maternal deaths causes of death Conﬁdential Vital data Hogan** enquiries (last period) estimates 2000

Austria 1980–98 191 119 38.0 NA NA 5 (4–7) 2003/04*** 10 6.4 Denmark 1985–94 60 9.8 7 (5–9) 2003/04 12 9.3 Finland 1999 6 3 50.0 5.3 2.6 7 (5–9) 2003/04 9 7.9 France 1999 58 47 19.0 NA 7.4 11 (10–13) 2001–06 463 384 17.1 9.6 8.0 2003/04 107 7.0 Italy 2000–07 118 NA 74.6 11.8 NA 5 (4–6) 2003/04 17 3.2 The Netherlands 1993–2005 309 208 32.7 12.1 8.1 8 (10–11) 2003/04 32 8.8 Norway 1976–95 61**** 5.5** 7 (5–10) 2003/04 4 3.5 Slovenia 2003–05 8***** 1 87.5 9.4 1.9 21 (15–29) 2003/04 4 11.5 United Kingdom 2000–02 261 136 47.9 13.1 6.8 2003–05 295 149 49.5 13.9 7.0 8 (7–10) 2003/04 108 7.7

Sources, EURO-PERISTAT, Confidential enquiries and specific surveys. *Underestimation : c =(a) ) (b)/(a) · 100, except for Austria. **Hogan estimations are based on modelling of vital data. ***All 2003/04 data are taken from EURO-PERISTAT. ****Norway recorded direct obstetric causes of deaths only. *****Slovenia among the eight deaths, three were late maternal deaths (‡42 days). the official MMRs under-ascertained maternal mortality in audits or linkages or confidential enquiries (Finland, all the European countries where it was possible to make France, the Netherlands, Slovenia and the UK) with the comparisons. For Norway and Denmark, the confidential official MMR in countries that do not, but that have a sim- enquiries were performed at earlier dates and cannot be ilar level of infant mortality (Austria, Belgium, Czech compared with the EURO-PERISTAT data. Slovenia pub- Republic, Denmark, Germany, Italy, Luxembourg, Poland, lished a first report on its systematic audit from national Portugal, Spain, Sweden). In all five countries with specific registries resulting in an MMR of 9.8 per 100 000 for the audits, linkage or enquiries, MMRs were at least seven per period 1985–94. This is equal to the rate included in the 100 000 live births, whereas seven of the 11 other countries EURO-PERISTAT report, but covers a longer period. The (63%), had MMRs <7. under-ascertainment was generally between 20 and 50%. Often, underestimation was higher in countries with a Maternal morbidity lower official rate. We also compared the official MMRs in Sixteen member states provided data for at least one of the countries that routinely monitor maternal deaths using indicators of SAMM but only three provided data about all

884 ª 2012 The Authors BJOG An International Journal of Obstetrics and Gynaecology ª 2012 RCOG Maternal health in European surveillance systems the categories, including admission into intensive-care units (Table 3). Discussion Hysterectomy for PPH and eclampsia were the two com- This attempt to gather data on SAMM and maternal mor- plications for which data were most often available. There tality at a European level using routine national systems was a five-fold country difference in the estimates of show that currently available data are insufficient. Fortu- incidence, but the range was moderate at 0.2–1.0 per 1000 nately, some countries have data from enhanced systems women giving birth, as was the range for eclampsia: for identification of maternal death and these make it pos- 0.2–1.6 per 1000. In contrast, data on intensive-care unit sible to quantify the shortfalls in data from national statis- admission were not generally available and the between- tical offices. This is a crucial issue because the absence of country differences were large (six-fold); There were very good data on maternal mortality and morbidity under- wide variations in blood transfusion data, most probably mines national and European capacity to monitor maternal because of differences in inclusion criteria. Data on emboli- health in Europe, and to permit comparisons between sation of uterine arteries were available for only seven countries or surveillance of trends over time. Our results countries and the rates varied between 0.0 and 0.3 per suggest that calls for a greater focus on mothers are still 1000. highly relevant for European countries and may also be for

Table 3. Severe maternal morbidity rates per 1000 women by pathologies or interventions, according to the countries

Countries No. of Eclampsia ICU Blood transfusion Hysterectomy Embolisation women admission whatever the number of units*

Austria** Belgium Flanders 59 956 NA NA 11.5 NA NA Cyprus** Czech Republic 96 771 0.2 NA NA 0.8 NA Denmark 63 781 0.3 NA 11.0 0.3 0.0 Estonia 13 879 0.6 NA NA 0.9 NA Finland 56 878 0.2 NA 0.1 0.2 0.2 France 774 870 1.1 0.5 2.1 0.3 0.3 Germany–Bavaria 105 490 0.7 3.1 10.7 1.0 0.0 Greece Hungary 93 913 0.5 NA NA 1.0 0.0 Ireland** Italy 534 568 1.6 NA 4.6 0.9 0.0 Latvia 20 256 0.4 NA 0.8 NA Lithuania** Luxembourg Malta 3838 1.3 NA 5.2 0.5 NA The Netherlands 187 910 0.7 2.2 6.4 0.3 0.3 Norway** Poland 213 190 0.2 NA NA NA NA Portugal** Scotland only 53 342 0.6 NA NA 0.2 NA Slovakia** Slovenia 17 629 1.1 NA 10.6 0.6 NA Spain–Valencia 38 389 0.3 NA 6.5 0.3 NA Sweden** United Kingdom 82 911 0.67*** NA NA 0.13*** NA (Wales and Scotland)

*The number of transfusion units provided by the countries were so heterogeneous (three units or more, ﬁve units or more, other amount, no units speciﬁed) that we summarised the data in only one category. **No data provided; Norway did not participate in the data collection. ***The rates were estimated from two nations only, Wales and Scotland.

ª 2012 The Authors BJOG An International Journal of Obstetrics and Gynaecology ª 2012 RCOG 885 Bouvier-Colle et al. other high-income countries, as recently affirmed in the tems of confidential enquiries into maternal deaths exist in USA.7 France, the Netherlands, Slovenia and the UK and these For the measure of maternal mortality, there are two also rely in part on linkage systems for identifying problems, completeness of ascertainment and quality of cases.14,16,18,19 Implementing data linkage and especially coding. The comparison of data from routine cause of confidential enquiries in all European countries would sub- death certificates with enhanced systems for studying stantially improve the ascertainment of maternal death. maternal deaths identified substantial underestimation of There are also major differences between countries in between 20 and 50% in MMRs. Furthermore, countries reporting causes of death. This problem is more difficult to with continuous audits had higher reported MMR than resolve because of the small number of deaths and the countries that have not implemented these initiatives, prob- complexity of the multiple complications that lead to a ably because of increased awareness of maternal deaths, maternal death. Many women die in an intensive-care unit, which improves the routine reporting of these deaths. often days or weeks after delivery, and the certifying physi- From this perspective, the very low mortality ratios in cian may not always be aware of the details of the preg- some countries are highly suggestive of a failure to ade- nancy. A previous European study showed that the quately count maternal deaths. It would be interesting to differences in coding the underlying cause of death by the study why these systems are missing maternal deaths and national statistical offices can lead to underestimation or in particular, whether this is associated with the complete- overestimation of the MMR compared with standardised ness of ascertainment or the procedures for coding. The coding by a European panel of experts;24 these discrepan- absence of deaths from indirect obstetric causes in some cies were confirmed in a subsequent study comparing countries, for example Poland or Spain, might be the result Europe with the USA.13,25 of under-ascertainment, or coding procedures, or both. A Another important issue that limits the use of maternal further issue that affects completeness of ascertainment and mortality indicators for surveillance of maternal health in comparability is whether migrant or foreign citizens are pregnancy is the small number of deaths. For example, the included in official mortality statistics: for example, in relatively high MMR for Estonia is based on only eight France they are, but in Austria they are not. Migrant maternal deaths; On the other hand, one maternal death in women have been found to have higher MMR than non- Malta would have increased its MMR from 0.0 to 12.6 per migrant women in the countries to which they migrate.21,22 100 000. Given this degree of variation, we would recom- These routinely reported data are those regularly used at mend that future international data collection and report- the international level instead of the results from enhanced ing be based on averages over 5 years instead of 2 years to systems of maternal death recording. The recent paper by reduce the effects of variations in the MMRs caused by the Hogan et al.,20 an attempt at the international level to rec- small number of maternal deaths, in medium-sized as well tify the well-known under-ascertainment of MMRs by as small countries. This issue highlights the importance of using models based on vital statistics data in industrialised developing valid indicators of severe maternal morbidity countries, was based on national statistical office data. Con- which have a higher incidence and therefore have a greater sequently, their estimates were consistently below the true potential to measure trends in maternal health over time. values in the countries where alternative data sources were Our results show, however, that data on maternal mor- available. Nevertheless, in other countries, such as Greece, bidity are scarce and their quality is inadequate. We had Malta and Sweden, their corrections led to values that were expected that the incidence of embolism, eclampsia, blood two or three times higher than ratios reported in national transfusion and surgery for PPH would be readily derived statistics. from the data files compiled at hospital level. We know There are validated methods for improving statistics on that the majority of the European countries have hospital maternal mortality, including setting up systems of enqui- discharge registers or databases that are used to monitor ries. Adding a pregnancy check box to the national medical hospital activity and to allocate resources to hospitals. As death certificate is a first step, as shown recently in a study the morbidity outcomes and procedures we chose take from the USA,23 but this is not sufficient, as witnessed by place in hospital at or shortly after delivery, the cases continued under-reporting in countries, such as France should be included in these systems. Many countries were since 2000 and others where pregnancy check boxes have not able to report the number of women; these data may been implemented. To insure the completeness of registra- exist, but they were not currently available for this purpose. tion, more comprehensive solutions involve data linkage of The EURO-PERISTAT project compiled data that are cur- cause of death registers with medical birth registers or birth rently used to evaluate perinatal health outcomes. registration records (Denmark, England and Wales, Fin- A future line of research would be to request data on land, Norway and Slovenia), abortion records (Finland) hospital stays associated with childbirth from hospital and hospital records (Denmark and Finland). National sys- discharge databases and validate the accuracy of reporting

886 ª 2012 The Authors BJOG An International Journal of Obstetrics and Gynaecology ª 2012 RCOG Maternal health in European surveillance systems of some specific and crucial diagnoses for which there are LEMMoN28 study in the Netherlands, the recently started commonly accepted definitions, including eclampsia, Nordic project (NOSS) or the French Severe maternal thromboembolism and sepsis. Results from new national morbidity: measurement, determinants and quality of and international initiatives should be taken into consider- care project (EPIMOMS), are prospective but transversal ation when refining these definitions of severe maternal approaches limited to a specific time period. Neither the morbidity.26–28 UK study nor the others are designed to enable routine fol- Retrospective studies have been conducted in the USA, low up over time. Canada and Australia on hospital databases or discharge summaries, according to a definition of SAMM that com- Conclusion and recommendations bines codes of procedures and diagnoses and that therefore depends on the available information.29–31 This type of Despite the existence of longstanding cause of death regis- study has the advantage of feasibility, immediately available tration systems and hospital morbidity registers in Euro- data and a large sample size. Limitations include use of pean countries, currently available data for surveillance of codes from the ICD of the World Health Organization to maternal morbidity and mortality associated with preg- identify SAMM events; although the ICD has a section on nancy, childbirth and the postpartum period are inade- direct and indirect maternal complications, it does not quate. All countries should be encouraged to use validated define their severity. There is also significant variation in methods to improve the ascertainment of maternal deaths the reporting of diagnoses that are not the main diagnoses. and in particular confidential enquiries and data linkage. Moreover, hospital data are not able to distinguish morbid- Better use of data available in hospital discharge databases ity associated with pregnancy (temporal association) from should make it possible to identify indicators of morbidity maternal morbidity directly caused by pregnancy (temporal that can be validly compared. and causal association). In some systems also there is a risk of counting the same woman in the same pregnancy several Disclosure of interests times unless there is a unique identifier and admissions are The authors have no competing interests to declare. linked. The validity and accuracy of the reported conditions in the hospital data may be not checked32 in a detailed Contribution to authorship manner because of the large numbers of records involved. MHBC and JZ conceived and designed the study, wrote the Using hospital discharge data would also make it possi- first draft of the paper and revised the second version; ble to record admission to intensive-care units, to which AMM, MG, ZNA, CV and KS provided input on the analy- we are strongly favourable. Even though intensive-care sis and revised the initial draft of the manuscript; The admission depends in part on the way health care is organ- EURO-PERSTAT scientific committee participated in the ised, it can mark a critical event, a so called ‘near miss’.33 interpretation of the results and commented on the paper’s This identifies a situation in which resuscitation by an second and final drafts. intensive-care specialist was required, as confirmed by recent studies.34,35 Since 2006, the intensive-care national Details of ethics approval audit and research centre (ICNARC) has analysed admis- No specific ethics approval for this study was required sions of women of reproductive age among the admissions because outcome data were routinely collected at the aggre- to adult general intensive-care units, in England and Wales gate level by countries. and Northern Ireland.36 The LEMMoN Study in the Neth- erlands included a chapter about obstetric intensive-care Funding unit admissions.28 In France, these admissions are well The EURO-PERISTAT project received funding from the recorded in hospital discharge data.32 European Commission of, Directorate General for Health Epidemiological studies focusing on specific aspects of and Consumers Protection and Public Health Programme SAMM are more informative and provide an essential com- (no. 20101301). The funders had no role in the collection plement to routine reporting and some are already under of the data, the writing of the manuscript or the decision way. They are usually population-based surveys giving bet- to submit for publication. ter estimates of the severe maternal morbidity rates. The Scottish confidential audit of severe maternal morbidity is Acknowledgements the oldest survey since 200326 and allows calculation of We acknowledge the following contributors to the EURO- SAMM indicators annually. The United Kingdom Obstetri- PERISTAT perinatal health report: Austria Christian Vutuc, cal Surveillance System (UKOSS) covering the UK36 is an Abteilung fu¨r Epidemiologie Zentrum fu¨r Public Health der ongoing system that focuses in turn on specific types of Med. Univ. Wien; Jeannette Klimont, Statistics Austria; rare severe maternal morbidity. Other studies, such as the Belgium Sophie Alexander, Wei-Hong Zhang, Universite´ Libre

ª 2012 The Authors BJOG An International Journal of Obstetrics and Gynaecology ª 2012 RCOG 887 Bouvier-Colle et al. de Bruxelles, School of Public,Health, Reproductive Health Family Hospital; Portugal Henrique Barros, Sofia Correia Uni- Unit; Guy Martens SPE (Study Centre for Perinatal Epidemiol- versity of Porto Medical School, Department of Hygiene and ogy), Edwige Haelterman, Myriam De Spiegelaere Brussels Epidemiology; Slovakia Jan Cap, Jarmila Hajnaliova, National Health and Social Observatory; Cyprus Pavlos Pavlou, Maria Health Information Centre; Slovenia Zˇ iva Novak-Antolicˇ, Ivan Athanasiadou Ministry of Health, Health Monitoring Unit; Verdenik, University Medical Centre Division for Perinatology, Andreas Hadjidemetriou, Christina Karaoli, Neonatal Intensive Polonca Truden-Dobrin, Centre for Health and Health Care Care Unit, Makarios III Hospital; Czech Republic Petr Velebil, Research, Institute of Public Health of the Republic of Slove- Institute for the Care of Mother and Child, Vit Unzeitig, nia, Spain Francisco Bolumar, Universidad de Alcala Facultad Department of Obstetrics & Gynecology, Masaryk University de Medecina; Ramon Prats, Department de Salut Direccio Brno Denmark Jens Langhoff Roos, Obstetrics Clinic, Rigshos- General Salut Publica, Carmen Barona Perinatal Health Unit pitalet, Copenhagen University; Steen Rasmussen, Sundheds- Public Health Board, Isabel Rıó, CIBER Epidemiologıá y Salud styrelsen National Board of Health; Estonia Gleb Denissov, Pu´ blica (CIBERESP); Sweden Gunilla Lindmark, IMCH, Aka- Statistics Estonia, Luule Sakkeus, Kati Karelson, Mare Ruuge, demiska sjukhuset; Milla Bennis, National Board of Health National Institute for Health Development, Department of and Welfare; United Kingdom Gwyneth Lewis, Department of Health Statistics, Avi Tellmann, Estonian Medical Birth Regis- Health, Alison Macfarlane, Nick Drey, Department of Mid- try; Finland Mika Gissler, National Research and Development wifery, City University; Angela Bell,Health Promotion Agency Centre for Welfare and Health (STAKES); Anneli Pouta for Northern Ireland; Jim Chalmers, Etta Shanks, Information National Public Health Institute (KTL), Department of Child Services Division, NHS National Services Scotland; Di Good- and Adolescent Health; France Beátrice Blondel, Marie-He´le`ne win, Vital Statistics Output Branch, Office for National Statis- Bouvier-Colle, Ge´rard Breárt, Jennifer Zeitlin, Meagan Zim- tics; Clara Mmata, Information Centre for Health and Social beck INSERM U953; Christine Cans, SCPE Service d’Informa- Care, England; Kath Moser, Office for National Statistics; tion et d’Informatique Me´dicale (SIIM) Germany Nicholas Gwyneth Thomas Health Statistics and Analysis Unit, Statisti- Lack, Bavarian Working Group for Quality Assurance, Klaus cal Directorate, Welsh Assembly Government. Doebler, Federal Quality Assurance Office BQS; Greece Aris Antlaklis, Peter Drakis, Athens University Department of Supporting Information Obstetrics & Gynecology, Division of Maternal & Fetal Medi- cine; Hungary Istvań Berbik, Vaszary Kolos Teaching Hospital, Additional Supporting Information may be found in the Department of Obstetrics & Gynecology, Istvań Szabo´, Depart- online version of this article: ment of Obstetric and Gynaecology, Medical Faculty, Scientific Appendix S1. Data sources used for maternal mortality University of Pećs Ireland Sheelagh Bonham, Jacqueline and morbidity statistics in the EURO-PERISTAT project. O’Reilly, Economic and Social Research Institute (ESRI), Italy Appendix S2. Definitions used for maternal morbidity Marina Cuttini, Pediatric Hospital of Baby Jesus, Unit of indicators. Epidemiology; Sabrina Prati, Cinzia Castagnaro, Silvia Bruzz- Please note: Wiley-Blackwell are not responsible for the one, Marzia Loghi Istituto Nazionale di Statistica, ISTAT; content or functionality of any supporting information Latvia Jautrite Karaskevica, Irisa Zile, Health Statistics and supplied by the authors. Any queries (other than missing Medical Technologies State Agency; Ilze Kreicberga, Riga material) should be directed to the corresponding author. Maternity Hospital; Lithuania Aldona Gaizauskiene, Kotryna Paulauskiene, Lithuanian Health Information Centre; Luxem- Keypoints bourg Yolande Wagener, Ministe`re de la Sante´, Direction de la Sante´, Division de la Me´decine Pre´ventive et Sociale; Malta • The first comparison of official data compiled on mater- Miriam Gatt, Kathleen England, Department of Health Infor- nal health data in the European countries, with data mation and Research; Raymond Galea Department of Obstet- from enhanced systems of collecting maternal deaths. rics & Gynecology, University of Malta; the Netherlands • The majority of maternal mortality ratios extracted from Simone Buitendijk, Ashna D Mohangoo, Sabine Anthony, Ab vital data registration are almost certainly underesti- Rijpstra TNO Quality of Life, Prevention and Care, Section mated. Confidential enquiries into maternal deaths Reproduction and Perinatology, Leiden; Jan Nijhuis, Maas- should be implemented to improve surveillance. tricht University Medical Centre, Department of Obstetrics & • There are potential sources of information in hospital Gynecology; Chantal Hukkelhoven, The Netherlands Perinatal databases on maternal morbidity requiring further study. j Registry; Norway Lorentz Irgens, Kari Klungsoyr Melve, University of Bergen, Medical Birth Registry of Norway; Jon References Gunnar Tufta Medical Birth Registry of Norway; Poland Kat- arzyna Szamotulska, Department of Epidemiology, National 1 Rosenfield A, Maine D. Maternal mortality—a neglected tragedy. Research Institute of Mother and Child; Bogdan Chazan, Holy Where is the M in MCH?. Lancet 1985;2:83–5.

888 ª 2012 The Authors BJOG An International Journal of Obstetrics and Gynaecology ª 2012 RCOG Maternal health in European surveillance systems

2 Mahler H. The safe motherhood initiative: a call to action. Lancet 23 MacKay A, Berg C, Liu X, Duran C, Hoyert D. Changes in pregnancy 1987;8534:668–70. mortality ascertainment. Obstet Gynecol 2011;118:101–10. 3 Bouvier-Colle MH, Salanave B, Ancel PY, Varnoux N, Fernandez H, 24 Salanave B, Bouvier-Colle M, Varnoux N, Alexander S, Macfarlane A. Papiernik E, et al. Obstetric patients treated in Intensive care units and Classification differences in maternal deaths. The European study on maternal mortality. Eur J Obstet Gynecol Reprod Biol 1996;65:121–5. maternal mortality and morbidity surveys: MOMS. IJE 1999;28:64–9. 4 Wildman K, Bouvier-Colle M, Group atM. Maternal mortality as an 25 Deneux-Tharaux C, Berg C, Bouvier-Colle M, Gissler M, Harper M, indicator of obstetric care in Europe. BJOG 2004;111:164–9. Nannini A, et al. Underreporting of pregnancy-related mortality in 5 Alexander S, Wildman K, Zhang W, Langer M, Vutuc C, Lindmark the United States and Europe. Obstet Gynecol 2005;106:684–92. G. Maternal health outcomes in Europe. Eur J Obstet Gynaecol 26 Brace V, Penney G, Hall M. Quantifying severe maternal morbidity:a Reprod Biol 2003;111:S78–87. Scottish population study. BJOG 2004;111:481–4. 6 Zhang W, Alexander S, Bouvier-Colle M, MacFarlane A, Group aTM- 27 Say L, Souza JP, Pattinson R, WHO Working Group. Maternal near- B. Incidence of severe pre eclampsia, postpartum haemorrhage and miss—towards a standard tool for monitoring quality of maternal sepsis as a surrogate marker for severe maternal morbidity in a Euro- care. Best Pract & Res Clin Obstet Gynaecol 2009;23:287–96. pean population-based study: the MOMS-B survey. BJOG 28 Zwart J, Richters J, O¨ ry F, De Vries J, Bloemenkamp K, VanRoosma- 2005;112:89–96. len J. Severe maternal morbidity during pregnancy, delivery and 7 D’Alton ME. Where is the ‘M’ in maternal–fetal medicine? Obstet puerperium in the Netherlands: a nationwide population-based study Gynecol 2010;116:1401–4. of 371 000 pregnancies. BJOG 2008;115:842–50. 8 EUROPERISTAT. European Perinatal Health Report. Better statistics 29 Callaghan WM, MacKay AP, Berg CJ. Identification of severe mater- for better health for pregnant women and their babies. 2008: Euro- nal morbidity during delivery hospitalizations,United States, 1991– pean Union:280. 2003. Am J Obstet Gynecol 2008;199:133.e1–8. 9 Gissler M, Mohangoo A, Blondel B, Chalmers J, Macfarlane A, Ga- 30 Bacak S, Callaghan W, Dietz P, Crouse C. Pregnancy associated hos- izauskiene A, et al. Perinatal health monitoring in Europe: results pitalizations in the United States, 1999–2000. Am J Obstet Gynecol from the EURO-PERISTAT project. Inform Health Soc Care 2005;192:592–7. 2010;35:64–79. 31 Wen S, Huang L, Liston R. Severe maternal morbidity in Canada, 10 WHO. International Statistical Classification of Diseases and Related 1991–2001. Can Med J 2005;173:759–64. Health Problems. ICD-10. Geneva: WHO, 1992; vol 1,1241 pages. 32 Chantry A, Deneux-Tharaux C, Cans C, Ego A, Quantin C, Bouvier- 11 Karimian-Teherani D, Haidinger G, Waldhoer T, Beck A, Vutuc C. Colle M, et al. Hospital discharge data can be used for monitoring Under-reporting of direct and indirect obstetrical deaths in Austria, procedures and intensive care related to severe maternal morbidity. 1980–98. Acta Obstet Gynecol Scand 2002;81:323–7. J Clin Epidemiol 2011;64:1014–22. 12 Andersen B, Westergaard H, Bodker B, Weber T, Moller M, Sorensen 33 Pattinson R, Hall M. Near misses: a useful adjunct to maternal death J. Maternal mortality in Denmark, 1985–1994. Eur J Obstet Gynecol enquiries. Br Med Bull 2003;67:231–53. Reprod Biol 2009;142:124–8. 34 RCoA-RCOG-OAA. Female Admissions (aged 16–50 years) to Adult, 13 Gissler M, Deneux-Tharaux C, Alexander S, C B, Bouvier-Colle M, General Critical Care Units in England, Wales and Northern Ireland, Harper M, et al. Pregnancy-related deaths in four regions of Europe Reported as ‘‘Currently Pregnant’’ or ‘‘Recently Pregnant’’. London: and the United States in 1999–2000. Characteristics of unreported Intensive audit care national audit & research centre, 2007; 55. deaths. Eur J Obstet Gynecol Reprod Biol 2007;133:179–85. 35 Zwart J, Dupuis J, Richters J, O¨ ry F, van Roosmalen J. Obstetric inten- 14 CNEMM. Rapport du comite´ national d’experts sur la mortalite´ sive care unit admission: a two year nationwide population-based maternelle. Saint Maurice: InVS- INSERM, 2010. cohort study The Lemmon Study. Intensive Care Med 2010;36:256–63. 15 Donati S, Senatore S, Ronconi A, Group TRMMW. Maternal mortal- 36 Knight M, Kurinczuk J, Brocklehust P. UK obstetric surveillance sys- ity in Italy: a record linkage study. BJOG 2011;118:872–9. tem uncovered. RCM Midwives 2005;8:38–9. 16 Schutte JM, Steegers AP, Schuitemaker N, Santema JG, De Boer K, Pel M, et al. Rise in maternal mortlity in the Netherlands. BJOG 2010;117:399–406. Appendix 17 Andersgaard A, Langhoff-Roos J, OIan P. Direct maternal deaths in Norway 1976–1995. Acta Obstet Gynecol Scand 2008;87:856–61. The Euro-Peristat Scientific Committee 18 Kralj E, Mihevc-Ponikvar B, Balazic J. Maternal mortality in Slovenia: Christian Vutuc (Austria), Sophie Alexander (Belgium), Pavlos case report and the method of identifying pregnancy-associated Pavlou (Cyprus), Petr Velebil (Czech Republic), Jens Langhoff deaths. Forensic Sci Int Supplt Series 2009;1:52–7. Roos (Denmark), Luule Sakkeus (Estonia), Mika Gissler (Fin- 19 CEMACH. Saving mothers’ lives. Confidential Enquiries into Maternal land), Beátrice Blondel (France), Nicholas Lack (Germany), Deaths. London: CEMACH, 2007:267. 20 Hogan M, Foreman K, Naghavi M, Ahn SY, Wang M, Makela SM, Aris Antlaklis (Greece), Istvań Berbik (Hungary), Sheelagh et al. Maternal mortality for 181 countries, 1980–2008: a systematic Bonham (Ireland), Marina Cuttini (Italy), Jautrite Karaskevica analysis of progress towards Millennium Development Goal 5. Lan- (Latvia), Jone Jaselioniene (Lithuania), Yolande Wagener (Lux- cet 2010;375:1609–23. embourg), Miriam Gatt (Malta), Jan Nijhuis (The Nether- 21 Ibison JM, Swerdlow AJ, Head JA, Marmot M. Maternal mortality in lands), Lorentz Irgens (Norway), Katarzyna Szamotulska England and Wales 1970–1985: an analysis by country of birth. BJOG 1996;103:973–80. (Poland), Henrique Barros (Portugal), Ma´ria Chmelova´ 22 Philibert M, Deneux-Tharaux C, Bouvier-Colle MH. Can excess (Slovakia), Zˇ iva Novak-Antolic (Slovenia), Francisco Bolu´ mar maternal mortality among women of foreign nationality (Spain), Gunilla Lindmark (Sweden), Alison Macfarlane (Uni- be explained by suboptimal obstetric care? BJOG 2008;115: ted Kingdom). 1411–8.

ª 2012 The Authors BJOG An International Journal of Obstetrics and Gynaecology ª 2012 RCOG 889 Bouvier-Colle

Commentary on ‘What about the mothers? An analysis of maternal mortality and morbidity in perinatal health surveillance systems in Europe’

This important paper from EURO-PERISTAT concludes correctly that routine registration data on maternal mortality are insufficient to monitor trends over time and for comparing countries. Unfortunately the World Health Organiza- tion relies solely on such data when producing maternal mortality estimates worldwide (Hogan et al., Lancet 2010;375:1609–23). More sophisticated systems such as confidential enquiries reveal substantial under-reporting, with rates of almost 20% in France, 33% in the Netherlands, 38% in Austria, almost 50% in the UK, 50% in Poland, 75% in Italy and 87% in Slovenia. This makes routinely collected vital statistics less useful and comparison between countries meaningless. Sweden will soon publish data from active surveillance, which will illuminate the extremely low official maternal mortality ratio in that country. Under-reporting has several causes, including incomplete ascertainment or misclassifications of maternal deaths as non-maternal. When immigrant deaths are not included (as in Austria) or indirect maternal deaths are left out (as in Norway), comparison between different countries becomes meaningless, especially as we know that immigrants are often disproportionally represented and indirect deaths comprise a high proportion. In this study, five countries reported zero indirect deaths. That may be understandable when the total numbers are low (Luxembourg, four; Lithuania, six and Estonia, eight). It is, however, likely that Spain with 41 and Poland with 31 deaths, did not (like Norway) report their indirect deaths. Substantial differences within countries exist, with large variations between regions, cities, provinces and even neighbourhoods (Saucedo et al., BJOG 2012;119:573–81; de Graaf et al., BJOG 2012;119:582–8). Higher maternal mortality ratios in deprived areas (as detected by postal codes) point to the socio-economic and multi-ethnic determinants of health, indicating that serious health inequalities still exist within our welfare states. The relatively short period of data collection leading to few maternal deaths being recorded, e.g. Luxembourg (n = 2 for 2000–04), Slovenia, Sweden and Norway (n = 4), Greece (n = 2) is a limitation. Causes of maternal deaths vary substantially between countries, although obstetric haemorrhage remains the most frequent cause. Deaths from haemorrhage range between 0 and 50% of all maternal deaths, and those from amniotic fluid embolism are between 0 and 20%. Amniotic fluid embolism often involves serious haemorrhage, so classifications are bound to overlap. As maternal deaths tend to become litigation cases, clinicians sometimes may prefer the ‘less avoidable’ cause of amniotic fluid embolism to obstetric haemorrhage. Unknown causes differed between 0 and 47% of the deaths. In ten of the 16 countries no maternal deaths of ‘unknown origin’ were reported. This gives rise to doubts about the figures, because documentation is often a problem, even in a confidential enquiry. Moreover, maternal deaths are often complex. Recently, when 25 experts from the International Network of Obstetric Survey Systems (INOSS) at their second annual INOSS meeting in Leiden tried to classify cases of maternal mortality from France and the Netherlands, consensus could not easily be reached, and an underlying cause could not always be assigned. Such differences can only be resolved by in-depth audit of every case of maternal death, with prospective data collection. This can be achieved when countries have national enquiries in place, and that is the most important recommendation from this study. j J van Roosmalen Department of Obstetrics, Leiden University Medical Centre, the Netherlands

890 ª 2012 The Author BJOG An International Journal of Obstetrics and Gynaecology ª 2012 RCOG The European Journal of Public Health Advance Access published January 7, 2013 European Journal of Public Health, 1–7 ß The Author 2013. Published by Oxford University Press on behalf of the European Public Health Association. All rights reserved. doi:10.1093/eurpub/cks176 ...... Reporting of perinatal health indicators for international comparisons—enhancing the appearance of geographical plots

Nicholas Lack1, Beatrice Blondel2, Ashna D. Mohangoo3, Luule Sakkeus4, Christine Cans5, Marie H. Bouvier-Colle2, Alison Macfarlane6, Jennifer Zeitlin2

1 BAQ, Bavarian Institute for Quality Assurance, Department of Methods and Perinatology, Munich, Germany 2 Epidemiological Research Unit on Perinatal and Women’s and Children’s Health, INSERM UMRS 953, Universite´ Pierre- et-Marie Curie Paris6, Paris, France 3 TNO Behavioral and Societal Sciences, Department Child Health, Leiden, The Netherlands 4 Estonian Institute for Population Studies, Tallinn University, Tallinn, Estonia 5 Laboratoire TIMC-IMAG, ThEMAS, Universite Grenoble 1, CHU, Grenoble, France 6 Maternal and Child Health Research Centre, City University London, UK

Correspondence: Nicholas Lack, BAQ, Westenriederstrasse 19, 80331 Munich, Germany, tel: 0049 89 211 590 0, Downloaded from fax: 0049 89 211 590 20, e-mail: [email protected]

Background: Tabulating annual national health indicators sorted by outcome may be misleading for two reasons. The implied rank order is largely a result of heterogeneous population sizes. Distinctions between geographically adjacent regions are not visible. Methods: Regional data are plotted in a geographical map shaded in terms of

percentiles of the indicator value. Degree of departure is determined relative to control limits of a corresponding http://eurpub.oxfordjournals.org/ funnel plot. Five methods for displaying outcome and degree of departure from a reference level are proposed for four indicators selected from the 2004 European Perinatal Health Report. Results: Spread of indicator values was generally largest for small population sizes, with results for large populations lying mostly close to respective European medians. The high neonatal mortality rate for Poland (4.9 per 1000); high low-birthweight rates for England and Wales (7.8%), Germany (7.3%) and Estonia (4.5%); and high caesarean section rates for Italy (37.8%), Poland (26.3%), Portugal (33.1%) and Germany (27.3%) were statistically significant exceptions to this pattern. Estonia also showed an extreme result for maternal mortality (29.6 per 100 000). Conclusion: Extreme deviations from EU reference levels are either correlated with small population sizes or may be interpreted in terms of differing medical practices, as in the case of caesarean section rate. EURO-PERISTAT has now decided to use 5-year averages for maternal mortality to reduce the variance in outcome. Use of two colours in three intensities

and solid fill versus crosshatching is best suited to display rate and significance of difference. at INSERM / ICGM on June 26, 2014 ......

Introduction lower end of the gestational age distribution.7 After due consider- ation and suitable adjustment for methodological difficulties8 he chief objective of any collection of indicators of health for a and availability and comparability issues,9 however, there remain Tnumber of countries will be to highlight and interpret differences questions of presentation of these data, e.g. how to order a table found between countries, with the ultimate aim of suggesting or what to display on a geographical map. possible causes associated with risk factors or health care. Current A tabular presentation is attractive because it can provide a large applications of international comparisons of health and health care amount of numerical information such as rates for each country or are published on the websites of EUROSTAT1 and WHO2 and in population size and because of its naturally inherent ranking option. the European Community Atlas of Avoidable Death.3 In addition, An alphabetical ordering such as adopted in the European Perinatal a wide range of commercially available databases and programs can Health Report6 may be considered objective, as reporting sequence is be found on the Internet. solely a function of lexical ordering. However, this objectivity is at In the perinatal field, each country collects data used to derive the expense of any form of structured data display that might help indicators through civil registration, hospital discharge systems or interpretation. On the other hand, the ranks of countries in terms specialist registers. The indicators include mothers’ and children’s of observed rates will be instable10,11 because absolute differences health, clinical practices and maternal risk factors. Thus, in the between observed rates are frequently small. An instable ranking European Union, all countries produce sufficient data to construct may lead to misinterpretation of observed raw outcome rates. at least 10 core indicators, to describe their situation.4 Comparison Longitudinal variation is best displayed graphically. Although of these indicators across countries is an essential tool for defining graphs of temporal trends can reveal valuable additional informa- national policy priorities, or generating hypotheses on factors that tion, we cannot always expect data to be available for long time might explain differences such as risk factors or health policies. To series. This is so especially for the more sophisticated perinatal achieve this aim, it is essential to display data in a way that is sim- indicators. These may not be available for all countries on a yearly ultaneously accurate as well as informative without being too basis. cumbersome and voluminous to digest. Funnel plots of rates against population size, as pioneered by The recent claim that the validity of international comparisons is Macfarlane12 and later adopted for the National Health Service compromised by variations in the registration of births at the more generally by Spiegelhalter,11 are increasingly used in per- borderline of viability5 may hold for some publications, although formance monitoring. A major advantage is the inclusion of a not so for the European Perinatal Health Report,6 as care was taken formal assessment of degree of departure from specified reference to ensure comparability by appropriate censoring of indicators at the levels with simultaneous display of outcome–volume relationships. 2of7 European Journal of Public Health

Funnel plots were recently proposed for monitoring and assessing perinatal health. An enhanced choropleth map was constructed improvements13,14 in surgical interventions in Germany. for caesarean section rates. This indicator was chosen because of The plotting of performance indicators for countries in a geo- its status as a core indicator, its availability in all but one country graphical context is possible because of the availability of graphical and its good quality of documentation. information software. This encourages and facilitates regional interpretation. Correlations between rates in adjacent states are immedi- Tabular summary ately apparent3,15. Holland3 suggests a number of steps that might be taken in studying geographical variation. One is to see how these A table of denominators ni, observed rates yi and p-values indicating differences change over time; another is to map individual deaths significance of departure from respective EU medians y0 was con- and to investigate clustering. structed in alphabetical order of country. The median was selected to prevent large countries from dominating the comparisons. In this article, we take these ideas one step further and suggest 1 moving the information contained in funnel plots into choropleth The p-values are computed as pi = È (zi) from the inverse maps to improve their legibility and enhance their interpretability. normal distribution function where The proposed method is applied to perinatal health indicators from yi y0 7 zi ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffi the European Perinatal Health Report. y0ðÞ1y0 n Methods EURO-PERISTAT data Standard choropleth maps Downloaded from The EURO-PERISTAT6 data are available for 25 European Union We plot the observed rates using colour coding derived from six (EU) member states plus Norway. Data to construct a number of percentiles of the distributional characteristics of the outcome indicators were compiled for the year 2004, in most cases at national measures and with upper limits at 10, 25, 50, 75, 90 and 100. We population–based level. In the absence of population-based data, chose shades of the neutral colour blue, with dark shades corresponding to high rates, in keeping with the examples in the European data were provided on a survey basis for the province of Valencia 6

Perinatal Health Report. http://eurpub.oxfordjournals.org/ in Spain and for France. Population-based data for the Brussels and 16 Flanders regions of Belgium, for the countries of Northern Ireland, The maps were constructed with the SAS GMAP procedure. The Scotland and England and Wales jointly were available separately, perimeters of country boundaries were extracted from the SAS/ MAPS16 database and linked to observed rates through the thus yielding 25 European countries/regions for analysis. The data 16 were mainly obtained from civil registration based on birth and country identifiers ci. Version 9.1 of SAS provides data at death certificates or from other national sources, including specific national level. Subsets of the database corresponding to the perinatal databases. country of Scotland and the Flanders and Brussels region of Belgium were generated by straight-line dissection of existing national coordinate files. Countries for which no data were Selection of indicators available are left blank.

To explore the effects of various forms of presentation, four at INSERM / ICGM on June 26, 2014 indicators were chosen from the EURO-PERISTAT set of ‘core’ Funnel plots indicators for 2004. They were chosen because they differed with 17 respect to their average rate and with the further requirement that Funnel plots were constructed using R by plotting the observed comparable observations were available for the majority of outcome against population size as a proxy measure of precision. countries. In ascending order of mean incidence, these are Control limits were drawn vertically around the median rate for all maternal mortality (ratio of number of pregnancy and childbirth- countries. To assess degree of departure from the median rate y0, related maternal deaths to live births) with a ratio of 4–6 per statistical control limits were computed from the standard normal 100 000 live births, crude neonatal mortality (deaths at 0–27 days approximation to the binomial distribution as sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi after live birth) with a rate of 3–5 per 1000 live births, birthweight ÀÁ ÀÁy0 1 y0 below 2500 g among all live and stillbirths with a rate of 6% and y z 0 = caesarean sections (all elective and emergency caesarean sections 1 2 ni combined) with a rate of 15% per total number of births. Owing to its low overall incidence, maternal mortality was for levels of 5.0 and 0.1% and 0 < y0 <1. compiled for the years 2003–2004 combined, except for Germany (2004), Italy (2002) and Luxembourg (2000–2004), to increase the Enhanced choropleth maps stability of these estimates. For countries with extremely small Two basic options for indicating significance of difference from a population sizes such as Estonia and Malta, only 2-year averages reference level were considered, either by superimposing suitable were available. No data were available on maternal deaths from symbols or by application of different shading patterns. The five Cyprus, Ireland and Slovakia. No data were available for neonatal variants of these two approaches are listed in table 1 and figure 1. mortality for Ireland, which only provided data on early neonatal Whereas colouring schemes A, B and C use a neutral colouring of mortality (deaths at 0–6 days after birth), whereas Cyprus provided percentile bands, schemes D and distinguish deviations above and no data on the distribution of birthweight, and no data were below the reference by different colours. Schemes A and B use available on mode of delivery for Greece. superimposed symbols, schemes C and E differentiate by pattern Throughout, the data sets are defined by observations (c ,r and i i and scheme D uses bold perimeters to indicate significant differ- n ) denoting country, number of cases in the numerator and i ences. A significance level of 0.001 was used throughout. denominator, respectively, with the subscript i running through all 25 countries/regions. The raw rates are computed as yi =ri/ni with 0 < yi <1. Results Table 2 shows denominators, rates and significance of departure Presentation from respective EU medians for the four selected indicators at In addition to a tabular summary, standard choropleth maps and regional or country level. The denominators vary considerably funnel plots were constructed for these four selected indicators of according to the demographic characteristics. The population of Reporting of perinatal health indicators 3of7

Table 1 Choices of colouring schemes for enhanced choropleth maps

Attribute A B C D E

Percentile bands Shades of blue Shades of blue Shades of blue Shades of red and green Shades of red and green Significance Plus and minus symbols Up and down arrows Solid fill Bold contours Solid fill

Intensity of shading corresponds to respective percentile band. Darker shades of blue and red colours indicate values above the median, lighter shades of blue and green colours indicate values below the median. Extreme shadings correspond to values below the 10th or above the 90th percentile. Downloaded from http://eurpub.oxfordjournals.org/

Figure 1 Colouring schemes for enhanced choropleth maps: Schemes A, B and C use three darker intensities of blue shading above and three lighter intensities (two of which are shown) of blue below the median. Significance is indicated by appropriate symbols superimposed in schemes A and B, by solid colouring in schemes C and E and by bold perimeters in scheme D at INSERM / ICGM on June 26, 2014

Table 2 Denominators (n), regional rates and P-values for EU member states for maternal deaths (MD) per 100 000 live births, birthweight under 2500 g (LBW) as percentage of total births, neonatal deaths (NND) per 1000 live births and caesarean sections (CS) as percentage of total births

Country Code n MD Pn LBW Pn NND Pn CS P

Austria AT 155 912 6.4 0.34 79 229 7.0 0.00 78 934 2.7 0.48 79 229 23.5 0.00 Belgium-Brussels BE.b 32 400 6.2 0.41 15 842 6.7 0.48 16 200 3.1 0.14 15 561 17.2 0.00 Belgium-Flanders BE.f 119 167 4.2 0.11 60 921 6.8 0.21 60 672 2.4 0.07 60 921 18.9 0.00 Czech Republic CZ 191 349 9.9 0.09 98 056 6.9 0.00 97 664 2.3 0.01 96 098 16.3 0.00 Denmark DK 129 466 9.3 0.20 64 585 5.4 0.00 64 521 3.6 0.00 63 767 20.5 0.50 England and Wales UK.ew 1 261 190 7.2 0.47 642 050 7.8 0.00 639 721 3.4 0.00 612 888 23.1 0.00 Estonia EE 27 028 29.6 0.00 14 017 4.5 0.00 13 990 4.2 0.00 14 020 17.7 0.00 Finland FI 114 018 7.9 0.40 57 734 4.3 0.00 57 569 2.4 0.11 57 752 17.1 0.00 France FR 1 529 280 7.0 0.34 14 534 7.2 0.01 767 816 2.6 0.01 14 696 20.2 0.20 Germany DE 692 802 5.3 0.03 648 667 7.3 0.00 705 622 2.7 0.30 647 685 27.3 0.00 Hungary HU 190 274 7.4 0.48 95 537 8.6 0.00 95 137 4.4 0.00 95 613 25.6 0.00 Italy IT 539 066 3.2 0.00 541 270 6.8 0.00 539 066 2.8 0.05 539 235 37.8 0.00 Latvia LV 41 340 12.1 0.13 20 492 5.4 0.00 20 355 5.7 0.00 20 256 19.6 0.00 Lithuania LT 61 017 9.8 0.23 29 633 5.1 0.00 29 480 4.6 0.00 29 595 17.4 0.00 Luxembourg LU 27 252 7.3 0.50 5300 4.8 0.00 5469 2.0 0.16 5422 25.3 0.00 Malta MT 7923 0.0 0.22 3899 8.0 0.00 3887 4.4 0.02 3902 28.3 0.00 Netherlands NL 362 012 8.8 0.14 182 263 6.7 0.44 181 006 3.5 0.00 182 135 15.1 0.00 Northern Ireland UK.n 43 786 9.1 0.32 22 503 6.0 0.00 22 362 3.0 0.25 22 378 27.6 0.00 Norway NO 113 409 3.5 0.07 57 356 5.0 0.00 57 111 2.1 0.00 57 368 15.6 0.00 Poland PL 707 203 4.4 0.00 358 381 6.4 0.00 356 697 4.9 0.00 350 048 26.3 0.00 Portugal PT 221 945 7.7 0.42 109 444 7.8 0.00 109 356 2.6 0.17 107 195 33.1 0.00 Scotland UK.s 106 389 12.2 0.03 53 255 7.5 0.00 52 911 3.0 0.07 52 893 24.7 0.00 Slovenia SI 34 907 11.5 0.18 17 946 6.2 0.00 17 846 2.6 0.42 17 937 14.4 0.00 Spain ES 896 472 4.6 0.00 435 748 7.6 0.00 454 591 2.6 0.16 38 290 24.6 0.00 Sweden SE 200 316 2.0 0.00 100 219 4.3 0.00 100 158 2.1 0.00 100 081 17.4 0.00

The p-values indicate significance of difference from respective EU medians. 4of7 European Journal of Public Health Downloaded from http://eurpub.oxfordjournals.org/ at INSERM / ICGM on June 26, 2014

Figure 2 Choropleth panel plot of selected EURO-PERISTAT indicators. Colour grading is by quintiles of outcome rate. Maternal mortality is based on the period 2003–2004 for the majority of countries (see text)

England and Wales is 160 times that of Malta and 50 times that deviate considerably from the median, again seemingly irrespective of Estonia and Luxembourg. The median rates are 7.3 per 100 000 of population size. Italy stands out due to its high caesarean section for maternal deaths, 6.7% for the proportion of births below 2500 g, rate and its large population size. 2.7 per 1000 live born for neonatal deaths and 20.5% for caesarean Figure 4 shows an enhanced choropleth plot for caesarean sections sections. using scheme E from figure 1. The width of the upper percentile Figure 2 shows a corresponding panel of standard choropleth band is 5 times that of the lower band as a result of the skewed maps, with intensity of blue corresponding to percentile band for distribution, with Italy and Portugal showing comparatively high all four indicators. Countries for which no data were available are caesarean section rates. All countries except Denmark, Latvia and left blank. The percentile grading is reflected in the broader ranges at France show significant differences from the EU median rate of the extremes of the distributions. Some of the subranges are not 20.5% (table 2). Twenty-two countries deviate significantly from contiguous owing to their automatic generation by SAS/GRAPH the EU median, 11 above and 11 below. software16 based on the actual distribution of observed rates. Figure 3 shows the corresponding panel of funnel plots. In the case of maternal mortality, most countries fall within the control limits. Estonia stands out with eight maternal deaths in 27 028 live Discussion births between 2003 and 2004 (29.6 per 100 000 live births vs. 0.0– Standard choropleth maps 12.2 for the remaining countries/regions). For neonatal mortality rates, the majority of smaller countries (with <150 000 births) lie The chief advantage of choropleth maps lies in making regional within the limits. Although there are several other countries with patterns transparent. At a glance, one can discern the strong high neonatal mortality rates, all new member states of the European contrast in neonatal mortality rates between the Eastern European Union as well as Poland stand out due to its large population. The states and Nordic states or the high caesarean section rates in Italy, distribution of low birthweight shows countries largely falling either Germany and Portugal (figure 2). Displays such as these were used below, within or above the control limits, irrespective of population for reporting perinatal indicators,7 and joint inspection may help to size. In the case of caesarean sections, only one, Denmark (DK) generate hypotheses about possible common causes for rates in alone, lies close to the European median rate. All other countries selected regions. Reporting of perinatal health indicators 5of7

Maternal Mortality Low birth weight 2004

EE HU MT PT UK.ew UK.s ES FR DE ATCZ BE.bBE.f NL IT PL SI LVSIUK.s UK.n CZ UK.nLTDK NL LUFIHUPT UK.ew FR LV DK BE.bAT LT DEPL ES 5678 NO BE.fNO IT LU SE EE

0102030 MT FISE Percentage of total births Percentage Deaths per 100,000 live birthsDeaths per 100,000 live 0 5 10 15 0246

Numbers of live births, 100,000s Numbers of total births, 100,000s Downloaded from Neonatal Mortality 2004 Caesarean Sections 2004

LV IT http://eurpub.oxfordjournals.org/

PL PT LT MT HU EE MT UK.n DE PL LU HU DK ESUK.s NL UK.ew AT UK.ew BE.b UK.s DK UK.n IT FRLV SI AT ES DE BE.f FI PT FR BE.bEELTFISE BE.fCZ CZ

NO at INSERM / ICGM on June 26, 2014 NOSE NL 15 20 25 30 35 2345 LU SI Percentage of total births Percentage

0246 0246 Neonatal deaths per 1,000 live birthsNeonatal deaths per 1,000 live Numbers of live births, 100,000s Numbers of total births, 100,000s

Figure 3 Funnel plots of indicators shown in figure 2 against total number of deliveries for European countries and regions with control limits drawn at 95.0 and 99.9%. Some of the axes exclude zero to enhance readability. For maternal mortality, the x-axis is on a wider scale due to the inclusion of the period 2003–2004 for the majority of countries (see text)

Panel plots are also well suited for comparisons across different also between selected indicators of perinatal health (figure 3). This performance indicators, which are highly correlated. For instance, strongly suggests non-random causes for the differences in perinatal one may see that the Nordic states have lower rates for caesarean health, which is particularly marked for low birthweight and sections, for low birthweight and, with the exception of Finland, for caesarean sections. Presenting indicators of perinatal health in maternal mortality. A further advantage lies in the information a panel facilitates the generation of hypotheses about differences contained in geographical size of the region. This, albeit rough, in the provision and effectiveness of health care. This form of pres- proxy measure of population size can be of help in interpreting entation also makes the fact that population size differs between the results. units more visible. EURO-PERISTAT has now decided to collect Six categories were chosen for grading the outcome rates in terms of maternal mortality for a 5-year period because of the difficulties distribution percentiles for all performance indicators to differentiate associated with low sample sizes. This is expected to result in a countries with respect to the median. Although this indicates distinct lower maternal death ratio for Estonia and conversely a higher categories of absolute rates, it allows no conclusion about the signifi- ratio for Malta, where data were available only for 2 successive cance of departure of this rate from a reference value. This can be years. When data are presented as rates, it is easy to forget how misleading, especially for small countries. It must be borne in mind few maternal deaths there are. that although identically coloured regions imply identical percentiles across different choropleth maps, this will not in itself indicate how far an individual rate differs from the mean or median. Choice of pattern Although initially favoured due to their widespread use, quintile grading was not considered, as this would lead to ambiguous Funnel plots colouring and shading in the case of countries within the central This shortcoming is remedied with funnel plots. They reveal a con- quintile deviating significantly in opposite directions. Symmetrically siderable heterogeneity in spread not only between countries but arranged percentiles were selected, with a focus on extreme lower 6of7 European Journal of Public Health Downloaded from http://eurpub.oxfordjournals.org/

Figure 4 Enhanced choropleth plot for caesarean section rates using colouring scheme E. Regions with solid-fill patterns differ significantly at INSERM / ICGM on June 26, 2014 from the EU median rate of 20.5%. The abundance of such countries indicates a bimodal distribution with evident grouping of rates below and above the median and upper 10%. The use of superimposed symbols may lead to dis- Supplementary data tortions and graphical difficulties for small countries such as Luxembourg. In scheme A, there is an inherent bias towards Supplementary data are available at EURPUB online. higher rates because of the larger appearance of the ‘plus’ sign, which is not present in scheme B. Although initially appealing, the discrimination of significance in terms of thickness of country Acknowledgements perimeters has the serious disadvantage that bold black borders The EURO-PERISTAT Scientific Committee: Gerald Haidinger are less distinct for regions with darker shadings. To avoid this (Austria), Sophie Alexander (Belgium), Pavlos Pavlou (Cyprus), bias and also because of obvious problems for small regions, Petr Velebil (Czech Republic), Jens Langhoff Roos (Denmark), scheme D was dropped. In the trade-off between colour bias Luule Sakkeus (Estonia), Mika Gissler (Finland), Be´atrice Blondel (scheme E) and difficulty in readily identifying countries above (France), Nicholas Lack (Germany), Aris Antsaklis (Greece), Istva´n from those below the median (scheme C), the final choice fell on Berbik (Hungary), Sheelagh Bonham (Ireland), Marina Cuttini scheme E. (Italy), Janis Misins (Latvia), Jone Jaselioniene (Lithuania), Yolande Wagener (Luxembourg), Miriam Gatt (Malta), Jan Nijhuis (The Netherlands), Kari Klungsoyr (Norway), Katarzyna Interpreting an enhanced choropleth map Szamotulska (Poland), Henrique Barros (Portugal), Ma´ria ˇ Figure 4 illustrates the advantage of enhancing a standard Chmelova´ (Slovak Republic), Ziva Novak-Antolic (Slovenia), choropleth map with colouring scheme E applied to caesarean Francisco Bolu´ mar (Spain), Karin Gottvall (Sweden) and Alison section rates. The distribution of caesarean section rates is Macfarlane (United Kingdom). Further acknowledgements to con- noteworthy, as one observes that most countries have either sig- tributors to the 2008 European Perinatal Health Report can be nificantly high or significantly low rates, with only Denmark, found in the accompanying supplementary file. Latvia and France in between, reflecting a bimodal distribution. It is attractive to use these data to entertain a hypothesis about Funding patterns of mode of delivery across Europe. Local choices made by pregnant women and clinical custom varying across countries The EURO-PERISTAT project is funded by the Public Health as well as interactions between these factors may well account for Programme of the European Commission, Directorate General of the observed pattern. Public Health (Agreement 2010 13 01). The Directorate General of Reporting of perinatal health indicators 7of7

Public Health had no role in the collection of the data, the writing of 5 Joseph KS, Liu S, Rouleau J, et al. Influence of definition based versus pragmatic the manuscript or the decision to submit for publication. Data birth registration on international comparisons of perinatal and infant mortality: collection for Estonia was supported by grant numbers EE ESF population based retrospective study. BMJ 2012;344:1–9. SF0130018s11 and EE ESF 8325. 6 EURO-PERISTAT: http://www.europeristat.com Accessed [25.05.2012]. 7 Zimbeck M, Mohangoo AD, Zeitlin J. The European perinatal health report: Conflicts of interest: None declared. Delivering comparable data for examining differences in maternal and infant health. Eur J Obstet Gynecol Reprod Biol 2009;146(Issue 2): 149–151. 8 Lack N, Zeitlin J, Krebs L, Ku¨nzel W, Alexander S. Methodological difficulties in the Key points comparison of indicators of perinatal health across Europe. Eur J Obstet Gynecol Reprod Biol 2003;Suppl 111:33–44. Tabular summaries of rates, population sizes and z-scores 9 Gissler M, Mohangoo AD, Blondel B, et al. Perinatal health monitoring in Europe: provide a maximum of statistical information, while being results from the EURO-PERISTAT project. Inform Health Soc Care unwieldy and not easy to read. 2010;35(2):64–79. Funnel plots also contain most of this information and enable rapid identification of outliers; however, they do 10 Goldstein H, Spiegelhalter DJ. League tables and their limitations: statistical issues in not provide information on regional patterns. comparisons of institutional performance. Journal of the Royal Statistical Society Choropleth plots with enhanced patterns where colour (Series A: Statistics in society) 1996;159:385–443. reflects polarity of a country from a reference level, 11 Spiegelhalter DJ. Funnel plots for comparing institutional performance. Statistics in intensity of colouring to indicate size of deviation and Medicine 2005;24(Issue 8): 1185–1202. solid fill versus crosshatching to indicate significance of 12 Macfarlane A. Some statistical approaches to comparisons between perinatal Downloaded from difference can add valuable additional information and mortality rates for small areas. In: Chalmers I, McIlwaine G, editors. Perinatal Audit facilitate the interpretation of regional data. and Surveillance: Proceedings of the Eighth Study Group of the Royal College of As customary geographical maps of indicators of perinatal Obstetricians and Gynaecologists 1980; 274–92. health tend to exaggerate differences between regions, the 13 Lack N, Gerhardinger U. Qualita¨tsvergleiche mit Funnelplots - Pla¨doyer fu¨r eine indication of significance of such differences is expected to einheitliche Methodik (Comparison of quality by means of funnel plots a plea

improve monitoring and assessment of public health policy for a uniform methodology). Zeitschrift fu¨r Evidenz, Fortbildung und Qualita¨tim http://eurpub.oxfordjournals.org/ and practice. Gesundheitswesen 2009;103(Issue 8): 536–41. 14 Lack N, Gerhardinger U. Wirksamkeitsanalyse externer Qualita¨ssicherungsmaßnahmen anhand von Vera¨nderungen in Qualita¨tskennzahlen einzelner Krankenha¨user (An analysis of the effectiveness of external quality References assurance programmes using changes in quality indicators of individual hospitals). Zeitschrift fu¨r Evidenz, Fortbildung und Qualita¨t im Gesundheitswesen 1 EUROSTAT: http:// http://epp.eurostat.ec.europa.eu/portal/page/portal/eurostat/ 2010;104(Issue 6): 503–11. home/ Accessed [25.05.2012]. 15 Ezzati M, Friedman AB, Kulkarni SC, Murray CJL. The reversal of fortunes: Trends 2 WHO: World Health Statistics. http:// http://www.who.int/research/en/ Accessed in county mortality and cross-county mortality disparities in the United States. [25.05.2012]. PLoS Medicine 2008;5(4). at INSERM / ICGM on June 26, 2014 3 Holland WW, editor. European Community Atlas of Avoidable Death. Volume 1 16 SAS Product Documentation: http://support.sas.com/documentation/onlinedoc/ and 2. 3rd edition, Oxford University Press, Oxford, UK, 1997. 91pdf/index_913.html Accessed [25.05.2012]. 4 Zeitlin J, Mohangoo AD, Cuttini M. The European Perinatal Health Report: 17 R Core Team (2012). R: A language and environment for statistical computing. comparing the health and care of pregnant women and newborn babies in Europe. Vienna, Austria, R Foundation for Statistical Computing, ISBN 3-900051-07-0, J Epidemiol Community Health 2009;63(Issue 9): 681–82. URL http://www.R-project.org/. Editorial

The changes since 2004 have not elimi- The second European Perinatal Health nated the wide differences in perinatal health outcomes in Europe, however. Fetal Report: documenting changes over mortality rates at or after 28 weeks of gestation still range from under 2.0 per 1000 6 years in the health of mothers total births in the Czech Republic and Iceland to 4.0 or more in France, Latvia, and babies in Europe the region of Brussels in Belgium and Romania. The countries of the UK also 1 2 1 3 J Zeitlin, A D Mohangoo, M Delnord, M Cuttini, have relatively high fetal mortality rates, 3.8 the EURO-PERISTAT Scientific Committee in England and Wales and 3.6 in Scotland. Neonatal mortality is lower than 2 per 1000 live births in Iceland, Finland and The second European Perinatal Health In some cases, these improvements fol- Sweden but over 4 in Malta and over 5 in Report from the EURO-PERISTAT project lowed public health actions deliberately Romania. Infant mortality ranges from was released on May 27 of this year.1 undertaken at national level. In the about 2 per 1000 in Iceland and Finland to Thirty indicators, compiled from routine Netherlands, the public debate following more than 5 in Malta and Latvia, and statistics in 29 countries, are analysed and this country poor ranking in fetal and reaches 9.8 in Romania. Documenting grouped into four main areas: fetal, neo- neonatal mortality with 2000 and 2004 these differences is important because it natal and child health, maternal health, data led to a series of policy efforts, raises important questions about differences characteristics of the populations and including audits of perinatal deaths in between populations, the effectiveness of healthcare. The report results from a 3-year term babies and establishing a national national maternity care policies and the role collaboration between researchers, clini- commission on perinatal care.4 Also, in of evidence in maternity care. cians and official statisticians in Europe. It 2007 prenatal screening for congenital Healthcare indicators continue to reveal also contains data from two other anomalies was implemented nationwide. marked variations in the approach to European projects: Surveillance of Cerebral As a consequence, between 2004 and childbirth in Europe. Caesarean section Palsy in Europe (SCPE) and European 2010 fetal mortality at or after 28 weeks rates range from 14.8% in Iceland to Surveillance of Congenital Anomalies of gestation declined from 4.3 to 2.9 per 52.2% in Cyprus, instrumental delivery (EUROCAT). 1000 births, and neonatal mortality at or rates range from 0.5% in Romania to Common definitions and inclusion cri- after 24 weeks declined from 2.8 to 2.2 16.4% in Ireland, and episiotomy rates teriamakeitpossibletoovercomesomeof per 1000 live births, while the country’s range from under 7% in Denmark and the differences between countries in the low caesarean rates were maintained. Sweden to over 70% in Cyprus and recording of births and deaths and improve Relating improvements in outcomes to Portugal. The sizes of the maternity units the comparability of the data presented.23 changes in distributions of risk factors is vary as well: from no births in maternity Both results for the year 2010 and compari- more problematic. The prevalence of some units with 5000 or more deliveries in the sons with the 2004 data published in the risk factors has increased in European coun- region of Flanders in Belgium and first European Perinatal Health Report are tries, while others have become less preva- Slovenia to 55.1% in Ireland. There were, included.3 lent. The proportion of mothers aged 35 however, some common trends: caesarean Between 2004 and 2010, fetal, neonatal and older has increased in all countries rates rose in all countries apart from and infant mortality decreased almost except Finland, but the negative impact of Sweden and Finland where rates declined. everywhere. Denmark, Italy and the this change on the health of pregnant Episiotomy rates also tended to decline Netherlands experienced the largest abso- women and neonates may have been moder- over this period, although not in countries lute declines in fetal mortality rates (a ated by better maternal general health and with already low rates in 2004 such as reduction of 1.4 per 1000 total births). care. Multiple birth rates have also increased, England, Latvia and Norway. Absolute declines in neonatal mortality probably as a result of rising maternal age The EURO-PERISTAT network has now were greatest in countries where rates were and more widespread use of assisted repro- been in place for over 10 years, showing higher in 2004 such as some of the Eastern duction techniques. In contrast, smoking that long-lasting multidisciplinary inter- European countries. However, declines during pregnancy declined in almost all national collaboration can be achieved. were also observed in countries with low countries where data were available. Over the years, the number of participating rates in 2004 such as Finland and Sweden, Trends in the rates of preterm live births countries has increased from 15 in 2000 to showing that further decreases are still vary between European countries (figure 1). the current 27 out of 28 European Union possible. Many countries experienced declines in (EU) member states plus Norway, overall rates, as seen in an earlier Switzerland and Iceland. Data were col-

1 EURO-PERISTAT analysis of singleton lected for three different years, 2000, 2004 UMRS 953, Epidemiological Research Unit on 5 and 2010 with the first results from the Perinatal and Women’s and Children’s Health, INSERM, births, while elsewhere rates remained 2 fi Paris, France; Department Child Health, TNO, almost constant. Overall, these ndings year 2000 having been published as a Netherlands Organization for Applied Scientific suggest that the much quoted increase in special issue of the European Journal of Research, Leiden, South Holland, The Netherlands; overall preterm birth rates over the past Obstetrics and Gynaecology in 2003.6 3 ’ Unit of Epidemiology, Bambino Gesù Children s 15 years may now be coming to an end. In Thus, time trends in population characteris- Hospital, Roma, Italy some countries, however, preterm birth rates tics and outcomes can now be explored. Correspondence to Dr Marina Cuttini, Unit of did increase. Understanding the reasons for Yet to build a truly informative Epidemiology, Bambino Gesù Children’s Hospital, Viale Ferdinando Baldelli 41, Roma 00146, Italy; these diverse trends could help shape effect- system, further actions are needed to [email protected] ive preventive public health policies. ensure that each country can provide the

Zeitlin J, et al. J Epidemiol Community Health December 2013 Vol 67 No 12 983 Editorial

System run by Eurostat? Should the European Centre for Disease Control expand its scope, following the example of its US counterpart, traditionally in charge of monitoring perinatal risk factors and outcomes? Or should a speciﬁcpro- gramme linking the various perinatal monitoring initiatives be created under the EU Directorate for Health? A solution is urgently needed to sustain long-term routine projects like this and other important European public health initiatives, as this would build on successful experiences and on a considerable amount of expertise and dedicated human resources.

Collaborators EURO-PERISTAT Scientific Committee: Gerald Haidinger (Austria), Sophie Alexander; (Belgium), Pavlos Pavlou (Cyprus), Petr Velebil (Czech Republic), Jens Langhoff Roos (Denmark), Luule Sakkeus (Estonia), Mika Gissler (Finland), Béatrice Blondel (France), Nicholas Lack (Germany), Aris Antsaklis (Greece), István Berbik (Hungary), Helga Sól Ólafsdóttir (Iceland), Sheelagh Bonham (Ireland), Marina Cuttini (Italy), Janis Misins (Latvia), Jone Jaselioniene (Lithuania), Yolande Wagener (Luxembourg), Miriam Gatt (Malta), Jan Nijhuis (the Netherlands), Kari Klungsøyr (Norway), Katarzyna Szamotulska (Poland), Henrique Barros (Portugal), Mihai Horga (Romania), Ján Čáp (Slovakia), Živa Novak-Antolic (Slovenia), Francisco Bolúmar (Spain), Karin Gottvall (Sweden), Sylvie Berrut (Switzerland) and Alison Macfarlane (UK). Writing committee for the 2010 European Perinatal Health Report: Jennifer Zeitlin, Ashna Mohangoo, Marie Delnord (Editors); Sophie Alexander, Béatrice Blondel, Marie-Hélène Bouvier-Colle, Nirupa Dattani, Mika Gissler, Alison Macfarlane, Karin van der Pal, Katarzyna Szamotulska and Wei Hong Zhang. Contributors JZ, ADM and MD were editors of the European Perinatal Health Report 2010. JZ and MC produced a first draft of this editorial. ADM and MD carried out analyses of submitted data and provided substantive comments to the manuscript. Members of the EURO-PERISTAT Group contributed to analyses and interpretation of data. Funding The EURO-PERISTAT project is co-financed by the Health Programme of the European Union Figure 1 Percentage of preterm live births in 2004 and difference between 2010 and 2004. Directorate General for Health and Consumers which NOTES: Countries ranked according to increasing differences between 2010 and 2004. Rate in also provides funding for SCPE and EUROCAT (grant 2010 can be computed by adding 2004 rate and difference between 2010 and 2004. number 2003131). Competing interests None. Provenance and peer review Commissioned; internally peer reviewed. full set of EURO-PERISTAT indicators immediate gains in many countries. The fi To cite Zeitlin J, Mohangoo AD, Delnord M, et al. using common de nitions and agreed eli- use of individually linked records, anon- J Epidemiol Community Health 2013;67:983–985. gibility criteria. For instance, valid com- ymised to protect confidentiality, would Received 22 August 2013 parisons of mortality rates for extremely provide opportunities for a better under- Accepted 26 August 2013 preterm neonates are still not possible standing of the relationship between Published Online First 19 September 2013 between European countries due to differ- changes in risk factors, healthcare pro- ences in registration criteria for births and vided and outcomes. in practices for recording late terminations The biggest question, however, con- of pregnancy in routine data systems.2 cerns the long-term sustainability of this Agreements about common recording initiative so far based on national guidelines as well as wider linkage of data resources, with central funding provided Editor’s choice from different sources, building on by a series of ad hoc projects under the Scan to access more free content methods already in use in some parts of Health Programme of the European Europe, could enable fuller use of data Union. Should perinatal data monitoring J Epidemiol Community Health 2013;67:983–985. already being collected and yield be included in the European Statistical doi:10.1136/jech-2013-203291

984 Zeitlin J, et al. J Epidemiol Community Health December 2013 Vol 67 No 12 Editorial

REFERENCES 3 Euro-Peristat project in collaboration SCPE, EUROCAT 5 Zeitlin J, Szamotulska K, Drewniak N, et al. Preterm 1 Euro-Peristat project with SCPE and Eurocat. European and EURONEOSTAT. Better statistics for better health birth time trends in Europe: a study of 19 countries. Perinatal health report. The health of pregnant women for pregnant women and their babies in 2004. BJOG 2013;120:1356–65. and babies in Europe in 2010. 2013. http://www. European Perinatal Health Report 2008. http://www. 6 Zeitlin J, Wildman K, Breart G, et al. Selecting an europeristat.com europeristat.com indicator set for monitoring and evaluating perinatal 2 Mohangoo AD, Buitendijk SE, Szamotulska K, et al. 4 Stuurgroep zwangerschap en geboorte. Een goed health in Europe: criteria, methods and results from Gestational age patterns of fetal and neonatal begin. Veilige zorg rond zwangerschap en the PERISTAT project. Eur J Obstet Gynecol Reprod mortality in Europe: results from the Euro-Peristat geboorte. Utrecht: Stuurgroep zwangerschap en Biol 2003;111(Suppl 1):5–14. Project. PLoS ONE 2011;6:e24727. geboorte, 2009.

Zeitlin J, et al. J Epidemiol Community Health December 2013 Vol 67 No 12 985 International Comparisons of Fetal and Neonatal Mortality Rates in High-Income Countries: Should Exclusion Thresholds Be Based on Birth Weight or Gestational Age?

Ashna D. Mohangoo1*,Be´atrice Blondel2, Mika Gissler3,4, Petr Velebil5, Alison Macfarlane6, Jennifer Zeitlin2,7, the Euro-Peristat Scientific Committee" 1 Department of Child Health, TNO, Netherlands Organization for Applied Scientific Research, Leiden, The Netherlands, 2 Epidemiological Research Unit on Perinatal and Women’s and Children’s Health, INSERM UMRS 953, Universite´ Pierre-et-Marie Curie Paris6, Paris, France, 3 Information Department, National Institute for Health and Welfare, Helsinki, Finland, 4 Nordic School of Public Health, Gothenburg, Sweden, 5 Institute for the Care of Mother and Child; Perinatal Centre, Prague, Czech Republic, 6 Maternal and Child Health Research Centre; City University London, London, United Kingdom, 7 UPMC University Paris 06, Paris, France

Abstract

Background: Fetal and neonatal mortality rates are essential indicators of population health, but variations in recording of births and deaths at the limits of viability compromises international comparisons. The World Health Organization recommends comparing rates after exclusion of births with a birth weight less than 1000 grams, but many analyses of perinatal outcomes are based on gestational age. We compared the effects of using a 1000-gram birth weight or a 28-week gestational age threshold on reported rates of fetal and neonatal mortality in Europe.

Methods: Aggregated data from 2004 on births and deaths tabulated by birth weight and gestational age from 29 European countries/regions participating in the Euro-Peristat project were used to compute fetal and neonatal mortality rates using cut-offs of 1000-grams and 28-weeks (2.8 million total births). We measured differences in rates between and within countries using the Wilcoxon signed rank test and 95% confidence intervals, respectively.

Principal Findings: For fetal mortality, rates based on gestational age were significantly higher than those based on birth weight (p,0.001), although these differences varied between countries. The use of a 1000-gram threshold included 8823 fetal deaths compared with 9535 using a 28-week threshold (difference of 712). In contrast, the choice of a cut-off made little difference for comparisons of neonatal deaths (difference of 16). Neonatal mortality rates differed minimally, by under 0.1 per 1000 in most countries (p = 0.370). Country rankings were comparable with both thresholds.

Conclusions: Neonatal mortality rates were not affected by the choice of a threshold. However, the use of a 1000-gram threshold underestimated the health burden of fetal deaths. This may in part reflect the exclusion of growth restricted fetuses. In high-income countries with a good measure of gestational age, using a 28-week threshold may provide additional valuable information about fetal deaths occurring in the third trimester.

Citation: Mohangoo AD, Blondel B, Gissler M, Velebil P, Macfarlane A, et al. (2013) International Comparisons of Fetal and Neonatal Mortality Rates in High- Income Countries: Should Exclusion Thresholds Be Based on Birth Weight or Gestational Age? PLoS ONE 8(5): e64869. doi:10.1371/journal.pone.0064869 Editor: Linda Wright, National Institute of Child Health and Human Development, United States of America Received August 16, 2012; Accepted April 19, 2013; Published May 20, 2013 Copyright: ß 2013 Mohangoo et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Funding: The results from this study are based on data from the Euro-Peristat project, a European project for monitoring and evaluating perinatal outcomes on the European level. The Euro-Peristat project was co-financed by the European Commission (DG-SANCO), grant numbers 2003131 and 20101301. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. No additional external funding was received for this study. Competing Interests: The authors have declared that no competing interests exist. * E-mail: [email protected] " Membership of the Euro-Peristat Scientific Committee is provided in the Acknowledgments.

Introduction Norway or 20 completed weeks in the United States to 26 completed weeks in Italy and Spain [3,5]. Denmark and Sweden There is an ongoing debate about the value of international recorded fetal deaths beginning at 28 completed weeks until 2004 comparisons of fetal and neonatal mortality rates, given differences and 2008, respectively. While only a small proportion of births between countries in recording of births and deaths at borderline occur before 24 completed weeks of gestation (about 1 per 1000) viability [1,2]. Fetal and neonatal mortality rates are highly [6], survival is rare and most of them are either fetal deaths or live sensitive to these inclusion criteria [1,3,4]. Differences in recording births followed by a neonatal death. These births have a criteria are most acute for fetal deaths [5]. These deaths are substantial impact on perinatal mortality statistics [3]. Valid recorded from as early as 16 completed weeks of gestation in analyses of fetal and neonatal mortality across countries thus

PLOS ONE | www.plosone.org 1 May 2013 | Volume 8 | Issue 5 | e64869 Comparisons of Fetal and Neonatal Mortality Rates require specifying common inclusion limits. These criteria are available, population-based regional data could be provided based either on gestational age, or on birth weight or on a instead. combination of these. The Euro-Peristat core indicator list includes fetal and neonatal The World Health Organization (WHO) recommends the use mortality. The fetal mortality rate is defined as the number of of a 1000-gram threshold for international comparisons of deaths before or during birth in a given year per 1000 live and perinatal mortality rates [7]. This limit makes it possible to stillbirths in the same year. The neonatal mortality rate is defined provide a measure of the health burden of third trimester perinatal as the number of deaths at 0 to 27 days after live birth in a given deaths, since 1000 grams corresponds approximately to the birth year per 1000 live births in the same year. Euro-Peristat collects weight at 28 completed weeks of gestation, the beginning of the data on births and deaths at or after 22 weeks of gestation, third trimester. This measure provides only a partial view of regardless of birth weight. Aggregated data on the number of live overall mortality, since a large proportion of deaths in high- births, fetal and neonatal deaths by each week of gestation and by income countries (between 25–60%) occur to babies born in the birth weight intervals of 500 grams were collected. These data second trimester [3,6], but using this threshold has the benefit of were used to calculate fetal and neonatal mortality rates for births enabling greater comparability between countries. Participants in and deaths weighing 1000 grams and over and for those born at or a recent international collaboration on stillbirths agreed that an after 28 completed weeks. analysis of third trimester deaths has public health relevance for Twenty-seven countries were able to provide data to calculate international comparisons in high-income countries [8]. fetal mortality rates with birth weight and gestational age The aim in international comparisons is to maximise both thresholds. Germany, Hungary, Ireland and Italy did not have comparability and scientific and policy relevance. The primary data on neonatal deaths by birth weight or gestational age. Some aim of using a birth weight threshold for international comparisons countries could only provide data on some regions (Valencia in is to ensure comparability because birth weight measures are Spain, Brussels and Flanders in Belgium). Data from France came considered to be less prone to error than calculations of gestational from a one-week national perinatal survey in October 2003, vital age. When the date of the last menstrual period (LMP) is used registration, and neonatal death certificates. Data on neonatal alone to calculate gestational age, the results can be inaccurate [9], deaths from England and Wales related to 2005 and data from especially if the woman has no antenatal care or if antenatal care Italy were for 2003. Table S1 presents additional information starts late in pregnancy. However, most high-income countries use about the data sources. These constraints reflect the diversity of a clinical estimate of gestational age that incorporates information sources for perinatal health data in Europe [5]. from dating ultrasounds and is therefore of better quality [10]. Birth weight data also have limitations since babies who are Missing Data stillborn or die before they can be transferred to a neonatal unit Most countries had fewer than 5% of data missing for fetal and may not be systematically weighed [11]. neonatal deaths by birth weight and gestational age, as presented Gestational age is generally considered to be a more relevant in Table S2. However, there were some exceptions. The variable than birth weight for studying perinatal outcomes. Recent percentages of fetal deaths with birth weights missing were European cohorts have analysed the outcome for very preterm 30.7% in Denmark, 25.0% in Italy, 22.7% in Brussels, 13.9% in rather than very low birth weight babies, as gestational age has a Valencia, 6.4% in Portugal, 5.9% in Luxembourg, and 5.1% in better prognostic value [12–16]. Furthermore, when obstetricians France. Gestational age was missing for 17.0% of fetal deaths in are making decisions during pregnancy, they have reasonably Brussels, 11.7% in Valencia and 9.5% in Portugal. We excluded precise information about gestational age but not about birth countries where the proportion of fetal deaths with missing birth weight. Finally, birth weight distributions differ between and weight was significantly different from the proportion with missing within populations and European comparisons have found that gestational age. These were Denmark (30.7% of birth weights vs. the birth weight at which mortality is lowest varies between 4.2% of gestational ages) and Italy (25.0% of birth weights and 0% European countries [17]. Using birth weight cut-offs will exclude for gestational age). These divergent proportions of missing data relatively more births and deaths in countries where average birth would have biased our ability to compare rates. weights are lower and this may introduce bias. For neonatal deaths, fewer countries had high proportions of While the hypothesis underlying the current WHO recommen- data missing. Over 5% of birth weights were missing for Denmark dation is that the 1000-gram threshold provides a good (14.8%), Luxembourg (9.1%), Sweden (7.1%), Scotland (6.8%), approximation for the 28th week of gestation or the beginning of and Valencia (5.8%). Gestational ages were missing for over 5% in the third trimester, this hypothesis has not been tested. The aim of Luxembourg (9.1%), Denmark (7.0%), Portugal (6.8%), and this analysis was therefore to compare the use of a 1000-gram birth Valencia (6.8%). As with fetal deaths, we excluded countries with weight threshold with a 28-week gestational age threshold in terms of their impact on reporting of fetal and neonatal mortality rates highly divergent proportions of missing birth weights and within European countries and on comparisons between Europe- gestational ages. We therefore excluded Denmark (14.8% of birth an countries. weights vs. 7.0% of gestational ages) and Sweden (7.1% of birth weights and 0% for gestational age). Live birth data were missing for less than 5% with the exception of Brussels and Valencia where Methods 6.3% and 5.5% of gestational ages were missing respectively. This study was embedded within Euro-Peristat, which devel- For countries included in the analyses, we excluded missing data oped a list of valid and reliable indicators for monitoring and from our primary analyses as this would reflect the reality if these evaluating perinatal health in the European Union (EU) [18]. cut-offs were used, but we also did a second set of analyses with Twenty-five EU member states and Norway participated. Detailed missing data distributed according to observed birth weight and information on the design and methods is available elsewhere gestational age distributions for live births, and fetal and neonatal [5,19,20]. National population-based data for each indicator for deaths separately. the year 2004 were requested in aggregated form from members of the Euro-Peristat Scientific Committee. If national data were not

PLOS ONE | www.plosone.org 2 May 2013 | Volume 8 | Issue 5 | e64869 Comparisons of Fetal and Neonatal Mortality Rates

Statistical Analysis based on gestational age were still significantly higher than those We calculated fetal and neonatal mortality rates with 95% based on birth weight (p = 0.002); while the choice of a cut-off confidential intervals, using both birth weight and gestational age made no difference for comparisons of neonatal mortality rates thresholds. We also computed differences between rates with (p = 0.380). confidence intervals to test whether these were significant within Fetal and neonatal mortality rates computed using a birth countries. To test whether there was a systematic difference weight threshold were highly correlated with rates computed using between countries in rates based on birth weight versus gestational a gestational age threshold, with Spearman rank correlations of age we used the non-parametric Wilcoxon signed rank test. In r = 0.952 (p,0.001, n = 25) for fetal mortality and r = 0.963 addition, we used the Wilcoxon rank sum test to assess whether (p,0.001, n = 21) for neonatal mortality. Country rankings were rates based on a 28-week threshold minus a 1000-gram threshold therefore similar with a few exceptions such as Brussels which differed significantly across countries. Finally, we tested the ranked sixth for birth weight and thirteenth for gestational age. correlation between these two rates using the Spearman rank test Even for the Netherlands and England and Wales, where to assess how these affected country rankings. These statistical tests differences in fetal mortality rates calculated using the two were repeated on recalculated rates after imputation of missing definitions were significantly different, their ranks only differed observations, as described above, to ensure that the addition of by two places (19 and 18 out of 25 for birth weight to 21 and 20 these data would not change our results. Analyses were done with out of 25 for gestational age, respectively (data not shown in table). SPSS version 17.0 for Windows (SPSS Inc, Chicago, IL, USA). Discussion Results Our analysis showed that fetal mortality rates in European Table 1 presents fetal and neonatal mortality rates using a birth countries were higher when based on a 28-week gestational age weight cut-off of 1000 grams. The range for fetal deaths was 1.6 to threshold compared with a 1000-gram birth weight threshold, 4.7 per 1000 live and stillbirths and the range of neonatal deaths whereas the choice of a threshold made little difference for was 1.1 to 4.3 per 1000 live births. Also shown are the same rates neonatal mortality rates. These results suggest that a substantial with a gestational age cut-off of 28 weeks. They ranged from 1.7 to proportion of fetal deaths occurring at or after 28 weeks of 4.9 per 1000 for fetal deaths and 1.3 to 4.0 per 1000 for neonatal gestation have a birth weight under 1000-grams. Despite this deaths. difference, however, the selection of a cut-off did not change Except for the Czech Republic (0.16%) and Estonia (0.21%), countries’ relative positions. The small differences between where rates were 0.2 per 1000 higher with a birth weight cut-off, neonatal mortality rates calculated using birth weight and most countries had higher rates of fetal deaths when a gestational gestational age cut-offs and the similarity in rankings suggest that age cut-off was used, as illustrated in Figure 1. For seven out of 25 differences in average birth weight between populations do not countries/regions, the two rates were very similar with minimal create a bias when rates are computed using a birth weight cut-off. differences of 0.1 per 1000 or less. The widest differences were There are a number of limitations to this study. Most notably its 0.8 per 1000 in Brussels (0.82%) and France (0.76%). At an reliance on aggregated data meant we could not cross tabulate the individual country level, differences between fetal mortality rates birth weight and gestational age distributions of the excluded based on gestational age and those based on birth weight were not births and deaths. Our data also date from 2004 and practices in significantly different from zero, except in the Netherlands where registration and care of very preterm infants may have changed the difference was 0.50 per 1000 with 95% confidence interval since this time. However, these changes most likely occurred for 0.09–0.91 (p = 0.018) and England and Wales where the rate births with a birth weight under 1000 grams or before 28 difference was 0.45 per 1000 with 95% confidence interval 0.23– completed weeks of gestation which are excluded from our 0.66 (p,0.001). analysis. At the time these data were compiled, a widespread In contrast, differences between neonatal mortality rates were consensus in Europe existed about the importance of active care minimal, with 15 out of 21 countries/regions having differences for infants born at or after 28 weeks of gestation or 1000 grams or between 20.1 and +0.1 per 1000 (Figure 1). Rates calculated with more [21]. Furthermore, in all countries live and stillbirths born at a gestational age cut-off were not significantly higher or lower than these thresholds were included in routine data collection systems those with a birth weight cut-off, although in Latvia (20.35%), [5–7]. Brussels (+0.27%) and Malta (+0.25%) differences were 0.25 per Another limitation relates to missing data; many countries had 1000 or more. some birth weight and gestational age data missing and this will Differences between countries in fetal mortality rates based on have affected absolute rates. While we excluded countries with gestational age compared with those based on birth weight were highly divergent proportions of birth weight and gestational age significant (p,0.001 for Wilcoxon signed rank test). The corre- data missing, some countries still had more data missing among sponding neonatal mortality rates did not differ significantly deaths than among live births. Missing data could be more between countries (p = 0.370), however twelve countries had a prevalent among extremely preterm or very low birth weight positive difference while eight had a negative and there was one babies, which would limit their influence on analyses of rates using tie. In total, 8823 fetal deaths were included when a 1000-gram 28 weeks or 1000 grams thresholds. Because we were using threshold was used compared with 9535 with a 28-week threshold, aggregated data, we were limited in our ability to investigate this a difference of 712 fetal deaths (7.5% of all fetal deaths). In further. However, even if this were not the case, these missing data contrast, the difference in neonatal deaths was minimal, 4710 are unlikely to change our conclusions. This was shown when we using a 1000-gram threshold versus 4726 using a 28-week repeated our analyses including missing data based on observed threshold (a difference of 16). distributions of birth weight and gestational age for fetal and Results did not change when the observed birth weight and neonatal deaths and live births. Nonetheless, this analysis showed gestational age distributions for fetal and neonatal deaths and live that proportions of missing gestational age and birth weight varied births were used to include births and deaths with missing birth between countries and this may have an impact on the comparison weights and gestational ages in the analyses. Fetal mortality rates of mortality rates when these thresholds are used, regardless of the

PLOS ONE | www.plosone.org 3 May 2013 | Volume 8 | Issue 5 | e64869 Comparisons of Fetal and Neonatal Mortality Rates

Table 1. Fetal mortality rates per 1000 total births and neonatal mortality rates per 1000 live births with 95% confidence intervals [CI].

Fetal mortality Neonatal mortality

Birth weight $1000 grams Gestational age $28 weeks Birth weight $1000 grams Gestational age $28 weeks

Country/region Total births Rate [95% CI] Total births Rate [95% CI] Live births Rate [95% CI] Live births Rate [95% CI]

Austria 78820 2.33 [2.0–2.7] 78794 2.49 [2.1–2.8] 78636 1.44 [1.2–1.7] 78598 1.39 [1.1–1.6] Belgium: Brussels 15752 2.54 [1.8–3.3] 15176 3.36 [2.4–4.3] 15712 1.91 [1.2–2.6] 15125 2.18 [1.4–2.9] Belgium: Flanders 60642 2.67 [2.3–3.1] 60679 2.85 [2.4–3.3] 60480 1.37 [1.1–1.7] 60506 1.39 [1.1–1.7] Czech Republic 97544 2.56 [2.2–2.9] 97480 2.40 [2.1–2.7] 97294 1.12 [0.9–1.3] 97365 1.25 [1.0–1.5] Estonia 13945 3.37 [3.3–4.3] 13939 3.16 [2.2–4.1] 13898 2.52 [1.7–3.4] 13895 2.66 [1.8–3.5] Finland 57482 1.97 [1.6–2.3] 57407 2.04 [1.7–2.4] 57369 1.20 [0.9–1.5] 57290 1.29 [1.0–1.6] France 14551 4.12 [3.1–5.2] 14540 4.88 [3.8–6.0] 761290 1.50 [1.4–1.6] 765752 1.48 [1.4–1.6] Germany 644654 2.39 [2.3–2.5] 645401 2.55 [2.4–2.7] Hungary 94801 3.55 [3.2–3.9] 94900 3.73 [3.3–4.1] Ireland 62077 3.82 [3.3–4.3] 62097 4.28 [3.8–4.8] Latvia 20393 4.71 [3.8–5.6] 20382 4.86 [3.9–5.8] 20297 4.34 [3.4–5.2] 20283 3.99 [3.1–4.9] Lithuania 29510 3.83 [3.1–4.5] 29502 3.93 [3.2–4.6] 29397 2.89 [2.3–3.5] 29386 2.93 [2.3–3.5] Luxembourg 5296 2.45 [1.1–3.8] 5384 2.79 [1.4–4.2] 5283 1.51 [0.5–2.6] 5369 1.30 [1.7–5.5] Malta 3889 3.86 [1.9–5.8] 3894 3.85 [1.9–5.8] 3874 3.36 [1.5–5.2] 3879 3.61 [1.7–5.5] The Netherlands 181014 3.77 [3.5–4.1] 178710 4.27 [4.0–4.6] 180332 1.96 [1.8–2.2] 177947 1.93 [1.7–2.1] Norway 57450 2.75 [2.3–3.2] 57004 2.84 [2.4–3.3] 56911 1.32 [1.0–1.6] 56925 1.35 [1.1–1.7] Poland 356571 3.54 [3.3–3.7] 356734 3.77 [3.6–4.0] 355307 2.92 [2.7–3.1] 355389 3.00 [2.8–3.2] Portugal 108948 2.64 [2.3–2.9] 109136 2.69 [2.4–3.0] 108660 1.51 [1.3–1.7 108842 1.45 [1.2–1.7] Slovenia 17840 3.48 [2.6–4.3] 17849 3.53 [2.7–4.4] 17778 1.29 [0.8–1.8] 17786 1.35 [0.8–1.9] Slovak Republic 52301 1.63 [1.3–2.0] 52332 1.66 [1.3–2.0] 52216 1.63 [1.3–2.0] 52245 1.70 [1.3–2.1] Spain: Valencia 49505 2.95 [2.5–3.4] 48279 3.11 [2.6–3.6] 49359 1.22 [0.9–1.5] 48129 1.25 [0.9–1.6] Sweden 99928 2.87 [2.5–3.2] 100111 3.16 [2.8–3.5] UK: England and Wales 637653 3.68 [3.5–3.8] 637521 4.13 [4.0–4.3] 640374 1.59 [1.5–1.7] 637521 1.59 [1.5–1.7] UK: Northern Ireland 22351 3.62 [2.8–4.4] 22355 3.76 [3.0–4.6] 22270 1.53 [1.0–2.0] 22271 1.44 [0.9–1.9] UK: Scotland 52907 4.06 [3.5–4.6] 52860 4.58 [4.0–5.2] 52692 1.54 [1.2–1.9] 52618 1.48 [1.2–1.8]

Cyprus and Greece (no data on fetal and neonatal death by birth weight and gestational age), Germany, Hungary, Ireland and Italy (no data on neonatal death by birth weight and gestational age), Denmark and Italy (excluded from fetal death comparisons, because of highly divergent missing data on birth weight versus gestational age), Denmark and Sweden (excluded from neonatal death comparisons, because of highly divergent missing data on birth weight versus gestational age). doi:10.1371/journal.pone.0064869.t001 choice of threshold. These proportions should be reported in As concluded by a recent review of stillbirths in high-income comparative analyses. countries, small for gestational age is the pregnancy condition with Finally, while the Euro-Peristat project requests data based on the highest population attributable risk (measured at one out of the best obstetric estimate of gestational age in weeks from clinical four for stillbirths) [25]. Fetal growth restriction is a particularly records, it was not possible to evaluate differences in the ways in important risk factor for antepartum deaths, which constitute over which participating countries actually measure gestational age. In 80% of fetal deaths in high-income countries [4]. Second, fetal most European countries, however, dating ultrasounds are a weight loss after antepartum death may also contribute to lower standard component of care during pregnancy and most women birth weights, although the extent of this phenomenon is still have their first antenatal visit in the first trimester [22–24]. unknown [26]. Finally, some deaths may predate delivery and this Using a birth weight cut-off of 1000 grams resulted in 712 fewer would lead to lower average birth weights for stillbirths. More fetal deaths overall compared with using a cut-off of 28 completed detailed analysis of stillbirths with a gestational age of 28 weeks weeks of gestation leading to systematically lower fetal death rates. and over, but birth weights under 1000 grams is needed to better A previous review also suggested that stillbirth rates were higher understand the relative contribution of these different explana- using gestational age limits based on Norwegian data showing that tions. a 500 grams cut-off point excluded more stillbirths than a 22 week While growth restriction is also a risk factor for neonatal death, cut-off point; neonatal deaths were not included in this study [4]. the magnitude of the association may be less strong, especially in By comparing neonatal deaths with fetal deaths, our results show the gestational age and birth weight bands considered in this that this effect specifically relates to stillbirths and does not reflect analysis. Recent studies in France and New Zealand found that the gestational age and birth weight distribution of all births. 17% and 13% of all neonatal deaths were below the tenth There are several possible explanations for this finding. First, percentiles of national standards [27,28]. This compared to studies using a birth weight cut-off may exclude growth restricted fetuses. of stillbirth where between 40% and 60% are associated with

PLOS ONE | www.plosone.org 4 May 2013 | Volume 8 | Issue 5 | e64869 Comparisons of Fetal and Neonatal Mortality Rates

Figure 1. Differences in mortality rates based on gestational age $28 weeks minus birth weight $500 grams. Austria (AT), Brussels (BE: BR), Flanders (BE: FL), Cyprus (CY), Czech Republic (CZ), Denmark (DK), Estonia (EE), Finland (FI), France (FR), Germany (DE), Greece (GR), Hungary (HU), Ireland (IE), Italy (IT), Latvia (LV), Lithuania (LT), Luxembourg (LU), Malta (MT), the Netherlands (NL), Norway (NO), Poland (PL), Portugal (PT), Slovenia (SI), Slovak Republic (SK), Valencia region of Spain (ES), Sweden (SE), and the United Kingdom (UK): England and Wales combined (UK: EW), Northern Ireland (UK: NI), and Scotland (UK: SC). doi:10.1371/journal.pone.0064869.g001 growth restriction [29]. Growth restriction in this context reflects a Acknowledgments wide range of underlying pregnancy complications which contribute to poor growth and adverse perinatal outcomes. The members of the Euro-Peristat Scientific Committee are: Gerald Haidinger, Department of Epidemiology, The Medical University of Vienna (Austria); Sophie Alexander, School of Public Health, Universite´ Conclusions Libre de Bruxelles (Belgium); Pavlos Pavlou, Health Monitoring Unit, In the European countries included in our analysis, fetal Ministry of Health (Cyprus); Petr Velebil, Institute for the Care of Mother mortality rates calculated using a threshold of 28 weeks of and Child, Perinatal Centre Prague (Czech Republic); Jens Langhoff Roos, gestation were higher than those based on birth weight cut-offs of Obstetrics Clinic Rigshospitalet, Copenhagen University (Denmark); Luule 1000 grams, probably due in part to the role of intra-uterine Sakkeus, Estonian Institute for Population Studies, Tallinn University (Estonia); Mika Gissler, THL National Institute for Health and Welfare growth restriction in antepartum fetal deaths. Assessing the health (Finland); Be´atrice Blondel, Epidemiological Research Unit on Perinatal burden of third trimester fetal deaths using a cut-off based on and Women’s and Children’s Health, INSERM UMRS 953, Universite´ gestational age provides valuable additional information. Com- Pierre-et-Marie Curie Paris6 (France); Nicholas Lack, Bavarian Working parisons based on this cut-off are possible in countries where a Group for Quality Assurance (Germany); Aris Antsaklis, Department of clinical estimate of gestational age is recorded in routine data Obstetrics and Gynaecology, Athens University (Greece); Istva´n Berbik, sources and where women have access to early antenatal care and Department of Obstetrics and Gynaecology, Vaszary Kolos Teaching dating ultrasound as is the case in European and other high- Hospital (Hungary); Sheelagh Bonham, National Perinatal Reporting income countries. Scheme, Economic and Social Research Institute (Ireland); Marina Cuttini, Unit of Epidemiology, Pediatric Hospital of Baby Jesus (Italy); Janis Misins, Center for Disease Prevention and Control (Latvia); Jone Supporting Information Jaselioniene, Department Epidemiology and Biostatistics, Institute of Hygiene (Lithuania); Yolande Wagener, Department of Health, Ministry Table S1 Data sources used for data on live births, fetal of Health (Luxembourg); Miriam Gatt, Department of Health Information and neonatal deaths in Europe in 2004. and Research, National Obstetric Information Systems (NOIS) Register (DOCX) (Malta); Jan Nijhuis, Department Obstetrics and Gynaecology, Maastricht University Medical Centre (The Netherlands); Kari Klungsoyr, Medical Table S2 Percentage of missing birth weights (BW) and Birth Registry of Norway, University of Bergen (Norway); Katarzyna gestational ages (GA). *Denmark (fetal and neonatal mortal- Szamotulska, Department of Epidemiology, National Research Institute of ity), Italy (fetal mortality) and Sweden (neonatal mortality) were Mother and Child (Poland); Henrique Barros, Department of Hygiene and excluded from analysis because of substantial difference in missing Epidemiology, University of Porto Medical School (Portugal); Ma´ria data by birth weight and gestational age. Proportions of 5% and Chmelova´, National Health Information Centre (Slovak Republic); Zˇ iva over missing data are presented in bold. Novak-Antolic, Perinatology Unit, University Medical Center (Slovenia); (DOCX) Francisco Bolu´mar, Department of Health Sciences and Social Medicine, University of Alcala´ (Spain); Karin Gottvall, Department of Statistics, Monitoring and Evaluation, The National Board of Health and Welfare, (Sweden); Alison Macfarlane, Maternal and Child Health Research Centre, City University London (United Kingdom).

PLOS ONE | www.plosone.org 5 May 2013 | Volume 8 | Issue 5 | e64869 Comparisons of Fetal and Neonatal Mortality Rates

The authors acknowledge the following contributors to the 2008 Research; Raymond Galea, Department of Obstetrics and Gynecology, European Perinatal Health Report: Austria Christian Vutuc, Abteilung University of Malta; The Netherlands Sabine Anthony, Simone fu¨r Epidemiologie Zentrum fu¨r Public Health der Med. Univ. Wien; Buitendijk, Ashna Mohangoo, Ab Rijpstra, TNO Quality of Life, Jeannette Klimont, Statistics Austria; Belgium Sophie Alexander, Wei- Department Prevention and Care, Section Reproduction and Perinatology, Hong Zhang, Universite´ Libre de Bruxelles, School of Public Health, Leiden; Jan Nijhuis, Maastricht University Medical Center, Department of Reproductive Health Unit; Guy Martens, Study Center for Perinatal Obstetrics and Gynecology; Chantal Hukkelhoven, The Netherlands Epidemiology (SPE), Edwige Haelterman, Myriam De Spiegelaere, Perinatal Registry; Norway Lorentz Irgens, Kari Klungsoyr Melve, Brussels Health and Social Observatory; Cyprus Pavlos Pavlou, Maria University of Bergen, Medical Birth Registry of Norway; Jon Gunnar Athanasiadou, Ministry of Health, Health Monitoring Unit; Andreas Tufta, Medical Birth Registry of Norway; Poland Katarzyna Szamo- Hadjidemetriou, Christina Karaoli, Neonatal Intensive Care Unit, tulska, Department of Epidemiology, National Research Institute of Makarios III Hospital; Czech Republic Petr Velebil, Institute for the Mother and Child; Bogdan Chazan, Holy Family Hospital; Portugal Care of Mother and Child, Vit Unzeitig, Department of Obstetrics and Henrique Barros, Sofia Correia, University of Porto Medical School, Gynecology, Masaryk University Brno; Denmark Jens Langhoff Roos, Department of Hygiene and Epidemiology; Slovak Republic Jan Cap, Obstetrics Clinic, Rigshospitalet, Copenhagen University; Steen Rasmus- Jarmila Hajnaliova, National Health Information Center; Slovenia Zˇ iva sen, Sundhedsstyrelsen National Board of Health; Estonia Luule Sakkeus, Novak-Antolicˇ, Ivan Verdenik, University Medical Centre, Perinatology Kati Karelson, Mare Ruuge, National Institute for Health Development, Unit, Polonca Truden-Dobrin, Center for Health and Health Care Department of Health Statistics; Finland Mika Gissler, National Research Research, Institute of Public Health of the Republic of Slovenia; Spain and Development Centre for Welfare and Health (STAKES); Anneli Pouta Francisco Bolumar, Universidad de Alcala Facultad de Medecina; Ramon National Public Health Institute (KTL), Department of Child and Prats, Departament de Salut Direccio General Salut Publica; Carmen Adolescent Health; France Beátrice Blondel, Marie-He´le`ne Bouvier- Barona, Perinatal Health Unit Public Health Board, Isabel Rıó, CIBER Colle, Ge´rard Breárt, Jennifer Zeitlin, Meagan Zimbeck, INSERM U953; Epidemiologıá y Salud Pu´blica (CIBERESP); Sweden Gunilla Lindmark, Christine Cans, SCPE Service d’Information et d’Informatique Me´dicale IMCH, Akademiska sjukhuset; Milla Bennis, National Board of Health and (SIIM); Germany Nicholas Lack, Bavarian Working Group for Quality Welfare; United Kingdom, Alison Macfarlane, Nick Drey, City Assurance, Klaus Doebler, Federal Quality Assurance Office BQS; University London; Angela Bell, Health Promotion Agency for Northern Greece Aris Antlaklis, Peter Drakakis, Athens University, Department Ireland CEMACH; Jim Chalmers, Etta Shanks, Information Services of Obstetrics and Gynecology, Division of Maternal and Fetal Medicine; Division, NHS National Services Scotland; Di Goodwin, Kath Moser, Hungary Istvań Berbik, Vaszary Kolos Teaching Hospital, Department Nirupa Dattani, Office for National Statistics; Gwyneth Thomas, Health of Obstetrics and Gynecology; Istvań Szabo´, Department of Obstetrics and Statistics and Analysis Unit, Statistical Directorate, Welsh Assembly Gynaecology, Medical Faculty, Scientific University of Pećs; Ireland Government. Sheelagh Bonham, Jacqueline O’Reilly, Economic and Social Research Institute (ESRI); Italy Marina Cuttini, Pediatric Hospital of Baby Jesus, Author Contributions Unit of Epidemiology; Sabrina Prati, Cinzia Castagnaro, Silvia Bruzzone, Marzia Loghi, Istituto Nazionale di Statistica, ISTAT; Latvia Jautrite Conceived and designed the experiments: ADM JZ. Performed the Karaskevica, Irisa Zile, Health Statistics and Medical Technologies State experiments: ADM JZ. Analyzed the data: ADM JZ. Contributed Agency; Ilze Kreicberga, Riga Maternity Hospital; Lithuania Aldona reagents/materials/analysis tools: ADM JZ Euro-Peristat Scientific Gaizauskiene, Kotryna Paulauskiene, Lithuanian Health Information Committee Members. Wrote the paper: ADM JZ. Substantially comment- Centre; Luxembourg Yolande Wagener, Ministe`re de la Sante´, Direction ed on the manuscript: BB MG PV AM. Provided the data and commented de la Sante´, Division de la Me´decine Pre´ventive et Sociale; Malta Miriam on the final draft of the manuscript: Euro-Peristat Scientific Committee Gatt, Kathleen England, Department of Health Information and Members.

References 1. Joseph KS, Liu S, Rouleau J, Lisonkova S, Hutcheon JA, et al. (2012) Influence 11. Lawn JE, Blencowe H, Pattinson R, Cousens S, Kumar R, et al. (2011) ofdefinition based versus pragmatic birth registration on international compar- Stillbirths: Where? When? Why? How to make the data count? Lancet. isons of perinatal and infant mortality: population based retrospective study. 377(9775): 1448–1463. BMJ 344: e746 doi:10.1136/bmj.e746. 12. Larroque B, Ancel PY, Marret S, Marchand L, Andre´ M, et al. (2008) 2. Zeitlin J, Blondel B, Mohangoo AD, Cuttini M, Macfarlane A, et al. (2012) Neurodevelopmental disabilities and special care of 5-year-old children born Rapid response to: Influence of definition based versus pragmatic birth before 33 weeks of gestation (the EPIPAGE study): a longitudinal cohort study. registration on international comparisons of perinatal and infant mortality: Lancet. 371(9615): 813–820. population based retrospective study. BMJ. 13. Vanhaesebrouck P, Allegaert K, Bottu J, Debauche C, Devlieger H, et al. (2004) 3. Mohangoo AD, Buitendijk SE, Szamotulska K, Chalmers J, Irgens LM, et al. The EPIBEL study: outcomes to discharge from hospital for extremely preterm (2011) Gestational age patterns of fetal and neonatal mortality in Europe: results infants in Belgium. Pediatrics. 114(3): 663–675. from the Euro-Peristat project. PloS one 6(11): e24727. 14. Moore T, Hennessy EM, Myles J, Johnson SJ, Draper ES, et al. (2012) 4. Frøen JF, Gordijn SJ, Abdel-Aleem H, Bergsjø P, Betran A, et al. (2009) Making Neurological and developmental outcome in extremely preterm children born in stillbirths count, making numbers talk - issues in data collection for stillbirths. England in 1995 and 2006: the EPICure studies.BMJ. 4;345: e7961. BMC Pregnancy Childbirth 9: 58. doi:10.1136/bmj.e7961. 5. Gissler M, Mohangoo AD, Blondel B, Chalmers J, Macfarlane A, et al. (2010) 15. EXPRESS Group (2010) Incidence of and risk factors for neonatal morbidity Perinatal health monitoring in Europe: results from the EURO-PERISTAT after active perinatal care: extremely preterm infants study in Sweden project. Inform Health Soc Care 35(2): 64–79. (EXPRESS).Acta Paediatr. 99(7): 978–92. doi:10.1111/j.1651- 6. EURO-PERISTAT project in collaboration with SCPE, EUROCAT and 2227.2010.01846.x. Epub 2010 Apr 26. EURONEOSTAT (2008) Better statistics for better health for pregnant women 16. Wood NS, Marlow N, Costeloe K, Gibson AT, Wilkinson AR (2000). and their babies in 2004. European Perinatal Health Report. Available at www. Neurologic and developmental disability after extremely preterm birth. EPICure europeristat.com. Study Group. N Engl J Med. 343(6): 378–384. 7. World Health Organization. ICD-10: International Statistical Classificationof 17. Graafmans WC, Richardus JH, Borsboom GJ, Bakketeig L, Langhoff-Roos J, Diseases and Related Health Problems - Instruction Manual. 2. In: WHO, ed. et al. (2002) Birth weight and perinatal mortality: a comparison of ‘‘optimal’’ Geneva: WHO; 2004: http: //www.who.int/classifications/icd/ birth weight in seven Western European countries. Epidemiology. 13(5): 569– icdonlineversions/en/index.html. Accessed http: //www.who.int/ 574. classifications/icd/icdonlineversions/en/index.html accessed 25.07.2011. 18. Zeitlin J, Wildman K, Breart G (2003) Perinatal health indicators for Europe: an 8. Flenady V, Middleton P, Smith GC, Duke W, Erwich JJ, et al. (2011) Stillbirths: introduction to the PERISTAT project. Eur J Obstet Gynecol Reprod Biol. 111 the way forward in high-income countries. Lancet. 377(9778): 1703–1717. Suppl 1: S1–4. 9. Kramer MS, McLean FH, Boyd ME, Usher RH (1988) The validity of 19. Zeitlin J, Wildman K, Breart G, Alexander S, Barros H, et al. (2003) gestational age estimation by menstrual dating in term, preterm, and postterm PERISTAT: indicators for monitoring and evaluating perinatal health in gestations. JAMA. 260(22): 3306–8. Europe. Eur J Public Health. 13(3 Suppl): 29–37. 10. Ananth CV (2007) Menstrual versus clinical estimate of gestational age dating in 20. Zeitlin J, Wildman K, Breart G, Alexander S, Barros H, et al. (2003) Selecting the United States: temporal trends and variability in indices of perinatal an indicator set for monitoring and evaluating perinatal health in Europe: outcomes. Paediatr Perinat Epidemiol. 21 Suppl 2: 22–30. criteria, methods and results from the PERISTAT project. Eur J Obstet Gynecol Reprod Biol. 111 Suppl 1: S5–S14.

PLOS ONE | www.plosone.org 6 May 2013 | Volume 8 | Issue 5 | e64869 Comparisons of Fetal and Neonatal Mortality Rates

21. Kollee LA, Cuttini M, Delmas D, Papiernik E, den Ouden AL, et al. (2009) 26. Chard T (2001) Does the fetus lose weight in utero following fetal death: a study Obstetric interventions for babies born before 28 weeks of gestation in Europe: in preterm infants. BJOG 108(11): 1113–1115. results of the MOSAIC study. Bjog 116(11): 1481–91. Epub 2009 Jul 7. 27. Battin MR, McCowan LM, George-Haddad M, Thompson JM (2007) Fetal 22. Langhoff-Roos J, Kesmodel U, Jacobsson B, Rasmussen S, Vogel I (2006) growth restriction and other factors associated with neonatal death in New Spontaneous preterm delivery in primiparous women at low risk in Denmark: Zealand. Aust N Z J Obstet Gynaecol. 47(6): 457–63. population based study. Bmj 332(7547): 937–9. 28. Carayol M, Bucourt M, Cuesta J, Zeitlin J, Blondel B (2012) Neonatal mortality 23. Norman JE, Morris C, Chalmers J (2009) The effect of changing patterns of in Seine-Saint-Denis: Analysis of neonatal death certificates. J Gynecol Obstet obstetric care in Scotland (1980–2004) on rates of preterm birth and its neonatal Biol Reprod (Paris). doi: pii:S0368-2315(12)00335-3. 10.1016/ consequences: perinatal database study. PLoS medicine 6(9): e1000153. j.jgyn.2012.10.012. [Epub ahead of print]. 24. Schaaf JM, Mol BW, Abu-Hanna A, Ravelli AC (2011) Trends in preterm birth: 29. Gardosi J, Kady SM, McGeown P, Francis A, Tonks A (2005) Classification of singleton and multiple pregnancies in the Netherlands, 2000–2007. BJOG stillbirth by relevant condition at death (ReCoDe): population based cohort 118(10): 1196–204. study. BMJ. 331(7525): 1113–7. Epub 2005 Oct 19. 25. Flenady V, Koopmans L, Middleton P, Frøen JF, Smith GC, et al. (2011) Major risk factors for stillbirth in high-income countries: a systematic review and meta- analysis. Lancet. 377(9774): 1331–1340.

PLOS ONE | www.plosone.org 7 May 2013 | Volume 8 | Issue 5 | e64869 Gestational Age Patterns of Fetal and Neonatal Mortality in Europe: Results from the Euro-Peristat Project

Ashna D. Mohangoo1*, Simone E. Buitendijk1, Katarzyna Szamotulska2, Jim Chalmers3, Lorentz M. Irgens4, Francisco Bolumar5, Jan G. Nijhuis6, Jennifer Zeitlin7,8, the Euro-Peristat Scientific Committee" 1 Department Child Health, TNO Netherlands Organization for Applied Scientific Research, Leiden, The Netherlands, 2 Department of Epidemiology, National Research Institute of Mother and Child, Warsaw, Poland, 3 Information Services Division, NHS National Services Scotland, Edinburgh, Scotland, 4 Department of Public Health and Primary Health Care, University of Bergen and Medical Birth Registry of Norway, Norwegian Institute of Public Health, Norway, 5 Department of Public Health Sciences, University of Alcala´, Madrid, Spain, 6 Maastricht University Medical Center, GROW School for Oncology and Developmental Biology, Maastricht, The Netherlands, 7 INSERM, UMRS 953, Epidemiological Research Unit on Perinatal and Women’s and Children’s Health, Paris, France, 8 UPMC Univ Paris 06, Paris, France

Abstract

Background: The first European Perinatal Health Report showed wide variability between European countries in fetal (2.6– 9.1%) and neonatal (1.6–5.7%) mortality rates in 2004. We investigated gestational age patterns of fetal and neonatal mortality to improve our understanding of the differences between countries with low and high mortality.

Methodology/Principal Findings: Data on 29 countries/regions participating in the Euro-Peristat project were analyzed. Most European countries had no limits for the registration of live births, but substantial variations in limits for registration of stillbirths before 28 weeks of gestation existed. Country rankings changed markedly after excluding deaths most likely to be affected by registration differences (22–23 weeks for neonatal mortality and 22–27 weeks for fetal mortality). Countries with high fetal mortality $28 weeks had on average higher proportions of fetal deaths at and near term ($37 weeks), while proportions of fetal deaths at earlier gestational ages (28–31 and 32–36 weeks) were higher in low fetal mortality countries. Countries with high neonatal mortality rates $24 weeks, all new member states of the European Union, had high gestational age-specific neonatal mortality rates for all gestational-age subgroups; they also had high fetal mortality, as well as high early and late neonatal mortality. In contrast, other countries with similar levels of neonatal mortality had varying levels of fetal mortality, and among these countries early and late neonatal mortality were negatively correlated.

Conclusions: For valid European comparisons, all countries should register births and deaths from at least 22 weeks of gestation and should be able to distinguish late terminations of pregnancy from stillbirths. After excluding deaths most likely to be influenced by existing registration differences, important variations in both levels and patterns of fetal and neonatal mortality rates were found. These disparities raise questions for future research about the effectiveness of medical policies and care in European countries.

Citation: Mohangoo AD, Buitendijk SE, Szamotulska K, Chalmers J, Irgens LM, et al. (2011) Gestational Age Patterns of Fetal and Neonatal Mortality in Europe: Results from the Euro-Peristat Project. PLoS ONE 6(11): e24727. doi:10.1371/journal.pone.0024727 Editor: Philippa Middleton, The University of Adelaide, Australia Received March 18, 2011; Accepted August 19, 2011; Published November 16, 2011 Copyright: ß 2011 Mohangoo et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Funding: The results from this study are based on data from the Euro-Peristat project, a European project for monitoring and evaluating perinatal outcomes on the European level. The Euro-Peristat project was co-financed by the European Commission (DG-SANCO), grant number 2003131. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. No additional external funding was received for this study. Competing Interests: The authors have declared that no competing interests exist. * E-mail: [email protected] " For a list of Euro-Peristat Scientific Committee members, please see the Acknowledgments.

Introduction stillbirths [4]. These registration differences could be one explanation for the observed variability between countries. In December 2008, the Euro-Peristat network produced the first Preterm birth is an important risk factor for mortality during the European Perinatal Health Report (EPHR) with data from 25 perinatal period and a key to understanding the etiology of both participating EU member states and Norway [1]. This report fetal and neonatal deaths. One of the recommendations of the showed great variations between European countries in fetal and Euro-Peristat project was therefore to collect and present data on neonatal mortality rates in 2004 [1–3]. The highest mortality rates fetal and neonatal mortality by gestational age to allow for were approximately 3.5 times higher than the lowest. Fetal exclusion of gestational age groups when differences in registration mortality rates ranged from 2.6 to 9.1 per 1000 total births, and are most marked and to permit more meaningful analysis of neonatal mortality rates from 1.6 to 5.7 per 1000 live births. The variations between countries by comparing gestational age-specific EPHR also showed that, despite efforts of the World Health mortality rates. Differences in health care policies and practices Organization (WHO) to promote the use of common inclusion may contribute to the variation in observed fetal and neonatal criteria, there were still substantial differences in limits for mortality rates by gestational age, including, for instance, policies registration of live and stillbirths in Europe in 2004, especially related to screening and terminations for congenital anomalies [5–

PLoS ONE | www.plosone.org 1 November 2011 | Volume 6 | Issue 11 | e24727 Patterns of Fetal and Neonatal Mortality in Europe

6]. In countries where terminations are not legal, some babies with was collected afterwards, when it appeared that TOP were not severe congenital anomalies probably die at later gestational ages systematically included as fetal deaths. either during pregnancy or during the neonatal period. In other All countries/regions provided data on neonatal mortality. countries, these pregnancies are more likely to be terminated at Ireland provided data only on early neonatal mortality and earlier gestational ages. Furthermore, prevention strategies for Germany only on early neonatal mortality by gestational age. reducing mortality may differ for very preterm versus at and near Cyprus, the Czech Republic, Greece, Hungary, and Italy provided term births between European countries. For instance, programs no data on neonatal mortality by gestational age. to improve the regionalization of perinatal care can contribute to Spanish data by gestational age came only from the Valencia reduced mortality in the very preterm population [7]. region. Data from France by gestational age was based on several The aim of this study was to analyze gestational age-related sources: a one-week national perinatal survey that was conducted differences in fetal and neonatal mortality between countries in in October 2003, vital registration, and neonatal death certificates. order to assess which part of inter-country variation is due to Data on the gestational age distribution for live births from variations in registration of births and deaths and which part is due England and Wales related to 2005, since these data were not to real differences in health and quality of care. We also sought to available at a national level before. Detailed information on the improve our understanding of differences between low versus high data sources used is presented in Table S1. mortality countries by identifying patterns of mortality by gestational age. The following research questions will be Statistical analysis addressed: (i) How did preterm deaths, and in particular early The annual number of births ranged from 3902 to 774 870. preterm deaths, contribute to differences in the variability of fetal France, Germany, England and Wales, and Italy were among the and neonatal mortality rates? (ii) Were absolute mortality rates countries with more than 500 000 births, while Malta, Luxem- associated with a specific pattern? That is, did countries with low bourg, and Cyprus had less than 10 000. We therefore calculated mortality rates have higher proportions of preterm deaths (which confidence intervals using the binomial distribution to deal with might be considered less preventable) and did high mortality statistical variation of observed mortality rates between countries. countries have higher proportions of at and near term deaths? (iii) Rates were not calculated if there were fewer than 10 births. Rates How did the timing of mortality (fetal/early neonatal/late based on fewer than 10 deaths are noted in the tables. neonatal) differ in countries with high versus low mortality? Low and high mortality countries were defined by choosing the 25th and 75th percentiles respectively as cut-off levels. Differences in Methods the proportions of fetal/neonatal deaths between low versus high mortality countries were tested with the x2-test. We used the non- This study was embedded within the Euro-Peristat project, parametric Spearman test to assess correlations and thus minimize which developed a list of valid and reliable indicators for the effects of outliers. Spearman’s rho (r) was used to interpret the monitoring and evaluating perinatal health in the European strength of correlations. All analyses were performed with SPSS Union [8]. Twenty-five EU member states and Norway partici- version 17.0 for Windows (SPSS Inc, Chicago, IL, USA). pated. Detailed information on the design and methods of the Euro-Peristat project is available elsewhere [4,9–10]. Results Data collection was coordinated in the Netherlands. Data about 29 countries/regions for the year 2004 were analyzed: Austria (AT), Registration of live and stillbirths Belgium (BE): regions Brussels (BE.BR) and Flanders (BE.FL), Table 1 shows that most European countries had no limits for Cyprus (CY), Czech Republic (CZ), Denmark (DK), Estonia (EE), registration of live births in 2004, but that the legal limits for Finland (FI), France (FR), Germany (DE), Greece (GR), Hungary registration of stillbirths varied substantially. In some countries (HU), Ireland (IE), Italy (IT), Latvia (LV), Lithuania (LT), stillbirths weighing less than 500 grams were not registered, while Luxembourg (LU), Malta (MT), the Netherlands (NL), Norway other countries had a legal gestational age limit of 24 or even 28 (NO), Poland (PL), Portugal (PT), Slovenia (SI), Slovakia (SK), weeks. Because of this Hungary and Sweden could not adhere to the Spain or the Valencia region of Spain (ES), Sweden (SE), and the Euro-Peristat definition of 22 completed weeks of gestation for United Kingdom (UK): England and Wales combined (UK.EW), stillbirths in 2004. Voluntary notification of late fetal deaths at 22– Northern Ireland (UK.NI), and Scotland (UK.SC). 23 weeks existed in the United Kingdom and Portugal and therefore Portugal, Northern Ireland, and Scotland were able to include these Euro-Peristat definitions deaths in their data. Italy and Luxembourg had a legal limit for Within Euro-Peristat the fetal mortality rate is defined as the registration of 180 days of pregnancy, but late fetal deaths starting at number of deaths before or during birth at or after 22 completed 22 weeks were available in the register of spontaneous abortions and weeks of gestation in a given year per 1000 live and stillbirths in were included. Table 1 also indicates that most countries did not the same year. The neonatal mortality rate is defined as the include TOP as fetal deaths. Exceptions were France, the Nether- number of deaths during the neonatal period (day 0 to 27) at or lands and Scotland. Elsewhere TOP were registered separately and after 22 completed weeks of gestation in a given year expressed per not included in national mortality statistics. 1000 live births in the same year. Early preterm deaths were defined as deaths that occurred at 22–27 weeks of gestation, and at Variation in fetal and neonatal mortality rates and near term deaths as deaths at 37 weeks and above. A large range was observed for overall fetal (2.6–9.1%) and neonatal (1.6–5.7%) mortality rates in the 28 participating Availability of fetal and neonatal mortality countries/regions, as shown in Table 2. In these countries/ If participating countries/regions were unable to provide regions, 25 360 fetal deaths and 4 733 268 births, and 14 212 numbers on fetal and neonatal mortality according to the Euro- neonatal deaths and 4 713 200 live births were registered. The Peristat definition, the local definition was used. Cyprus provided proportion of fetal deaths represented by TOP varied between no data on fetal mortality, and the data from Greece and Italy countries, as measured by supplemental data provided by a few were for 2003. Information on fetal deaths with and without TOP countries (data not shown in table). If TOP were included as fetal

PLoS ONE | www.plosone.org 2 November 2011 | Volume 6 | Issue 11 | e24727 Patterns of Fetal and Neonatal Mortality in Europe

Table 1. Criteria for registration of live births and stillbirths and inclusion of terminations of pregnancy in Europe in 2004.

TOP included in a separate data Live births Stillbirths TOP included system

Austria No limit $500 grams No No BE: Brussels No limit $22 weeks or $500 grams No No BE: Flanders No limit $500 grams No No Cyprus No limit No data available No data available Czech Republic $500 grams or any birth weight $22 weeks No Yes surviving the first 24 hours Denmark No limit $22 weeks No Yes Estonia No limit $22 weeks or $500 grams No Yes Finland No limit $22 weeks or $500 grams No Yes France $22 weeks or $500 grams $22 weeks or $500 grams Yes No Germany No limit $500 grams No Yes Greece No limit legal limit of $28 weeks No No Hungary No limit $24 weeks or $500 grams No Yes Ireland No limit $24 weeks or $500 grams TOP is not legal and not performed Italy No limit 180 days No Yes Latvia No limit $22 weeks No Yes Lithuania $22 weeks $22 weeks No Yes Luxembourg No limit 180 days No No Malta No limit $22 weeks or $500 grams TOP is illegal and not performed The Netherlands $22 weeks or $500 grams $24 weeks for civil registration Yes Yes $16 weeks for perinatal registry Norway $12 weeks $12 weeks No Yes Poland $500 grams $500 grams No Yes Portugal No limit $24 weeks No No Slovakia No limit $22 weeks or $500 grams No Yes Slovenia No limit $500 grams No Yes Spain No limit $26 weeks (national) No Yes $22 weeks (region Valencia) Sweden No limit $28 weeks No Yes UK: England and Wales No limit legal limit of $24 weeks voluntary Yes Yes notification at 22–23 weeks UK: Northern Ireland No limit legal limit of $24 weeks voluntary TOP is not legal* notification at 22–23 weeks UK: Scotland No limit legal limit of $24 weeks voluntary Yes Yes notification at 22–23 weeks

*The legislation which legalised abortion in the rest of the United Kingdom does not cover Northern Ireland, but TOP are occasionally done there under case law. doi:10.1371/journal.pone.0024727.t001 deaths, the fetal mortality rate would have increased from 5.7% to many countries, deaths before 28 weeks of gestation accounted for 5.9% in England and Wales (3.3% were TOP), from 3.2% to a substantial proportion of all deaths. Countries with the highest 3.7% in Finland (16% were TOP), and from 5.4% to 6.9% in overall fetal mortality rates did not necessarily have the highest Italy (28% were TOP). After excluding TOP, the fetal mortality fetal mortality rates at or after 28 weeks of gestation, and not all rate in Scotland would have declined from 6.7% to 6.6% (2.5% countries with the highest neonatal mortality rates had the highest were TOP), and the Netherlands estimated that their fetal neonatal mortality rates at or after 28 weeks of gestation. The fetal mortality rate would have declined from 7.0% to 6.8% after mortality rate declined dramatically for France when fetal deaths excluding TOP (2.9% were TOP). at 22–27 weeks were excluded and removing neonatal deaths at 22–23 weeks led to a large decline in neonatal mortality rates in Exclusion of early preterm deaths the Netherlands, Northern Ireland, and England and Wales. Figure 1 illustrates fetal and neonatal mortality rates by gestational-age subgroups and ranks countries by their mortality Adjusted proportions of fetal and neonatal deaths rate at 28 weeks and over. Excluding early preterm deaths reduced Figure 2 presents the percentage of fetal and neonatal deaths in the range of fetal mortality rates quite substantially, while a each gestational age subgroup after excluding fetal deaths at 22–27 moderate reduction was observed for neonatal mortality rates. For weeks and neonatal deaths at 22–23 weeks. The percentage of fetal

PLoS ONE | www.plosone.org 3 November 2011 | Volume 6 | Issue 11 | e24727 Patterns of Fetal and Neonatal Mortality in Europe

Table 2. Fetal and neonatal mortality rates in Europe in 2004.

Neonatal Mortality Number of Number of Fetal Mortality Rates Number of live Number of Rates per 1000 live Country/region total births fetal deaths per 1000 total births births neonatal deaths births

Austria 79 229 295 3.7 [3.3–4.1] 78 934 215 2.7 [2.4–3.1] BE: Brussels 16 288 88 5.4 [4.3–6.5] 16 200 51 3.4 [2.3–4.0] BE: Flanders 60 921 249 4.1 [3.6–4.6] 60 672 146 2.4 [2.0–2.8] Cyprus NA NA NA 8 309 13 1.6 [0.7–2.4] Czech Republic 98 051 387 3.9 [3.6–4.3] 97 671 224 2.3 [2.0–2.6] Denmark 64 853 332 5.1 [4.6–5.7] 64 521 230 3.6 [3.1–4.0] Estonia 14 053 63 4.5 [3.4–5.6] 13 990 59 4.2 [3.1–5.3] Finland 57 759 190 3.3 [2.8–3.8] 57 569 141 2.4 [2.0–2.9] France 774 870 7 054 9.1 [8.9–9.3] 767 816 1968 2.6 [2.5–2.7] Germany 648 860 2 261 3.5 [3.3–3.6] 705 622 1892 2.7 [2.6–2.8] Greece 104 858 503 4.8 [4.4–5.2] 104 355 282 2.7 [2.4–3.0] Hungary 95 594 476 5.0 [4.5–5.4] 95 137 423 4.4 [4.0–4.9] Ireland 62 400 334 5.4 [4.8–5.9] 62 066 NA NA Italy 542 003 2 937 5.4 [5.2–5.6] 539 066 1526 2.8 [2.7–3.0] Latvia 20 492 137 6.7 [5.6–7.8] 20 355 116 5.7 [4.7–6.7] Lithuania 29 633 153 5.2 [4.3–6.0] 29 480 136 4.6 [3.8–5.4] Luxembourg 5 486 17 3.2 [1.6–4.6] 5 469 11 2.0 [0.8–3.2] Malta 3 902 15 3.8 [1.9–5.8] 3 887 17 4.4 [2.3–6.4] The Netherlands 182 279 1 273 7.0 [6.6–7.4] 181 006 631 3.5 [3.2–3.8] Norway 57 368 257 4.5 [3.9–5.0] 57 111 118 2.1 [1.7–2.4] Poland 358 440 1 743 4.9 [4.6–5.1] 356 697 1731 4.9 [4.6–5.1] Portugal 109 778 422 3.8 [3.5–4.2] 109 356 280 2.6 [2.3–2.9] Slovak Republic 52 522 134 2.6 [2.1–3.0] 52 388 134 2.6 [2.1–3.0] Slovenia 17 946 100 5.6 [4.5–6.7] 17 846 47 2.6 [1.9–3.4] Spain 456 029 1 438 3.2 [3.0–3.3] 454 591 1199 2.6 [2.5–2.8] Sweden 100 474 316 3.1 [2.8–3.5] 100 158 210 2.1 [1.8–2.4] UK: England and Wales 643 407 3 686 5.7 [5.5–5.9] 639 721 2185 3.4 [3.3–3.6] UK: Northern Ireland 22 504 142 6.3 [5.3–7.3] 22 362 66 3.0 [2.2–3.7] UK: Scotland 53 269 358 6.7 [6.0–7.4] 52 911 161 3.0 [2.6–3.5]

Cyprus provided no data on fetal death. Ireland only provided data on early neonatal death. Data for countries that did not adhere to the Euro-Peristat definition are presented in italics. High mortality rates (.75th quartile) are presented in bold. doi:10.1371/journal.pone.0024727.t002 deaths in each gestational age subgroup differed significantly Substantial variation existed within all gestational age subgroups, between low and high fetal mortality countries (p,0.001). On even when rates based on small denominators were excluded. average, low fetal mortality countries had higher percentages of Countries with high neonatal mortality rates at or after 24 weeks of their fetal deaths at earlier gestational ages (at 28–31 weeks 24.5% gestation (Latvia, Poland, Malta, Lithuania, and Estonia) often had vs. 22.9%, at 32–36 weeks 36.3% vs. 32.9%), while high fetal the highest gestational age-specific neonatal mortality rates for all mortality countries had higher percentages at and near term (44.2% gestational age subgroups, except at 22–23 weeks. Some countries vs. 39.2%). In contrast, the percentage of neonatal deaths in each stood out in some subgroups. The Netherlands, for instance, had gestational age subgroup did not differ significantly between low high rates at 22–23 weeks and at 24–27 weeks, while Denmark and high neonatal mortality countries (p = 0.112): at 24–27 weeks had high rates at 22–23 weeks and at 37+ weeks. (30.0% vs. 31.7%), at 28–31 weeks (17.4% vs. 19.5%), at 32–36 weeks (18.5% vs. 19.6%), and at 37+ weeks (34.1% vs. 29.2%). Correlation between fetal and neonatal mortality rates Indeed, on average, high neonatal mortality countries had lower Although fetal mortality rates at or after 28 weeks and neonatal percentages of neonatal deaths at and near term. mortality rates at or after 24 weeks were significantly correlated (r = 0.646; p = 0.001), different patterns were observed for low Gestational age-specific neonatal mortality rates versus high neonatal mortality countries. Figure 3 shows that high Table 3 provides data on gestational age-specific neonatal neonatal mortality countries had high fetal mortality rates, but mortality for those countries that could provide both numerator countries with low and moderate neonatal mortality rates had and denominator data for the gestational age distribution. varying levels of fetal mortality. Finland, Czech Republic,

PLoS ONE | www.plosone.org 4 November 2011 | Volume 6 | Issue 11 | e24727 Patterns of Fetal and Neonatal Mortality in Europe

Figure 1. Impact of different inclusion criteria on fetal and neonatal mortality rates. Countries were sorted by mortality rate at or after 28 weeks of gestation with low mortality countries listed first. doi:10.1371/journal.pone.0024727.g001

Luxembourg, Norway, Spain, and Sweden, for example, all had to the variation in fetal and neonatal mortality rates in Europe neonatal mortality rates of 2.0%, but their fetal mortality ranged in 2004. But even after these early preterm deaths were from 2.0% through 3.5%. The correlation between fetal and excluded, fetal and neonatal mortality rates varied notably and neonatal mortality increased with increasing gestational age and in all gestational age subgroups, including those at and near was closest when all preterm deaths were excluded (r = 0.758; term. In addition, patterns of mortality differed for the p,0.001), see Figure S1. gestational age at which highest mortality was observed and for the association between fetal and neonatal mortality Correlation between early and late neonatal mortality rates. rates Our study has several limitations linked to measurement and Early and late neonatal mortality were related in different ways data availability, despite efforts within Euro-Peristat to ensure in countries with high (r = 0.261; p = 0.618) versus low and comparability in definitions and sources. The first limitation is moderate (r = 20.302; p = 0.184) neonatal mortality, as Figure 4 related to our ability to assess the completeness of registration of shows. Countries with high neonatal mortality had high rates of early preterm births, especially stillbirths [11]. Although we were both early and late neonatal mortality, while different patterns able to describe registration rules, our lack of knowledge about the were observed in other countries: some countries had high early extent to which these were applied uniformly in all countries limits neonatal mortality, but low late neonatal mortality (e.g. the our ability to assess real fetal mortality at 22–27 weeks and real Netherlands and Denmark) and several other countries had low neonatal mortality at 22–23 weeks. A related issue is the early neonatal mortality and high late neonatal mortality. Early registration of TOP, which is managed very differently from one neonatal mortality was related to total neonatal mortality in all country to another and terminations are not always included as countries (r = 0.915; p,0.001), but late neonatal mortality was fetal deaths. TOP varied from 3 to 28% of total fetal deaths in related to total neonatal mortality only in countries with high those countries where this proportion could be computed. In neonatal mortality (r = 0.812; p = 0.05), see Figure S2. France an even larger proportion of early fetal deaths are estimated to be TOP [12–13]. To our knowledge, this limitation Discussion affects all stillbirth data routinely reported to international agencies. This issue has not previously been highlighted, even Our analysis shows that early preterm deaths, most strongly though the influence of TOP on fetal mortality rates has been influenced by registration differences, contributed substantially discussed in previous studies [14–15].

PLoS ONE | www.plosone.org 5 November 2011 | Volume 6 | Issue 11 | e24727 Patterns of Fetal and Neonatal Mortality in Europe

PLoS ONE | www.plosone.org 6 November 2011 | Volume 6 | Issue 11 | e24727 Patterns of Fetal and Neonatal Mortality in Europe

Figure 2. Fetal deaths at or after 28 weeks of gestation (A) and neonatal deaths at or after 24 weeks (B) by gestational age subgroups. Countries were sorted by fetal (A) and neonatal (B) mortality rates with low mortality countries listed first. Fetal mortality rate at or after 28 weeks of gestation was calculated as follows: [(number of fetal deaths $28 weeks)/(number of total births $28 weeks)]61000. France, Latvia, Scotland, Ireland and The Netherlands had rates in the top quartile (.75th). * Percentages of fetal deaths were based on fewer than 10 events for Luxembourg and Malta, and for Estonia (at 28–31 and 32–36 weeks). Neonatal mortality rate at or after 24 weeks of gestation was calculated as follows: [(number of neonatal deaths $24 weeks)/(number of live births $24 weeks)]61000. Latvia, Poland, Malta, Lithuania, and Estonia had mortality rates in the top quartile (.75th). * Percentages of neonatal death were based on less than 10 events for Luxembourg and Malta, for Brussels, Estonia, and Slovenia (at 28–31 weeks and at 32–36 weeks), and for Northern Ireland (at 28–31 weeks). doi:10.1371/journal.pone.0024727.g002

The different sources of data used by each country may also population-based data. We addressed the question of small sample affect comparability [4,16]. Some countries provided mortality sizes by calling attention to rates based on fewer than 10 deaths, data from civil or perinatal registries or both, by linkage. These not reporting those based on fewer than 10 births, and presenting differences in methods may lead to differences in coverage [4]. confidence intervals. The number of annual births in each country also varied greatly A final limitation is related to the comparability of reported throughout Europe, from 3902 to 774 870, and random variability gestational ages. Euro-Peristat requested data in completed weeks from year to year is high when rates are based on small numbers. of gestation based on the best obstetrical estimate available, but we In Malta, for example, the relatively high gestational age-specific were unable to assess or evaluate differences in the measurement neonatal mortality rate at 28–31 weeks of 238% (5 out of 21 of gestational age. Preterm birth rates may differ depending to babies died) in 2004 fell drastically to 59% in the two subsequent whether they are measured by last menstrual period or by years (1 out of 17 babies died in both 2005 and 2006) [17]. In ultrasound [18–19]. Use of ultrasound measurements, by shifting addition, not all countries were able to provide national data; the entire gestational age distribution to the left, can increase the instead, some used either representative samples or regional preterm birth rate [18], but it can also decrease the rate by

Table 3. Gestational age-specific neonatal mortality rates per 1000 live births.

Country/region Gestational age in weeks 22–23 24–27 28–31 32–36 $37

Luxembourg – – * 76.9 [0.0–222] * 9.7 [0.0–20.7] * 0.6 [0.0–1.3] Czech Republic 546 [337–754] 218 [170–266] 52.1 [36.2–68.0] 5.9 [3.9–7.9] 0.5 [0.4–0.7] Norway 556 [326–785] 185 [126–243] 46.6 [26.1–67.0] 4.3 [2.1–6.5] 0.8 [0.6–1.1] ES: Valencia – 302 [215–389] 67.4 [41.4–93.5] 3.5 [1.7–5.4] 0.5 [0.3–0.7] Sweden 485 [314–655] 167 [121–212] 33.4 [19.3–47.4] 8.9 [6.4–11.4] 0.9 [0.7–1.1] Finland 867 [745–988] 293 [216–371] 52.6 [29.0–76.3] 6.0 [3.0–8.9] 0.7 [0.5–1.0] BE: Flanders 1000 [1000–1000] 311 [237–385] 51.5 [29.5–73.5] 7.2 [4.7–9.8] 0.6 [0.4–0.8] Austria 867 [767–966] 230 [182–279] 37.4 [24.0–50.7] 4.1 [2.7–5.5] 0.7 [0.5–0.9] Slovenia – 308 [182–433] * 36.5 [5.1–67.9] * 7.6 [2.4–12.9] 0.7 [0.3–1.1] Portugal 338 [285–391] 54.9 [38.2–71.7] 6.7 [4.7–8.8] 0.7 [0.6–0.9] France – – – – 0.8 [0.8–0.9] UK: Northern Ireland – 244 [151–337] * 30.3 [4.1–56.5] 9.0 [3.7–14.3] 0.8 [0.4–1.2] Slovakia 600 [352–848] 281 [203–359] 86.0 [56.6–115.4] 11.8 [7.8–15.8] 0.5 [0.3–0.7] UK: England and 903 [880–926] 237 [220–254] 36.6 [31.7–41.4] 6.1 [5.3–6.9] 0.9 [0.9–1.0] Wales{ UK: Scotland – 301 [234–367] 47.3 [27.6–67.0] 4.7 [2.4–7.0] 0.8 [0.6–1.1] The Netherlands 976 [950–1000] 325 [282–368] 54.5 [42.2–66.7] 7.5 [5.9–9.1] 1.1 [1.0–1.3] Denmark 947 [847–1000] 289 [220–358] 38.2 [21.3–55.0] 8.2 [5.3–11.1] 1.6 [1.3–2.0] BE: Brussels – 320 [191–449] * 84.9 [31.8–138] * 6.6 [1.3–11.8] 1.3 [0.7–1.9] Estonia – 321 [195–446] * 47.1 [2.0–92.1] * 8.8 [1.8–15.8] 2.1 [1.3–2.8] Lithuania 786 [571–1000] 488 [378–597] 90.9 [49.7–132] 13.2 [6.9–19.4] 1.9 [1.4–2.4] Malta – – * 238 [55.9–420] * 15.9 [0.4–31.3] * 1.4 [0.2–2.6] Poland 875 [822–928] 457 [428–486] 125 [112–137] 16.2 [14.5–17.9] 1.2 [1.1–1.4] Latvia – 477 [356–598] 82.9 [42.7–123] 15.3 [7.3–23.2] 2.7 [2.0–3.4]

Cyprus, Germany, Greece, Hungary, Ireland, and Italy had no data on neonatal death by gestational age. { Data from 2005. Countries were sorted by neonatal mortality rate at or after 24 weeks of gestation with low mortality countries listed first. High mortality rates are presented in bold (.75th quartile). Rates based on fewer than 10 deaths were denoted with *. Rates were not computed for cells with fewer than 10 births and were denoted with –. For France the number of term live births was estimated from the national perinatal survey and totals from the vital statistics data. doi:10.1371/journal.pone.0024727.t003

PLoS ONE | www.plosone.org 7 November 2011 | Volume 6 | Issue 11 | e24727 Patterns of Fetal and Neonatal Mortality in Europe

Figure 3. Correlation between fetal and neonatal mortality rates, after exclusion of deaths most likely influenced by registration differences. High mortality countries are presented in bold. Correlation for fetal and neonatal mortality: r = 0.646 (p = 0.001). doi:10.1371/journal.pone.0024727.g003

Figure 4. Correlation between early and late neonatal mortality rates. High neonatal mortality countries are presented in bold. r = 0.261 (p = 0.618) in high neonatal mortality countries versus r = 20.302 (p = 0.184) in other countries. doi:10.1371/journal.pone.0024727.g004

PLoS ONE | www.plosone.org 8 November 2011 | Volume 6 | Issue 11 | e24727 Patterns of Fetal and Neonatal Mortality in Europe reducing errors in gestational age estimates [20]. In most problems of comparability raised by TOP notification. For valid European countries, however, dating ultrasounds are now part European comparisons of gestation-specific fetal and neonatal of standard care during pregnancy, and most women have their mortality, all member states should aim to register fetal deaths first prenatal visit in the first trimester [1]. from 22 completed weeks of gestation, regardless of birth weight. The Euro-Peristat project chose to collect mortality data using a In July 2008, Sweden changed its limit for registration of stillbirths cut-off point of 22 completed weeks of gestation without a birth from 28 to 22 completed weeks of gestation. This change will make weight limit, although data were also collected by birth weight to it possible to compare their stillbirth rates with other European permit calculation of rates with a lower limit of 1000 grams as countries at earlier gestational ages in future studies. recommended by WHO for international comparisons. While the Use of these exclusion criteria had a substantial impact on the WHO definition of the perinatal period is based on gestational age ranking of countries. Not all countries with the highest or lowest (starting at 22 weeks of gestation), recommendations for data mortality rates also had the highest or lowest fetal mortality rates at collection are based primarily on birth weight (death of a fetus that or after 28 weeks or the highest or lowest neonatal mortality rates has reached a birth weight of 500 grams or if the birth weight is at or after 24 weeks. It is thus very important for international unavailable, a gestational age of 22 completed weeks or a crown- agencies to collect gestational age to allow like-with-like compar- to-heel length of 24 cm is used) [21]. These international isons. In this light, EUROSTAT’s updated directives, which recommendations are understandable because in many countries makes provision of fetal and neonatal mortality by birth weight valid data on gestational age simply do not exist, although it is not voluntary for EU member states and does not even request clear that stillbirths are systematically weighed [22]. In Europe, voluntary collection by gestational age, is a matter for concern however, legislation for registering stillbirths is based largely on [32]. gestational age. Regulations governing TOP are also specified with After exclusion of the early preterm deaths most likely to be respect to the length of gestation [13]. influenced by registration differences, the levels and patterns of It is important to note that gestational age-specific mortality mortality still varied significantly between countries. The clearest data are not produced routinely by any international or European pattern observed was among countries with highest neonatal agency (including EUROSTAT, OECD or WHO) and have not, mortality rates, all new member states of the European Union to our knowledge, ever been published in this way from routine (Latvia, Poland, Malta, Estonia, Lithuania, and Hungary). These data. Our analyses require the collection of fetal and neonatal countries had high fetal mortality and high early and high late (both early and late) deaths by individual week of gestation and neonatal mortality rates. This finding suggests that some of the these are not routinely provided, even in many national causes may be related to overall standards of living and resources publications. A further strength of our analysis is the number of available to the health care system. For Poland and Malta, the countries that contributed. restricted availability (Poland) and illegality of TOP (Malta) may From an analytic perspective, gestational-age based analyses are also contribute to higher rates. important because they distinguish premature birth from fetal In contrast, we observed different levels of fetal mortality in growth restriction. Recent European cohorts of very preterm other countries with the same overall level of neonatal mortality. infants are based on gestational age because of its superior Medical advances, such as antenatal steroids and surfactant for prognostic value [23] and because information on gestational age, very preterm babies, have been very successful in decreasing and not on birth weight, is available to obstetricians when making neonatal mortality, perhaps contributing to more convergent decisions during pregnancy and delivery [24–27]. Gestational age trends in neonatal mortality in countries with similar access to comparisons across Europe avoid biases related to population these technologies. In contrast, preventing stillbirths may be more differences in birth weight; European comparisons have found that complex. Stillbirths are strongly related to maternal social optimum birth weight, defined as the birth weight at which characteristics, high body mass index and smoking [33–34]. The mortality is lowest, varies between European countries [28]. prevalence of these risk factors and health system programs to Our analyses showed that it was necessary to exclude stillbirths reduce their impact may differ across countries. Factors suggesting at 22–27 weeks and neonatal deaths at 22–23 weeks from sub-optimal care are associated with a substantial proportion of European comparisons to minimize the effect of differences in stillbirths, especially for intrapartum deaths [35]. registration requirements and TOP legislation on mortality rates. We also found that some countries had higher mortality in some As survival is rare for babies at 22–23 weeks, including this gestational age subgroups. Denmark, for example, had both high gestational age group adds nearly as many deaths as births and proportions of fetal and neonatal deaths at 37+ weeks, and a high gives this subgroup a large weight in mortality statistics, although gestational age-specific neonatal mortality rate at 37+ weeks, but they only represent about 1 in 1000 live births. These infants make was not defined as a high mortality country when comparing rates up a high proportion of fetal and neonatal deaths in some at earlier gestational ages. France, on the other hand, had the countries. Although registration limits primarily affect fetal deaths, highest fetal mortality rate at or after 28 weeks, but did not have a fetal death registration is known to affect the completeness of live high proportion of fetal deaths at and near term. birth registration, especially, for example, for those weighing less Differences in policies and practices may explain these varying than 500 grams. The limit of 500 grams primarily concerns infants patterns of mortality. For instance, policies related to screening at 22–23 weeks, so excluding these infants also resolves the and terminations for congenital anomalies differ in Europe and problems of comparability presented by birth weight limits. By 24 can have a large impact on both fetal and neonatal mortality rates weeks of gestation, most babies have a birth weight above 500 [5–6]. TOP could be performed legally around 2004 in most grams [29]. Some countries had substantially lower gestational European countries, although the maximum gestational age limit age-specific neonatal mortality at 22–23 weeks than reported in varied and the notification procedures differed. Exceptions were specific studies of this population [30–31] suggesting that in some Ireland, Northern Ireland and Malta, where TOP are not legal countries many immediate neonatal deaths are simply not and cannot be performed. It is possible that some pregnancies that registered as live births in birth registers. For fetal mortality, a were found to have lethal congenital anomalies from these higher cut-off point was necessary to include Sweden, which only countries were terminated elsewhere or that fewer of these registered fetal deaths starting at 28 weeks, as well as to deal with pregnancies were terminated; in the latter cases, these babies,

PLoS ONE | www.plosone.org 9 November 2011 | Volume 6 | Issue 11 | e24727 Patterns of Fetal and Neonatal Mortality in Europe who died from the anomaly before, during or after birth, were references for neonatal and fetal mortality by gestational age included in their mortality statistics. Policies related to TOP may makes it possible for them to assess their specific weaknesses and explain the relatively high proportion of fetal deaths at and near generate ideas about how to improve outcomes. These data also term in the Netherlands, which, unlike most other European raise important questions for future research about the tradeoffs countries in 2004, had no system for systematic early detection of between fetal and neonatal mortality in many countries and the congenital anomalies through prenatal screening. On the other reasons for differing gestational age-specific patterns in neonatal hand, countries that systematically screened at earlier gestational mortality. ages and terminated more before the registration limit necessarily have lower mortality rates, both fetal and neonatal [13–14]. In Supporting Information contrast, countries that practice terminations at or after 22 weeks may end up with high stillbirth rates explained primarily by Table S1 Data sources used for the Euro-Peristat terminations. France is one example [13]. Late terminations for project data. fetal anomalies are rare in Europe, and prevalence rates vary (DOC) between European countries [13]. Ideally, it should be possible to Figure S1 Correlation between gestation-specific fetal remove TOP from fetal death statistics and to calculate rates with and neonatal mortality rates. Correlation for fetal and and without TOP. neonatal mortality $22 weeks: r = 0.502 (p = 0.010). Correlation Another area where European countries differ is in policies and for fetal and neonatal mortality $28 weeks: r = 0.612 (p = 0.002). attitudes to withdrawing and withholding care for preterm infants Correlation for fetal and neonatal mortality $37 weeks: r = 0.758 at the limit of viability. This affects the types of care babies receive, (p,0.001). their survival probabilities and the timing of death [36–37]. (TIF) Countries that are more likely to withhold care will have higher mortality rates at early gestational ages; for example the Nether- Figure S2 Correlation of early and late neonatal lands had high gestational age-specific neonatal mortality rates at mortality with total neonatal mortality. High neonatal 22–23 and at 24–27 weeks, because its active interventions for mortality countries are presented in bold. Correlation for early these early preterm live births were much more limited than in and total neonatal mortality: r = 0.915 (p,0.001). Correlation for other European countries [36–37]. These policies may also affect late and total neonatal mortality: r = 0.812 (p = 0.05) in high the proportion of infants that are live born at early gestational ages neonatal mortality countries versus r = 20.210 (p = 0.362) in other after medically indicated cesareans [38]. Cesarean delivery rates countries. for infants born between 24 and 25 weeks, excluding cesareans for (TIF) maternal indications, varied from 0–78% between European regions in 2003, for instance [39]. Without active intervention, Acknowledgments neonatal deaths also occur earlier. In the Netherlands, for The Euro-Peristat Scientific Committee: Christian Vutuc (Austria), Sophie example, high early neonatal mortality rates but low late neonatal Alexander (Belgium), Pavlos Pavlou (Cyprus), Petr Velebil (Czech mortality rates could reflect the lack of active intervention before Republic), Jens Langhoff Roos (Denmark), Luule Sakkeus (Estonia), Mika 26 weeks of gestation [36]. Several other countries had low early Gissler (Finland), Beátrice Blondel (France), Nicholas Lack (Germany), Aris neonatal mortality rates, but high late neonatal mortality rates, Antlaklis (Greece), Istvań Berbik (Hungary), Sheelagh Bonham (Ireland), which may indicate that babies were living longer and dying in the Marina Cuttini (Italy), Jautrite Karaskevica (Latvia), Jone Jaselioniene late neonatal period. (Lithuania), Yolande Wagener (Luxembourg), Miriam Gatt (Malta), Jan Finally, high mortality at and near term may reflect policies Nijhuis (The Netherlands), Lorentz Irgens (Norway), Katarzyna Szamo- tulska (Poland), Henrique Barros (Portugal), Ma´ria Chmelova´ (Slovakia), related to the care of term pregnancies, including policies to Zˇ iva Novak-Antolic (Slovenia), Francisco Bolu´mar (Spain), Gunilla Lind- induce delivery for post-term infants, which may differ substan- mark (Sweden), Alison Macfarlane (United Kingdom). tially [40] and the organization of maternity services for low-risk The authors acknowledge the following contributors to the European pregnancies which also varies greatly within Europe [41]. Having Perinatal Health Report: Austria Christian Vutuc, Abteilung fu¨r Epide- data on causes of death and on timing of death (intrapartum or miologie Zentrum fu¨r Public Health der Med. Univ. Wien; Jeannette antepartum) would add to Euro-Peristat’s capacity to explain Klimont, Statistics Austria; Belgium Sophie Alexander, Wei-Hong Zhang, differences in levels and patterns of gestational age-specific Universite´ Libre de Bruxelles, School of Public Health, Reproductive mortality and this will be a goal for future phases. Health Unit; Guy Martens, SPE (Study Center for Perinatal Epidemiol- ogy), Edwige Haelterman, Myriam De Spiegelaere, Brussels Health and Social Observatory; Cyprus Pavlos Pavlou, Maria Athanasiadou, Ministry of Conclusions Health, Health Monitoring Unit; Andreas Hadjidemetriou, Christina Registration differences contributed notably to the variability in Karaoli, Neonatal Intensive Care Unit, Makarios III Hospital; Czech fetal and neonatal mortality rates between countries. All European Republic Petr Velebil, Institute for the Care of Mother and Child, Vit countries should use common inclusion criteria for the registration Unzeitig, Department of Obstetrics and Gynecology, Masaryk University Brno; Denmark Jens Langhoff Roos, Obstetrics Clinic, Rigshospitalet, of live and stillbirths. To allow a common analysis of gestation- Copenhagen University; Steen Rasmussen, Sundhedsstyrelsen National specific mortality, it is important to have data starting from at least Board of Health; Estonia Luule Sakkeus, Kati Karelson, Mare Ruuge, 22 completed weeks of gestation. To comply with WHO National Institute for Health Development, Department of Health recommendations, this would require countries to collect data Statistics; Finland Mika Gissler, National Research and Development using both gestational age (22 weeks) and birth weight (500 grams) Centre for Welfare and Health (STAKES); Anneli Pouta National Public limits, as is already done in several European countries. Countries Health Institute (KTL), Department of Child and Adolescent Health; should also be able to calculate fetal mortality rates with and France Beátrice Blondel, Marie-He´le`ne Bouvier-Colle, Ge´rard Breárt, without late TOP. Nonetheless, differences in registration criteria Jennifer Zeitlin, Meagan Zimbeck, INSERM U953; Christine Cans, SCPE Service d’Information et d’Informatique Me´dicale (SIIM); Germany do not explain the variability in mortality rates between European Nicholas Lack, Bavarian Working Group for Quality Assurance, Klaus countries. Routine reporting of fetal and neonatal deaths by Doebler, Federal Quality Assurance Office BQS; Greece Aris Antlaklis, gestational age improves the usefulness of these data for Peter Drakakis, Athens University, Department of Obstetrics and surveillance and policy. Providing countries with international Gynecology, Division of Maternal and Fetal Medicine; Hungary Istvań

PLoS ONE | www.plosone.org 10 November 2011 | Volume 6 | Issue 11 | e24727 Patterns of Fetal and Neonatal Mortality in Europe

Berbik, Vaszary Kolos Teaching Hospital, Department of Obstetrics and Jarmila Hajnaliova, National Health Information Center; Slovenia Zˇ iva Gynecology; Istvań Szabo´, Department of Obstetric and Gynaecology, Novak-Antolicˇ, Ivan Verdenik, University Medical Centre, Perinatology Medical Faculty, Scientific University of Pećs; Ireland Sheelagh Bonham, Unit, Polonca Truden-Dobrin, Center for Health and Health Care Jacqueline O’Reilly, Economic and Social Research Institute (ESRI); Italy Research, Institute of Public Health of the Republic of Slovenia; Spain Marina Cuttini, Pediatric Hospital of Baby Jesus, Unit of Epidemiology; Francisco Bolumar, Universidad de Alcala Facultad de Medecina; Ramon Sabrina Prati, Cinzia Castagnaro, Silvia Bruzzone, Marzia Loghi, Istituto Prats, Departament de Salut Direccio General Salut Publica; Carmen Nazionale di Statistica, ISTAT; Latvia Jautrite Karaskevica, Irisa Zile, Barona, Perinatal Health Unit Public Health Board, Isabel Rıó, CIBER Health Statistics and Medical Technologies State Agency; Ilze Kreicberga, Epidemiologıá y Salud Pu´blica (CIBERESP); Sweden Gunilla Lindmark, Riga Maternity Hospital; Lithuania Aldona Gaizauskiene, Kotryna IMCH, Akademiska sjukhuset; Milla Bennis, National Board of Health and Paulauskiene, Lithuanian Health Information Centre; Luxembourg Yolande Welfare; United Kingdom, Alison Macfarlane, Nick Drey, Department of Wagener, Ministe`re de la Sante´, Direction de la Sante´, Division de la Midwifery, City University London; Angela Bell, Health Promotion Me´decine Pre´ventive et Sociale; Malta Miriam Gatt, Kathleen England, Agency for Northern Ireland CEMACH; Jim Chalmers, Etta Shanks, Department of Health Information and Research; Raymond Galea, Information Services Division, NHS National Services Scotland; Di Department of Obstetrics and Gynecology, University of Malta; The Goodwin, Kath Moser, Office for National Statistics; Gwyneth Thomas, Netherlands Sabine Anthony, Simone Buitendijk, Ashna Mohangoo, Ab Health Statistics and Analysis Unit, Statistical Directorate, Welsh Assembly Rijpstra, TNO Quality of Life, Deparment Prevention and Care, Section Government. Reproduction and Perinatology, Leiden; Jan Nijhuis, Maastricht Univer- sity Medical Center, Department of Obstetrics and Gynecology; Chantal Hukkelhoven, The Netherlands Perinatal Registry; Norway Lorentz Irgens, Author Contributions Kari Klungsoyr Melve, University of Bergen, Medical Birth Registry of Conceived and designed the experiments: ADM JZ. Performed the Norway; Jon Gunnar Tufta, Medical Birth Registry of Norway; Poland experiments: ADM JZ. Analyzed the data: ADM. Contributed reagents/ Katarzyna Szamotulska, Department of Epidemiology, National Research materials/analysis tools: Euro-Peristat Scientific Members. Wrote the Institute of Mother and Child; Bogdan Chazan, Holy Family Hospital; paper: ADM JZ. Revised the manuscript: SEB KS JC LMI FB JGN. Portugal Henrique Barros, Sofia Correia, University of Porto Medical Reviewed the manuscript: Euro-Peristat Scientific Members. School, Department of Hygiene and Epidemiology; Slovakia Jan Cap,

References 1. (2008) EURO-PERISTAT project in collaboration with SCPE, EUROCAT 17. (2011) Personal communication with Miriam Gatt, MD MSc, from the and EURONEOSTAT. Better statistics for better health for pregnant women Department of Health Information and Research in Malta. and their babies in 2004. European Perinatal Health Report 2008. Available at 18. Blondel B, Morin I, Platt RW, Kramer MS, Usher R, et al. (2002) Algorithms www.europeristat.com. for combining menstrual and ultrasound estimates of gestational age: 2. Zeitlin J, Mohangoo A, Cuttini M, Alexander S, Barros H, et al. (2009) The consequences for rates of preterm and postterm birth. Bjog 109: 718–720. European Perinatal Health Report: comparing the health and care of pregnant 19. Kramer MS, McLean FH, Boyd ME, Usher RH (1988) The validity of women and newborn babies in Europe. J Epidemiol Community Health 63: gestational age estimation by menstrual dating in term, preterm, and postterm 681–682. gestations. Jama 260: 3306–3308. 3. Zimbeck M, Mohangoo A, Zeitlin J (2009) The European perinatal health 20. Wingate MS, Alexander GR, Buekens P, Vahratian A (2007) Comparison of report: delivering comparable data for examining differences in maternal and gestational age classifications: date of last menstrual period vs. clinical estimate. infant health. Eur J Obstet Gynecol Reprod Biol 146: 149–151. Ann Epidemiol 17: 425–430. 4. Gissler M, Mohangoo AD, Blondel B, Chalmers J, Macfarlane A, et al. (2010) 21. World Health Organization (2004) ICD-10: International Statistical Classifica- Perinatal health monitoring in Europe: results from the EURO-PERISTAT tionof Diseases and Related Health Problems - Instruction Manual. 2. In:, , project. Inform Health Soc Care 35: 64–79. WHO, editor. Geneva: WHO. 5. Boyd PA, Devigan C, Khoshnood B, Loane M, Garne E, et al. (2008) Survey of 22. Lawn JE, Blencowe H, Pattinson R, Cousens S, Kumar R, et al. (2011) prenatal screening policies in Europe for structural malformations and Stillbirths: Where? When? Why? How to make the data count? Lancet 377: chromosome anomalies, and their impact on detection and termination rates 1448–1463. for neural tube defects and Down’s syndrome. Bjog 115: 689–696. 23. Verloove-Vanhorick SP, Verwey RA, Brand R, Gravenhorst JB, Keirse MJ, 6. van der Pal-de Bruin KM, Graafmans W, Biermans MC, Richardus JH, et al. (1986) Neonatal mortality risk in relation to gestational age and Zijlstra AG, et al. (2002) The influence of prenatal screening and termination of birthweight. Results of a national survey of preterm and very-low-birthweight pregnancy on perinatal mortality rates. Prenat Diagn 22: 966–972. infants in the Netherlands. Lancet 1: 55–57. 7. Lasswell SM, Barfield WD, Rochat RW, Blackmon L (2010) Perinatal 24. Larroque B, Ancel PY, Marret S, Marchand L, Andre M, et al. (2008) regionalization for very low-birth-weight and very preterm infants: a meta- Neurodevelopmental disabilities and special care of 5-year-old children born analysis. Jama 304: 992–1000. before 33 weeks of gestation (the EPIPAGE study): a longitudinal cohort study. 8. Zeitlin J, Wildman K, Breart G (2003) Perinatal health indicators for Europe: an Lancet 371: 813–820. introduction to the PERISTAT project. Eur J Obstet Gynecol Reprod Biol 111 25. Vanhaesebrouck P, Allegaert K, Bottu J, Debauche C, Devlieger H, et al. (2004) Suppl 1: S1–4. The EPIBEL study: outcomes to discharge from hospital for extremely preterm 9. Zeitlin J, Wildman K, Breart G, Alexander S, Barros H, et al. (2003) infants in Belgium. Pediatrics 114: 663–675. PERISTAT: indicators for monitoring and evaluating perinatal health in 26. Wood NS, Marlow N, Costeloe K, Gibson AT, Wilkinson AR (2000) Neurologic Europe. Eur J Public Health 13: 29–37. and developmental disability after extremely preterm birth. EPICure Study 10. Zeitlin J, Wildman K, Breart G, Alexander S, Barros H, et al. (2003) Selecting Group. N Engl J Med 343: 378–384. an indicator set for monitoring and evaluating perinatal health in Europe: 27. Zeitlin J, Draper ES, Kollee L, Milligan D, Boerch K, et al. (2008) Differences in criteria, methods and results from the PERISTAT project. Eur J Obstet Gynecol rates and short-term outcome of live births before 32 weeks of gestation in Reprod Biol 111 Suppl 1: S5–S14. Europe in 2003: results from the MOSAIC cohort. Pediatrics 121: e936–944. 11. Flenady V, Froen JF, Pinar H, Torabi R, Saastad E, et al. (2009) An evaluation 28. Graafmans WC, Richardus JH, Macfarlane A, Rebagliato M, Blondel B, et al. of classification systems for stillbirth. BMC Pregnancy Childbirth 9: 24. (2001) Comparability of published perinatal mortality rates in Western Europe: 12. Papiernik E, Zeitlin J, Delmas D, Draper ES, Gadzinowski J, et al. (2008) the quantitative impact of differences in gestational age and birthweight criteria. Termination of pregnancy among very preterm births and its impact on very BJOG 108: 1237–1245. preterm mortality: results from ten European population-based cohorts in the 29. Cole TJ, Williams AF, Wright CM (2011) Revised birth centiles for weight, MOSAIC study. Bjog 115: 361–368. length and head circumference in the UK-WHO growth charts. Ann Hum Biol 13. Garne E, Khoshnood B, Loane M, Boyd P, Dolk H (2010) Termination of 38: 7–11. pregnancy for fetal anomaly after 23 weeks of gestation: a European register- 30. Fellman V, Hellstrom-Westas L, Norman M, Westgren M, Kallen K, et al. based study. Bjog 117: 660–666. (2009) One-year survival of extremely preterm infants after active perinatal care 14. Gissler M, Ollila E, Teperi J, Hemminki E (1994) Impact of induced abortions in Sweden. Jama 301: 2225–2233. and statistical definitions on perinatal mortality figures. Paediatr Perinat 31. Lorenz JM (2001) The outcome of extreme prematurity. Semin Perinatol 25: Epidemiol 8: 391–400. 348–359. 15. Sachs BP, Fretts RC, Gardner R, Hellerstein S, Wampler NS, et al. (1995) The 32. (2010) Council of the European Union. 2010. Draft Commission Regulation impact of extreme prematurity and congenital anomalies on the interpretation of (EU) No …/… of […] implementing Regulation (EC) No. 1338/2008 of the international comparisons of infant mortality. Obstet Gynecol 85: 941–946. European Parliament and of the Council on community statistics on public 16. Lack N, Zeitlin J, Krebs L, Kunzel W, Alexander S (2003) Methodological health and health and safety at work, as regards statistics on causes of death. difficulties in the comparison of indicators of perinatal health across Europe. Available at: http://register.consilium.europa.eu/pdf/en/10/st17/st17002. Eur J Obstet Gynecol Reprod Biol 111 Suppl 1: S33–44. en10.pdf.

PLoS ONE | www.plosone.org 11 November 2011 | Volume 6 | Issue 11 | e24727 Patterns of Fetal and Neonatal Mortality in Europe

33. Smith GC, Fretts RC (2007) Stillbirth. Lancet 370: 1715–1725. 38. Hakansson S, Farooqi A, Holmgren PA, Serenius F, Hogberg U (2004) 34. Flenady V, Koopmans L, Middleton P, Froen JF, Smith GC, et al. (2011) Major Proactive management promotes outcome in extremely preterm infants: a risk factors for stillbirth in high-income countries: a systematic review and meta- population-based comparison of two perinatal management strategies. Pediatrics analysis. Lancet 377: 1331–1340. 114: 58–64. 35. Flenady V, Middleton P, Smith GC, Duke W, Erwich JJ, et al. (2011) Stillbirths: 39. Kollee LA, Cuttini M, Delmas D, Papiernik E, den Ouden AL, et al. (2009) the way forward in high-income countries. Lancet 377: 1703–1717. Obstetric interventions for babies born before 28 weeks of gestation in Europe: 36. De Leeuw R, Cuttini M, Nadai M, Berbik I, Hansen G, et al. (2000) Treatment results of the MOSAIC study. Bjog 116: 1481–1491. choices for extremely preterm infants: an international perspective. J Pediatr 40. Zeitlin J, Blondel B, Alexander S, Breart G (2007) Variation in rates of postterm 137: 608–616. birth in Europe: reality or artefact? Bjog 114: 1097–1103. 37. Verhagen AA, Janvier A, Leuthner SR, Andrews B, Lagatta J, et al. (2010) 41. Di Renzo GC, O’Herlihy C, van Geijn HP, Copray FJ (1992) Organization of Categorizing neonatal deaths: a cross-cultural study in the United States, perinatal care within the European community. Eur J Obstet Gynecol Reprod Canada, and The Netherlands. J Pediatr 156: 33–37. Biol 45: 81–87.

PLoS ONE | www.plosone.org 12 November 2011 | Volume 6 | Issue 11 | e24727 DOI: 10.1111/1471-0528.12281 Epidemiology www.bjog.org

Preterm birth time trends in Europe: a study of 19 countries

J Zeitlin,a,b K Szamotulska,c N Drewniak,a,b AD Mohangoo,d J Chalmers,e L Sakkeus,f L Irgens,g,h M Gatt,i M Gissler,j,k B Blondel,a,b The Euro-Peristat Preterm Study Group* a INSERM, UMRS 953, Epidemiological Research Unit on Perinatal and Women’s and Children’s Health, Paris, France b UPMC, Paris, France c Department of Epidemiology, National Research Institute of Mother and Child, Warsaw, Poland d Department Child Health, TNO Netherlands Organization for Applied Scientiﬁc Research, Leiden, the Netherlands e Information Services Division, NHS National Services Scotland, Edinburgh, UK f Estonian Institute for Population Studies, Tallinn University, Tallinn, Estonia g Department of Public Health and Primary Health Care, University of Bergen, Bergen, Norway h Medical Birth Registry of Norway, Norwegian Institute of Public Health, Bergen, Norway i Department of Health Information and Research, National Obstetric Information Systems (NOIS) Register, G’Mangia, Malta j Department of Information, THL National Institute for Health and Welfare, Helsinki, Finland k Nordic School of Public Health, Gothenburg, Sweden Correspondence: J Zeitlin, INSERM, UMRS 953, Epidemiological Research Unit on Perinatal and Women’s and Children’s Health, 53 avenue de l’Observatoire, 75014 Paris, France. Email [email protected]

Accepted 17 April 2013. Published Online 24 May 2013.

Objective To investigate time trends in preterm birth in Europe contributed to increases in the overall preterm birth rate. About by multiplicity, gestational age, and onset of delivery. half of countries experienced no change or decreases in the rates of singleton preterm birth. Where preterm birth rates rose, Design Analysis of aggregate data from routine sources. increases were no more prominent at 35–36 weeks of gestation – Setting Nineteen European countries. than at 32 34 weeks of gestation. Variable trends were observed for spontaneous and non-spontaneous preterm births in the 13 Population Live births in 1996, 2000, 2004, and 2008. countries with mode of onset data; increases were not solely attributed to non-spontaneous preterm births. Methods Annual risk ratios of preterm birth in each country were estimated with year as a continuous variable for all births and by Conclusions There was a wide variation in preterm birth trends in subgroup using log-binomial regression models. European countries. Many countries maintained or reduced rates of singleton preterm birth over the past 15 years, challenging a Main outcome measures Overall preterm birth rate and rate by widespread belief that rising rates are the norm. Understanding multiplicity, gestational age group, and spontaneous versus non- these cross-country differences could inform strategies for the spontaneous (induced or prelabour caesarean section) onset of labour. prevention of preterm birth.

Results Preterm birth rates rose in most countries, but the Keywords Europe, indicated preterm births, multiple births, magnitude of these increases varied. Rises in the multiple birth preterm births, time trends. rate as well as in the preterm birth rate for multiple births

Please cite this paper as: Zeitlin J, Szamotulska K, Drewniak N, Mohangoo A, Chalmers J, Sakkeus L, Irgens L, Gatt M, Gissler M, Blondel B. Preterm birth time trends in Europe: a study of 19 countries. BJOG 2013;120:1356–1365.

Introduction hood than infants born at term. In high-income countries, between two-thirds and three-quarters of neonatal deaths Infants born preterm, deﬁned as births at <37 completed occur in the 6–11% of infants born alive before 37 weeks of weeks of gestation, are at higher risk of mortality, morbidity, gestation.1 Infants born before 32 weeks of gestation are at and impaired motor and cognitive development in child- particularly high risk of adverse outcomes, with rates of infant mortality at 10–15% and of cerebral palsy at 5– *Euro-Peristat Preterm Study Group: C Vutuc (Austria); E Martens 10%,2,3 but moderate preterm birth (at 32–36 weeks of ges- (Flanders); P Velebil (Czech Republic); L Sakkeus (Estonia); M Gissler tation) is also associated with poor outcomes at birth and in – (Finland); B Blondel (France); N Lack, B Misselwitz, P Wenzlaff (Ger- childhood.4 6 Being born preterm predisposes infants to many); S Bonham (Ireland); J Jaselioniene (Lithuania); M Gatt (Malta); higher risks of chronic diseases and mortality later in life.7,8 A Mohangoo, J Nijhuis (the Netherlands); L Irgens, K Klungsøyr (Nor- Many countries have reported increased preterm birth way); K Szamotulska (Poland); H Barros (Portugal); Z Novak (Slovenia); – rates over the past two decades,9 15 and this general trend F Bolumar (Spain); K Gottvall (Sweden); James Chalmers (UK).

1356 ª 2013 The Authors. BJOG: An International Journal of Obstetrics & Gynaecology published by John Wiley and Sons on behalf of the Royal College of Obstetricians and Gynaecologists This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes. Preterm birth time trends in Europe was recently confirmed by a WHO global survey.16 There The time intervals were selected in order to allow com- are many reasons to expect preterm birth rates to rise. One parisons with other Euro-Peristat data collected in 2000 reason is increasing multiple pregnancy rates, associated and 2004. Countries that were unable to provide data for with the use of subfertility treatments and later maternal these years were asked to provide data from the closest age at childbirth.17,18 The preterm birth rate for multiples available time point. If data were not available nationally, is 40–60%, compared with 5–10% for singletons.19 Second, we requested population-based data from geographically the survival of very preterm infants has improved markedly defined regions. Appendix S1 describes data sources and over recent decades because of medical advances in neona- geographical coverage. tal care, such as antenatal corticosteroids and surfactants,20 Nineteen countries participated in the study. In Belgium, and their improved prognosis has changed perceptions of data came from Flanders, and in Germany, data came from the risk associated with prematurity versus other pregnancy three L€ander. Data from the UK came from Scotland (ges- complications. This has lowered the threshold for indicated tational age was added to routine birth registers in North- (alternatively termed non-spontaneous or provider-initi- ern Ireland, England, and Wales in 2005 only). In France, ated) preterm births, and has led to the rise in number of data came from a routine nationally representative survey – these births.21 23 Other risk factors for spontaneous and of all births. Spain and Portugal could only provide data non-spontaneous preterm birth, such as in vitro fertilisation by gestational age groups. The Czech Republic, the German (IVF), older maternal age, and higher maternal body mass L€ander, Ireland, and Malta had no data from 1996. Malta index (BMI), have also become more prevalent among and Sweden provided data from 2009 instead of 2008. Data childbearing women.10,15,24 Finally, progress in the preven- from the French survey were available for 1995, 1998, 2003, tion of preterm birth has been limited: the 2006 Institute and 2010. Most countries reported only minimal rates of of Medicine report on preterm birth and other reviews missing data for gestational age, with the exception of have concluded that the efforts for prevention have been Spain, where missing data were 11–19% depending on the largely unsuccessful.25,26 period. Missing data were minimal for other variables. In contrast to this general trend, however, recent studies Missing data were excluded from analyses. from Finland and the Netherlands have reported decreasing Austria, Ireland, Poland, Portugal, and Spain could not rates of preterm birth for singleton births.24,27 Data on pre- provide data on the onset of labour, and Slovakia only had term birth rates from the Euro-Peristat project, a collabora- this data for the last time point. Estonia, Lithuania, Malta, tion to monitor perinatal health in the European Union, also and Scotland collected data by whether the caesarean was raise the question of whether rates are rising in all countries. planned/elective or an emergency. For these latter coun- Preterm birth rates in 2004 ranged from 5 to 11%, and it is tries, planned caesarean sections were considered to occur possible that differences in trends over time explain some of before the onset of labour, although Estonia used data on this variation.1 This study was thus designed to investigate the presence of labour to recode elective caesarean sections time trends in preterm birth rates in the Euro-Peristat coun- that followed the onset of labour. tries, and how these trends differ for singleton versus multiple pregnancies, as well as preterm deliveries with a Analysis spontaneous versus a non-spontaneous onset of labour. We computed preterm birth rates for all births and for singleton and multiple births for each time point. We also Methods computed rates of multiple birth (multiple births/all births) and rates of spontaneous and non-spontaneous preterm Data birth separately, by multiplicity. We estimated risk ratios The scientific committee members of the countries partici- (RRs) of preterm birth with year as an independent contin- pating in the Euro-Peristat II project (25 European mem- uous variable in each country separately for all births, and ber states and Norway) were invited to take part in this by subgroup, using log-binomial regression models.29 Risk study.1 Aggregate data from routine population-based ratios were then transformed into percentage increases (risk sources were requested on number of births by gestational ratio À1) for presentation in graphs and tables. We used age (in completed weeks), by multiplicity, mode of delivery the exact time points available in each country. Random (vaginal or caesarean), and mode of onset of labour (cae- effects meta-analysis was used to test for heterogeneity in sarean section before labour, induction, or spontaneous), annual RRs across countries and to compute pooled mea- in 1996, 2000, 2004, and 2008. The definition of gestational sures. We also redid analyses after excluding births at 22– age was the final estimate in the obstetrical records. We 23 weeks of gestation because of concerns about cross- requested data on all live births, starting at 22 weeks of country differences in the recording of these infants, and gestation. Stillbirths were excluded because registration cri- confirmed that the results were similar. Correlations teria differ in routine sources across EU countries.28 between country-level variables were assessed with Spear-

ª 2013 The Authors. BJOG: An International Journal of Obstetrics & Gynaecology published by John Wiley and Sons on behalf of the Royal College of Obstetricians and Gynaecologists 1357 Zeitlin et al. man’s rank tests. Finally, we computed population-attribut- (0.1–0.3), and 1.3 (1.2–1.4) for all, singleton, and multiple able risks to assess the contribution of multiple births to births, respectively). Country-level trends by year for multi- the overall preterm birth rate; confidence intervals were ples and singletons were not significantly associated, computed using Walter’s limits.30 Data were analysed using although the Spearman’s correlation coefficient was positive STATA 10.0 (StataCorp LP, College Station, TX, USA). (q = 0.37, P = 0.12). Some countries experienced fluctuations in rates from Results one period to another, in particular for singletons. For instance, in Austria the rate increased over the period, but Rates and trends in preterm birth then declined slightly between 2004 and 2008. Furthermore, In 2008, preterm birth rates across Europe ranged from 5.5 not all countries could provide data for all time points. We to 11.1% for all live births, from 4.3 to 8.7% for singleton estimated annual trends for the period 2000–2008 in order births, and from 42.2 to 77.8% for multiple births to assess the sensitivity of our results to the selection of time (Table 1). The annual percentage increases in preterm birth points. Results were similar for all countries (Figure S1). were significantly >0 in 13 out of the 19 countries included To test whether countries with lower initial rates of pre- in the study for all live births (Figure 1). For singleton term birth experienced greater increases, we correlated pre- births, the percentage increases were positive for eight term birth rates in the first time period with annual trends. countries and negative in three countries. Thirteen coun- The Spearman’s correlation coefficients were negative, but tries experienced significant increases in preterm birth for the associations were not significant (all births, –0.266, multiple births, and no countries had significant decreases, P = 0.27; singleton births, –0.244, P = 0.31; and multiple although four countries had percentage changes <0 births, –0.321, P = 0.18). (Finland, the Netherlands, Sweden, and France). Meta-analysis found highly significant heterogeneity for all three Time trends in multiple births and population- measures using the Q–test; pooled RRs were over 1, but attributable risks given the extensive heterogeneity between countries, they Multiple births as a proportion of all live births ranged are of limited value (pooled measures: 0.7 (0.7–1.8), 0.2 from 2.4 to 4.0% in 2008 (Table 2). Over the study period,

Table 1. Rates of preterm birth from 1996 to 2008 in 19 European countries

Country: region/ All live births Singleton live births Multiple live births area n 1996 2000 2004 2008 n 1996 2000 2004 2008 n 1996 2000 2004 2008 (2008) % % % % (2008) % % % % (2008) % % % %

Austria 77 720 9.1 10.0 11.4 11.1 75 066 7.9 8.4 9.4 8.7 2654 58.2 67.5 74.6 77.8 Belgium: Flanders 69 187 7.0 7.8 8.1 8.0 66 672 5.2 6.0 6.3 6.2 2515 51.7 55.9 60.4 57.3 Czech Republic 119 455 5.4 7.7 8.3 114 722 4.2 6.0 6.3 4733 42.3 52.7 57.5 Estonia 16 031 5.5 5.9 5.9 6.2 15 506 4.9 5.1 4.9 4.6 525 38.5 46.2 47.6 51.0 Finland 59 486 5.8 6.1 5.6 5.5 57 767 4.5 4.7 4.4 4.3 1719 46.5 49.4 44.5 47.5 France* 14 696 5.4 6.2 6.3 6.6 14 261 4.5 4.7 5.0 5.5 435 40.5 48.2 44.3 42.1 Germany: 3 Lander€ 215 634 8.8 9.2 9.0 208 383 7.0 7.2 7.0 7251 61.7 61.8 64.2 Ireland 75 246 5.4 5.5 5.9 72 589 4.5 4.4 4.3 2657 41.8 42.3 49.9 Lithuania 31 287 5.3 5.3 5.3 5.9 30 510 4.5 4.6 4.5 4.7 777 41.3 42.6 42.7 49.4 Malta** 4152 6.0 7.2 6.7 4020 5.0 5.8 5.3 132 39.5 51.7 50.0 the Netherlands 175 160 7.8 7.7 7.4 7.4 168 829 6.2 6.0 5.7 5.7 6331 51.1 47.5 48.2 50.6 Norway 60 744 6.4 6.8 7.1 6.7 58 674 5.3 5.4 5.5 5.3 2070 43.4 43.9 49.2 48.3 Poland 414 480 6.8 6.3 6.8 6.6 404 452 6.1 5.5 5.8 5.5 10 028 43.1 44.0 50.2 51.2 Portugal 103 597 7.0 5.9 6.8 9.0 100 705 6.1 4.9 5.4 7.4 2892 45.9 49.6 54.9 63.5 Slovakia 53 624 5.1 5.4 6.3 6.8 52 227 4.4 4.5 5.2 5.6 1397 40.3 46.3 49.8 52.2 Slovenia 21 816 6.0 6.8 7.0 7.4 21 050 4.8 5.1 5.2 5.4 766 54.1 57.4 55.4 62.3 Spain 417 094 7.1 7.7 8.0 8.2 400 474 6.2 6.3 6.4 6.3 16 620 42.2 50.4 53.0 53.9 Sweden** 108 865 6.1 6.4 6.3 5.9 105 799 5.0 5.2 5.2 4.8 3066 44.1 43.4 45.2 43.3 UK: Scotland 58 275 7.0 7.4 7.6 7.7 56 423 5.8 6.1 6.3 6.1 1852 53.1 51.6 55.5 55.0

*Data from France come from a nationally representative sample of births, and the years are 1995, 1998, 2003, and 2010. **2009, instead of 2008 data.

1358 ª 2013 The Authors. BJOG: An International Journal of Obstetrics & Gynaecology published by John Wiley and Sons on behalf of the Royal College of Obstetricians and Gynaecologists Preterm birth time trends in Europe

All live births

Czech Republic* 5.1 Slovakia 2.6 Portugal 2.5 Austria 1.8 Slovenia 1.7 Malta* 1.2 Spain 1.1 Ireland* 1.1 Belgium: Flanders 1.1 France 1.0 Estonia 0.9 Lithuania 0.8 UK: Scotland 0.7 Norway 0.4 Germany: 3 Länder* 0.2 Poland –0.1 Sweden –0.3 Finland –0.6 The Netherlands –0.6 0 2 4 6 8 10

Singleton live births

Czech Republic* 4.6 Slovakia 2.2 Portugal 1.9 Belgium: Flanders 1.3 France 1.3 Slovenia 1.0 Austria 1.0 Malta* 0.7 UK: Scotland 0.5 Lithuania 0.3 Spain 0.1 Germany: 3 Länder* 0.1 Norway –0.0 Sweden –0.3 Ireland* –0.4 Estonia –0.5 Finland –0.5 Poland –0.7 The Netherlands –0.7 0 2 4 6 8 10

MulƟple live births

Czech Republic* 3.6 Portugal 2.8 Ireland* 2.5 Austria 2.3 Malta* 2.3 Slovakia 2.0 Estonia 2.0 Spain 1.6 Poland 1.6 Lithuania 1.4 Norway 1.1 Slovenia 1.1 Belgium: Flanders 0.9 Germany: 3 Länder* 0.5 UK: Scotland 0.4 Sweden –0.1 The Netherlands –0.1 Finland –0.1 France –0.3 0 2 4 6 8 10

Figure 1. Average annual percentage change for preterm birth by country, 1996–2008.* Data series begins in 2000.

ª 2013 The Authors. BJOG: An International Journal of Obstetrics & Gynaecology published by John Wiley and Sons on behalf of the Royal College of Obstetricians and Gynaecologists 1359 Zeitlin et al.

0.8 and –0.1). Divergent time trends are also observed in Table 2. Rates of multiple births per 100 live births, population- attributable risks, and average annual increases, 1996–2008 Poland, where decreases were larger for earlier preterm births. The group at 35–36 weeks of gestation represented Multiple birth Annual Population- a median of 60% of preterm births in participating rate 2008 increase attributable countries (interquartile range, 57–62%; range, 55–66%). risk 2008 Time trends in spontaneous and non-spontaneous – Austria 3.4 3.2 21.3 (19.6 23.1) preterm birth Belgium: Flanders 3.6 –0.6* 23.2 (21.1–25.2) Czech Republic 4.0 3.3* 24.5 (22.9–26.0) For singletons, the rates of non-spontaneous preterm births Estonia 3.3 5.5* 24.7 (20.2–29.2) ranged from 1.1 to 3.0% in 2008, whereas spontaneous Finland 2.9 –0.9* 22.5 (20.1–25.0) onset births ranged from 2.8 to 4.8% (Table 3). For multi- France** 3.0 0.4 16.5 (11.6–21.4) ples, the rates of non-spontaneous preterm birth ranged Germany: 3 Lander€ 3.4 0.3 21.5 (20.3–22.7) from 12.0 to 34.4%, and spontaneous onset births from Ireland 3.5 3.9* 27.2 (25.2–29.2) 15.1 to 38.2%. In each country, spontaneous preterm births – Lithuania 2.5 1.8* 18.9 (15.3 22.6) were more frequent than non-spontaneous preterm births, Malta*** 3.2 0.8 21.1 (12.1–30.1) the Netherlands 3.6 –0.3 22.1 (20.8–23.4) with a few exceptions (Germany and Norway for singleton Norway 3.5 1.2* 21.7 (19.4–24.0) and multiple births, France and Malta for singleton births, Poland 2.4 1.8* 16.8 (15.8–17.9) and Belgium, Czech Republic, and Lithuania for multiple Portugal 2.8 2.5* 17.4 (15.5–19.3) births). Slovakia 2.7 2.7* 17.8 (15.0–20.5) Countries had differing time trends for non-spontaneous Slovenia 3.5 2.6* 26.9 (23.2–30.5) and spontaneous births for singleton births (Figure 3). In – Spain 3.8 3.2* 23.1 (22.3 24.0) some countries both types of preterm birth increased (Bel- Sweden*** 2.8 –0.6* 18.4 (16.5–20.2) UK: Scotland 3.2 1.2* 20.2 (17.8–22.6) gium and Czech Republic), in others non-spontaneous preterm births increased, whereas spontaneous preterm births *Confidence interval does not include 0. either remained unchanged or declined (France, Norway, **Data from France come from a nationally representative sample and Sweden). Finally, some countries had increases in of births, and the years are 1995, 1998, 2003, and 2010. spontaneous preterm births with no change in non-sponta- ***2009, instead of 2008 data. neous preterm births (Scotland and Germany). For multiples, in contrast, non-spontaneous preterm births increased this proportion was stable or decreasing in Belgium, Fin- in almost all countries. In Sweden and the Netherlands, land, the Netherlands, and Sweden, and increased steeply where rates of multiple preterm births were stable, these in Austria, the Czech Republic, Estonia, Ireland, and Spain. increases were offset by the decline in spontaneous preterm There was a significant association between the increase in births. the proportion of multiple births and the increase in pre- q = = term birth (Spearman’s 0.66, P 0.021). The propor- Discussion tion of the overall preterm birth rate attributable to multiples in 2008 ranged from about 17% in France, Time trends in preterm births in Europe between 1996 and Poland, and Portugal, to 27% in Ireland and Slovenia. 2008 were highly heterogeneous, although the overall preterm birth rate and the multiple preterm birth rate Time trends by gestational age group increased in most countries. In contrast, singleton preterm Figure 2 displays annual trends by gestational age group birth rates were stable or decreased in about half of the for singletons and multiples. Countries are ordered as in countries in this analysis, challenging a widespread belief Figure 1. Although there was more variability in our esti- that rising rates have been the norm. In countries with rate mates because of the smaller samples, this figure shows that increases, these were observed for all gestational age groups, increases in preterm birth were less marked for births at not just the births closest to term. <32 weeks of gestation, in particular for multiples. Our study is limited by the data available from national Increases were not greatest for the 35–36 weeks of gestation systems: for instance, several countries did not have data group, and in many countries the largest proportional for all the requested time points. We estimated annual changes were observed between 32 and 34 weeks of gesta- trends using the available data points to compare across tion. Although many countries had similar trends for all countries despite this limitation; a sensitivity analysis com- gestational age groups, patterns could vary: the Netherlands puting trends from 2000 to 2008 showed that our results experienced increases for singleton births at <32 weeks of were robust to the choice of period. Because our question gestation (0.9), but decreases for the two other groups (– was whether rates were rising, we tested for linear trends.

1360 ª 2013 The Authors. BJOG: An International Journal of Obstetrics & Gynaecology published by John Wiley and Sons on behalf of the Royal College of Obstetricians and Gynaecologists Preterm birth time trends in Europe

Figure 2. Average annual percentage change for birth at <32 weeks of gestation, 32–34 weeks of gestation, and 35–36 weeks of gestation among singleton live births (A), and among multiple live births (B), 1996–2008.

Rate fluctuations occurred in some countries, but no con- the study period in some countries also confirms that large sistent patterns could be discerned, and we chose not to variations of this indicator are plausible. model these rises and falls. More generally, it was not possible to assess the quality Some countries could not provide data on the mode of of data collection and case ascertainment; previous work in the onset of labour, and among those that did, definitions the Euro-Peristat group has found significant heterogeneity differed (‘elective’ versus ‘pre-labour’ caesareans), although in routine data systems in Europe with respect to organisa- they were stable over the study period. Questions also exist tion and scope.33 However, this study was restricted to about the measure of gestational age. We requested gesta- population-based reporting systems with high coverage,33 tional age based on a common definition, the best obstetri- and used a pre-established protocol with common defini- cal estimate, but we were unable to assess how clinicians tions developed collaboratively with participating data pro- assigned this estimate.31 Dating pregnancies using ultra- viders. This represents a strength over previous sound shifts the gestational age distribution to the left, and international studies that have relied on data in published can increase the preterm birth rate,32 but it can also reports and were unable to specify a priori definitions.16 decrease the rate by reducing errors in gestational age esti- Missing data on gestational age were low, with the excep- mates.31 We cannot exclude the possibility that the rates of tion of Spain, where civil registration data rely on parental preterm birth were affected by an increased use of ultra- reports,34 and estimated trends in this case must be viewed sound for the dating of pregnancies over time, but in many with caution. European countries ultrasound dating was already widely We requested data on live births instead of total births used in the mid-1990s,11,13,27 and it is not clear whether because of the differences in registration of stillbirths this would lead to systematic upward or downward trends. between European countries.28,33 Although it is important A part of the wide variation in preterm birth rates across to consider the impact of stillbirths because many indicated countries (5–11%) may result from differences in how ges- preterm deliveries aim to reduce stillbirths, this exclusion is tational age is estimated; however, the fact that we unlikely to affect our conclusions as preterm stillbirth is a observed substantial changes in the preterm birth rate over rare outcome (about 2 per 1000 total births) compared

ª 2013 The Authors. BJOG: An International Journal of Obstetrics & Gynaecology published by John Wiley and Sons on behalf of the Royal College of Obstetricians and Gynaecologists 1361 Zeitlin et al.

Table 3. Spontaneous and non-spontaneous preterm births per 100 live births by multiplicity from 1996 to 2008

Country: region/area Singleton births Multiple births

Spontaneous onset Non-spontaneous onset Spontaneous onset Non-spontaneous onset

1996 2000 2004 2008 1996 2000 2004 2008 1996 2000 2004 2008 1996 2000 2004 2008

Austria Belgium: Flanders 3.8 3.9 4.3 4.2 1.5 2.1 2.1 2.0 29.0 33.0 33.4 30.6 22.7 22.9 27.0 26.7 Czech Republic 3.1 4.4 4.4 1.1 1.6 1.9 23.3 27.0 26.2 19.0 25.7 31.3 Estonia 3.4 3.9 3.8 3.6 1.4 1.1 1.1 1.1 29.9 30.2 30.5 33.5 8.7 16.0 17.1 17.5 Finland 3.3 3.7 3.5 3.2 1.1 1.0 0.9 1.1 30.7 35.9 29.0 31.9 15.8 13.5 15.5 15.5 France* 3.0 2.9 2.7 2.8 1.5 1.7 2.3 2.6 22.6 31.2 20.9 21.8 18.0 17.0 23.1 20.2 Germany: 3 Lander€ 3.8 4.0 4.0 3.0 3.1 3.0 27.4 27.8 32.0 32.3 33.1 32.1 Ireland Lithuania 3.0 3.1 3.1 3.2 1.5 1.5 1.4 1.5 23.0 23.0 23.9 15.1 18.3 19.5 18.5 34.4 Malta** 3.9 3.5 4.2 0.9 2.3 1.1 25.6 32.5 32.6 12.0 19.2 17.4 the Netherlands 4.4 4.4 4.2 3.9 1.7 1.6 1.5 1.8 34.3 32.1 32.8 29.9 15.9 15.4 15.4 20.7 Norway 3.2 3.3 3.1 3.1 1.6 2.1 2.3 2.1 24.5 24.5 25.6 25.3 14.7 19.0 23.1 21.6 Poland Portugal Slovakia 4.3 1.2 38.2 12.0 Slovenia 4.1 4.1 4.2 4.2 0.7 1.0 1.0 1.3 41.0 46.5 39.9 37.6 11.7 10.9 15.6 24.7 Spain Sweden** 3.2 3.3 3.4 3.1 1.6 1.7 1.7 1.7 27.4 28.5 28.1 25.0 16.6 13.9 16.5 17.9 UK: Scotland 4.5 4.8 5.0 4.8 1.3 1.2 1.2 1.3 39.9 36.3 36.9 36.3 13.2 15.3 18.6 18.8

*Data from France come from a nationally representative sample of births, and the years are 1995, 1998, 2003, and 2010. **2009, instead of 2008 data.

with live preterm birth.1 We set a common lower inclusion Belgium and Sweden.35 eSET has also been extensively pro- limit of 22 weeks of gestation for this study, and recom- moted in Finland, despite the fact that it is not mandatory puted time trends after the exclusion of births under nor an ofﬁcial policy.36 In contrast, other European coun- 24 weeks of gestation to verify that differences between tries have no such policies: in Italy, the law requires the countries in registration practices for live births at the lim- transfer of all fertilised embryos in each cycle, although it its of viability had no impact on our ﬁndings. limits the number of fertilised embryos to three.37 Data Our results show that the preterm birth rates for all collected by the European Society of Human Reproduction births rose in many European countries, as was also found and Embryology (ESHRE) from IVF centres documents by the recent WHO study of preterm birth trends based on wide differences in the rates of single embryo transfer publicly available data in 64 countries in developed regions, across Europe (from 10 to ~70%)17; countries in our analy- Latin America, and the Caribbean.16 Our results add to this sis with negative trends in their preterm birth rates, such as overview, however, by revealing that time trends can differ Belgium, Finland, and Sweden, had a high proportion of substantially between the overall preterm birth rate and the eSET (50.4, 62.1, and 69.5%, respectively). In contrast, singleton preterm birth rate, that trends were similar across countries with increases in their multiple birth rate had a gestational age groups, and by documenting changes in lower proportion of single embryo transfers (Austria, multiple births rates over time and their contribution to 22.6%; Ireland, 19.1; and Portugal, 19.0). the overall preterm rates. Multiple births also affected the overall preterm birth We found a strong correlation between increases in mul- rate because of increases in the preterm birth rate among tiple births and preterm birth, corroborating previous stud- multiples. For multiple births, and with the data on mode ies.18 Policies related to the use of assisted reproductive of onset of labour included in the analysis, non-spontane- technology (ART) are highly variable in Europe, and these ous preterm birth rates increased in almost all countries. In affect the multiple birth rate resulting from ART.17 For almost all countries with data on mode of onset of labour, instance, national elective single embryo transfer (eSET) non-spontaneous preterm birth rates increased. Overall, policies have been adopted by several countries, including our data showed that the population-attributable risk asso-

1362 ª 2013 The Authors. BJOG: An International Journal of Obstetrics & Gynaecology published by John Wiley and Sons on behalf of the Royal College of Obstetricians and Gynaecologists Preterm birth time trends in Europe

Figure 3. Average annual percentage change for spontaneous (A) and non-spontaneous (B) preterm births among singleton live births, and annual rate ratios for spontaneous (C) and non-spontaneous (D) preterm births among multiple live births, 1996–2008. ciated with multiple pregnancies was substantial, ranging Australia, Finland, and the Netherlands.11,13,15,24,27 Rates of from 17 to 27%. spontaneous preterm births rose in some countries, and We found that many countries had unchanging or where overall preterm birth rates decreased, these trends declining singleton preterm birth rates, as also shown by affected spontaneous preterm births. The reasons for trends studies from Finland and the Netherlands over different in the spontaneous preterm birth rate are poorly under- time periods,24,27 while elsewhere preterm birth rates rose stood, and countries with similar populations have experi- considerably. We found increases in non-spontaneous pre- enced divergent trends, as in Denmark and Finland, for term births in some countries, corroborating other studies instance.11,24 Researchers have proposed a range of factors concluding that these births were a driving force behind that could contribute to varying preterm birth rates rising preterm birth rates.13,15,22,38 However, we observed between populations, including older maternal age, obesity, extensive heterogeneity in the proportions of preterm births higher-risk migrant populations, smoking during preg- by mode of onset of labour, and in the evolution of non- nancy, use of IVF, diabetes, Chlamydia trachomatis infec- spontaneous preterm births over time. A consistent pattern tion, and previous induced abortions, but their relative of rising preterm birth rates driven primarily by non-spon- contribution remains to be established.11,13,15,24,27 Obstetric taneous preterm births was not detected. practices related to the management of preterm birth risk We also showed that spontaneous preterm births played (screening for short cervix, use of progesterone, and pre- a role in determining overall trends, as reported in other scription of bed rest, for instance) may differ across coun- in-depth studies of preterm birth in Denmark, Scotland, tries; however, we are not aware of any studies that have

ª 2013 The Authors. BJOG: An International Journal of Obstetrics & Gynaecology published by John Wiley and Sons on behalf of the Royal College of Obstetricians and Gynaecologists 1363 Zeitlin et al. assessed variations in these practices across countries and Contribution to authorship their impact on national preterm birth rates. The preva- JZ, BB, and KS conceived the study, ND carried out statis- lence of work leaves for pregnant women differ in Europe, tical analysis, ADM, JC, LS, LI, MG, and MG contributed and this may reduce the impact of work-related risk factors to the interpretation of the results and revised successive on preterm birth.39 Economic factors may also play a role: versions of the article. Members of the Euro-Peristat pre- some studies find that preterm birth rates have risen more term birth group were responsible for the provision, accu- steeply among women of lower socio-economic status.9 racy, and interpretation of data in their country: they Comparative cross-national studies provide an opportunity commented on initial and final versions of the article. All to test these multiple hypotheses; the Euro-Peristat network authors approved the final article. as well as birth cohorts that have been established in Eur- ope are promising platforms for future research in this Details of ethics approval area. This article is based on the analysis of aggregate data pro- Although annual changes in the rate of preterm birth were vided from routine data sources, and is exempt from ethical modest in most countries, the impact is substantial when approvals at INSERM in France. The transmission of data assessed in terms of the numbers of preterm infants. If every was consistent with existing authorisations for each routine country had experienced trends similar to Finland or the data source in terms of the allowable minimum cell sizes. Netherlands over the study period (–0.6% per year), over 24 000 fewer preterm babies would have been born in 2008, Funding or 1.2% of the over two million births in the participating This analysis was partially funded by a grant to the Euro-Peri- countries. Evaluating the health impact of rising rates is stat project from the European Commission (2010 13 01). JZ more complex than computing the number of ‘excess’ pre- also received funding from the European Commission, term infants, however. Several studies have suggested that Research Directorate, Marie Curie, IOF Fellowship, grant no. rises in the rate of indicated preterm births may be associ- 254171. The funding agency was not involved in the study. ated with better perinatal outcomes. For twins, more intensive prenatal care was related to higher rates of preterm Supporting Information birth, and mothers receiving more intensive care had lower neonatal mortality.40 For singletons, mortality rates were Additional Supporting Information may be found in the observed to decline more steeply among non-spontaneous online version of this article: than spontaneous preterm births.41 On the other hand, there Figure S1. Annual percentage changes of preterm birth – is a growing body of research documenting the adverse by year, 2000 2008. & short- and longer-term health consequences of being born Appendix S1. Data sources. preterm, even at later gestational ages.6,8 The large variability in the proportions of non-spontaneous preterm births sug- References gests that there are contrasting interpretations of the current 1 EURO-PERISTAT project in collaboration with SCPE, EUROCAT and evidence base related to the positive and negative conse- EURONEOSTAT. Better statistics for better health for pregnant quences of inducing a delivery before term. women and their babies in 2004. European Perinatal Health Report 2008. Available at www.europeristat.com, 2008. 2 Larroque B, Ancel PY, Marret S, Marchand L, Andre M, Arnaud C, Conclusion et al. Neurodevelopmental disabilities and special care of 5-year-old children born before 33 weeks of gestation (the EPIPAGE study): a Time trends in the rates of preterm birth since the mid- longitudinal cohort study. Lancet 2008;371:813–20. 1990s show a striking diversity in 19 European countries. 3 Zeitlin J, Draper ES, Kollee L, Milligan D, Boerch K, Agostino R, For multiples, rates have generally increased, although the et al. Differences in rates and short-term outcome of live births range is wide; for singletons, however, the direction of before 32 weeks of gestation in Europe in 2003: results from the – change differs. These results call for further examination of MOSAIC cohort. Pediatrics 2008;121:e936 44. 4 Kramer MS, Demissie K, Yang H, Platt RW, Sauve R, Liston R. The reproductive and perinatal health policies and medical contribution of mild and moderate preterm birth to infant mortality. practices in European countries, and for an assessment of Fetal and Infant Health Study Group of the Canadian Perinatal their impact on the population risk of preterm birth. To Surveillance System. JAMA 2000;284:843–9. enable comparative analyses, data on preterm birth need to 5 Gouyon JB, Vintejoux A, Sagot P, Burguet A, Quantin C, Ferdynus – be included in international health databases. C. Neonatal outcome associated with singleton birth at 34 41 weeks of gestation. Int J Epidemiol 2010;39:769–76. 6 Boyle EM, Poulsen G, Field DJ, Kurinczuk JJ, Wolke D, Alfirevic Z, Disclosure of interests et al. Effects of gestational age at birth on health outcomes at 3 The authors have no conflicts of interest or disclosures to and 5 years of age: population based cohort study. BMJ 2012;344: declare. e896.

1364 ª 2013 The Authors. BJOG: An International Journal of Obstetrics & Gynaecology published by John Wiley and Sons on behalf of the Royal College of Obstetricians and Gynaecologists Preterm birth time trends in Europe

7 Crump C, Sundquist K, Sundquist J, Winkleby MA. Gestational age at 25 Iams JD, Romero R, Culhane JF, Goldenberg RL. Primary, secondary, birth and mortality in young adulthood. JAMA 2011;306:1233–40. and tertiary interventions to reduce the morbidity and mortality of 8 Crump C, Sundquist K, Winkleby MA, Sundquist J. Early-term birth preterm birth. Lancet 2008;371:164–75. (37–38 weeks) and mortality in young adulthood. Epidemiology 26 Behrman RE, Butler AS, editors. Preterm Birth: Causes, Consequences, 2013;24:270–6. and Prevention. Institute of Medicine (US) Committee on Understanding 9 Auger N, Gamache P, Adam-Smith J, Harper S. Relative and Premature Birth and Assuring Healthy Outcomes. Washington (DC): absolute disparities in preterm birth related to neighborhood The National Academies Press (US), 2007. education. Ann Epidemiol 2011;21:481–8. 27 Schaaf JM, Mol BW, Abu-Hanna A, Ravelli AC. Trends in preterm 10 Keirse MJ, Hanssens M, Devlieger H. Trends in preterm births in birth: singleton and multiple pregnancies in the Netherlands, 2000– Flanders, Belgium, from 1991 to 2002. Paediatr Perinat Epidemiol 2007. BJOG 2011;118:1196–204. 2009;23:522–32. 28 Mohangoo AD, Buitendijk SE, Szamotulska K, Chalmers J, Irgens 11 Langhoff-Roos J, Kesmodel U, Jacobsson B, Rasmussen S, Vogel I. LM, Bolumar F, et al. Gestational age patterns of fetal and neonatal Spontaneous preterm delivery in primiparous women at low risk in mortality in europe: results from the Euro-Peristat project. PLoS ONE Denmark: population based study. BMJ 2006;332:937–9. 2011;6:e24727. 12 Martin JA, Hamilton BE, Sutton PD, Ventura SJ, Menacker F, 29 Spiegelman D, Hertzmark E. Easy SAS calculations for risk or prevalence Kirmeyer S, et al. Births: final data for 2005. Natl Vital Stat Rep ratios and differences. Am J Epidemiol 2005;162:199–200. 2007;56:1–103. 30 Walter SD. Calculation of attributable risks from epidemiological 13 Norman JE, Morris C, Chalmers J. The effect of changing patterns of data. Int J Epidemiol 1978;7:175–82. obstetric care in Scotland (1980–2004) on rates of preterm birth 31 Wingate MS, Alexander GR, Buekens P, Vahratian A. Comparison of and its neonatal consequences: perinatal database study. PLoS Med gestational age classifications: date of last menstrual period vs. 2009;6:e1000153. clinical estimate. Ann Epidemiol 2007;17:425–30. 14 Ooki S. The effect of an increase in the rate of multiple births on 32 Blondel B, Morin I, Platt RW, Kramer MS, Usher R, Breart G. low-birth-weight and preterm deliveries during 1975–2008. J Algorithms for combining menstrual and ultrasound estimates of Epidemiol 2010;20:480–8. gestational age: consequences for rates of preterm and postterm 15 Tracy SK, Tracy MB, Dean J, Laws P, Sullivan E. Spontaneous birth. BJOG 2002;109:718–20. preterm birth of liveborn infants in women at low risk in Australia 33 Gissler M, Mohangoo AD, Blondel B, Chalmers J, Macfarlane A, over 10 years: a population-based study. BJOG 2007;114:731–5. Gaizauskiene A, et al. Perinatal health monitoring in Europe: results 16 Blencowe H, Cousens S, Oestergaard MZ, Chou D, Moller AB, from the EURO-PERISTAT project. Inform Health Soc Care Narwal R, et al. National, regional, and worldwide estimates of 2010;35:64–79. preterm birth rates in the year 2010 with time trends since 1990 for 34 Juarez S, Alonso Ortiz T, Ramiro-Farinas D, Bolumar F. The quality of selected countries: a systematic analysis and implications. Lancet vital statistics for studying perinatal health: the Spanish case. 2012;379:2162–72. Paediatr Perinat Epidemiol 2012;26:310–5. 17 de Mouzon J, Goossens V, Bhattacharya S, Castilla JA, Ferraretti AP, 35 Cook JL, Collins J, Buckett W, Racowsky C, Hughes E, Jarvi K. Korsak V, et al. Assisted reproductive technology in Europe, 2007: Assisted reproductive technology-related multiple births: Canada in results generated from European registers by ESHRE. Hum Reprod an international context. J Obstet Gynaecol Can 2011;33:159–67. 2012;27:954–66. 36 Tiitinen A, Gissler M. Effect of in vitro fertilization practices on 18 Blondel B, Kogan MD, Alexander GR, Dattani N, Kramer MS, multiple pregnancy rates in Finland. Fertil Steril 2004;82:1689–90. Macfarlane A, et al. The impact of the increasing number of 37 La Sala GB, Nicoli A, Villani MT, Rondini I, Moscato L, Blickstein I. multiple births on the rates of preterm birth and low birthweight: The 2004 Italian legislation on the application of assisted an international study. Am J Public Health 2002;92:1323–30. reproductive technology: epilogue. Eur J Obstet Gynecol Reprod Biol 19 Blondel B, Macfarlane A, Gissler M, Breart G, Zeitlin J. Preterm birth 2012;161:187–9. and multiple pregnancy in European countries participating in the 38 Ananth CV, Joseph KS, Oyelese Y, Demissie K, Vintzileos AM. PERISTAT project. BJOG 2006;113:528–35. Trends in preterm birth and perinatal mortality among singletons: 20 Saigal S, Doyle LW. An overview of mortality and sequelae of United States, 1989 through 2000. Obstet Gynecol 2005;105(5 Pt preterm birth from infancy to adulthood. Lancet 2008;371:261–9. 1):1084–91. 21 Goldenberg RL, Gravett MG, Iams J, Papageorghiou AT, Waller SA, 39 Saurel-Cubizolles MJ, Zeitlin J, Lelong N, Papiernik E, Di Renzo GC, Kramer M, et al. The preterm birth syndrome: issues to consider in Breart G. Employment, working conditions, and preterm birth: creating a classification system. Am J Obstet Gynecol 2012; results from the Europop case-control survey. J Epidemiol 206:113–8. Community Health 2004;58:395–401. 22 Zhang X, Kramer M. The rise in singleton preterm births in the USA: 40 Kogan MD, Alexander GR, Kotelchuck M, MacDorman MF, Buekens the impact of labour induction. BJOG 2012;119:1309–15. P, Martin JA, et al. Trends in twin birth outcomes and prenatal care 23 MacDorman MF, Declercq E, Zhang J. Obstetrical intervention and utilization in the United States, 1981–1997. JAMA 2000;284:335–41. the singleton preterm birth rate in the United States from 1991– 41 Lisonkova S, Hutcheon JA, Joseph KS. Temporal trends in neonatal 2006. Am J Public Health 2010;100:2241–7. outcomes following iatrogenic preterm delivery. BMC Pregnancy 24 Jakobsson M, Gissler M, Paavonen J, Tapper AM. The incidence of Childbirth 2011;11:39. preterm deliveries decreases in Finland. BJOG 2008;115:38–43.

ª 2013 The Authors. BJOG: An International Journal of Obstetrics & Gynaecology published by John Wiley and Sons on behalf of the Royal College of Obstetricians and Gynaecologists 1365