SEARCH

SEARCH BY CITATION

Keywords:

  • International comparisons;
  • perinatal death;
  • preterm birth

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. Disclosure of interests
  10. Contribution to authorship
  11. Details of ethics approval
  12. Funding
  13. References
  14. Commentary on ‘The ecological design’

Please cite this paper as: Lisonkova S, Sabr Y, Butler B, Joseph K. International comparisons of preterm birth: higher rates of late preterm birth are associated with lower rates of stillbirth and neonatal death. BJOG 2012;119:1630–1639.

Objective  To examine international rates of preterm birth and potential associations with stillbirths and neonatal deaths at late preterm and term gestation.

Design  Ecological study.

Setting  Canada, USA and 26 countries in Europe.

Population  All deliveries in 2004.

Methods  Information on preterm birth (<37, 32–36, 28–31 and 24–27 weeks of gestation) and perinatal deaths was obtained for 28 countries. Data sources included files and publications from Statistics Canada, the EURO-PERISTAT project and the National Center for Health Statistics. Pearson correlation coefficients and random-intercept Poisson regression were used to examine the association between preterm birth rates and gestational age-specific stillbirth and neonatal death rates. Rate ratios with 95% confidence intervals were estimated after adjustment for maternal age, parity and multiple births.

Main outcome measures  Stillbirths and neonatal deaths ≥32 and ≥37 weeks of gestation.

Results  International rates of preterm birth (<37 weeks) ranged between 5.3 and 11.4 per 100 live births. Preterm birth rates at 32–36 weeks were inversely associated with stillbirths at ≥32 weeks (adjusted rate ratio 0.94, 95% CI 0.92–0.96) and ≥37 weeks (adjusted rate ratio 0.88, 95% CI 0.85–0.91) of gestation and inversely associated with neonatal deaths at ≥32 weeks (adjusted rate ratio 0.88, 95% CI 0.85–0.91) and ≥37 weeks (adjusted rate ratio 0.82, 95% CI 0.78–0.86) of gestation.

Conclusions  Countries with high rates of preterm birth at 32–36 weeks of gestation have lower stillbirth and neonatal death rates at and beyond 32 weeks of gestation. Contemporary rates of preterm birth are indicators of both perinatal health and obstetric care services.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. Disclosure of interests
  10. Contribution to authorship
  11. Details of ethics approval
  12. Funding
  13. References
  14. Commentary on ‘The ecological design’

Preterm birth is the leading cause of neonatal morbidity and mortality and an important public health concern.1,2 Preterm birth rates vary widely among industrialised countries in Europe and North America, ranging from 5.5 per 100 live births in Ireland to 12.4 per 100 live births in the USA.3,4 In the USA and Canada, high rates of preterm birth have been implicated as a major contributor to poor international ranking in infant mortality.3,5,6 Despite numerous preventive efforts, preterm birth rates have increased in most industrialised countries during the past two decades.7–12 In the USA, preterm birth increased by 36% from 9.4 to 12.3 per 100 live births between 1981 and 2008,10,11 while in Denmark, the preterm birth rate rose by 22% from 5.2 to 6.3 per 100 live births between 1995 and 2004.12 This trend has resulted in further calls for preventive action.13–15 For example, the US initiative Healthy People 2020 calls for a 10% reduction in preterm birth rates by 2020.15

The primary reason for this upward trend in preterm birth rates is the rise in obstetric intervention (i.e. medically indicated labour induction and caesarean delivery), partly in response to the increased prevalence of risk factors such as older maternal age, prepregnancy obesity and multiple pregnancies.11,16–22 Given the advances in neonatal care over the last few decades, the chances of intact survival of preterm newborns have increased and the benefits of preterm delivery have become increasingly more likely to outweigh the benefits of expectant management in pregnancies with fetal compromise, especially at late preterm gestation. In fact, most of the increase in iatrogenic preterm birth has occurred at late preterm gestation (34–36 weeks) among high-risk pregnancies and these increases in preterm birth have been accompanied by declines in stillbirth and neonatal death rates.7,16,17 Nevertheless, this seemingly counterintuitive development has not received adequate attention in the literature, much of which has focused on the increased rates of neonatal respiratory morbidity associated with late preterm birth.23–28

We carried out a study to compare rates of preterm birth among industrialised countries in Europe and North America. Our goal was to estimate the impact of increases in mild and moderate preterm birth (32–36 weeks of gestation) on the overall rates of preterm birth and to examine the association between regional differences in mild/moderate preterm birth and perinatal mortality.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. Disclosure of interests
  10. Contribution to authorship
  11. Details of ethics approval
  12. Funding
  13. References
  14. Commentary on ‘The ecological design’

International comparisons of preterm birth rates were carried out using 2004 data from Canada, Europe and USA. Canadian data were obtained from the vital statistics databases of Statistics Canada. Data from the province of Ontario were excluded because of documented problems with data accuracy.29 European data were abstracted from the European Perinatal Health Report (PERISTAT) which described and summarised detailed information on perinatal health indicators from most European countries.4 Countries without gestational age-specific information were excluded. Data from the US National Center for Health Statistics (NCHS) were used to obtain preterm birth rates for the US.30 Preterm birth rates were based on the clinical estimate of gestation, because this is more accurate than dating based on the last menstrual period.31–33 The clinical estimate of gestation in NCHS files was that provided by the healthcare provider, without specification of the source (i.e. whether based on clinical examination, ultrasound, etc.). California did not report clinical estimates of gestation for the study period and was not included in the analysis (13.5% of all births in the USA). In addition to the above information on preterm birth rates, stillbirth and neonatal death, we obtained information about each country’s proportion of older mothers (35 years of age or older), proportion of primiparous mothers and multiple birth rate from the same data sources.

Live births and stillbirths at <24 weeks of gestation were excluded from the study to avoid between-country variations in birth registration. The preterm birth rate was defined as the number of live births before 37 completed weeks of gestation per 100 live births. Categories of preterm birth included mild/moderate preterm birth (32–36 weeks of gestation), very preterm birth (28–31 weeks of gestation), and extremely preterm birth (between 24 and 27 weeks of gestation). We ranked Canada, Europe (26 countries) and the USA by their rates of preterm birth and by subcategories of preterm birth. Pearson correlation coefficients were used to examine the correlation between the overall preterm birth rate (<37 weeks of gestation) and mild/moderate preterm, very preterm and extremely preterm birth rates.

Pearson correlation coefficients were also used to examine ecological associations between the mild/moderate preterm birth rate and four perinatal outcomes: stillbirth rates at ≥32 and ≥37 weeks of gestation, and neonatal death rates at ≥32 and ≥37 weeks of gestation. Weighted Pearson correlation coefficients were calculated using the number of total births in each country for weighting. We fitted four random-intercept Poisson regression models (SAS GLIMMIX procedure; SAS Institute Inc., Cary, NC, USA) to analyse these associations (between mild/moderate preterm birth and the four mortality outcomes) while adjusting for the proportion of older mothers (35 years of age or older), the proportion of primiparous mothers and the multiple birth rate in each country. Each country represented one unit of analysis and a total of 28 countries were included in this ecological analysis. The number of total births in each country was used to offset the observed number of perinatal deaths.

Sensitivity analyses were carried out to estimate the influence of missing data on gestational age in the regression analyses examining the relation between mild/moderate preterm birth and stillbirth/neonatal death rates. In these analyses, stillbirths/neonatal deaths with missing gestational age were assumed to have gestational ages ≥32 weeks of gestation (or a gestational age ≥37 weeks of gestation). Sensitivity analyses were also carried out after excluding the USA, because this was a highly influential observation. Finally, sensitivity analyses were also carried out after excluding countries that had more than 1% of live births with missing gestational age. In addition, we compared preterm birth rates and gestational-age specific stillbirth and neonatal death rates between Canada (2003) and England, Wales and Northern Ireland (2004). Canadian data (excluding Ontario) were obtained from the Canadian Perinatal Health Report,29 while data for England, Wales and Northern Ireland were obtained from the Confidential Enquiry into Maternal and Infant Death.34 All analyses were performed using SAS statistical package version 9.2 (SAS Institute Inc.).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. Disclosure of interests
  10. Contribution to authorship
  11. Details of ethics approval
  12. Funding
  13. References
  14. Commentary on ‘The ecological design’

Preterm birth rates (<37 weeksof gestation) varied widely across Canada, Europe and the USA (Table 1), with the highest rates observed in Austria, the USA and Germany (11.4, 10.8 and 8.9 per 100 live births, respectively) and the lowest rates seen in Lithuania, Ireland and Finland (5.3, 5.5 and 5.6 per 100 live births, respectively). The inter-country variation in preterm birth <37 weeks and preterm birth between 32 and 36 weeks of gestation followed similar patterns and 79% of countries retained the same rank (±1 rank) for both indicators. However, only 29% and 25% of countries retained the same rank (±1 rank) when ranks based on preterm birth rates <37 weeks of gestation were contrasted with ranks based on very preterm (28–31 weeks) and extremely preterm (24–27 weeks) birth rates. Only a few countries had consistently high (e.g. USA) or low (e.g. Finland) rates across all preterm birth categories. In contrast, many countries with a relatively high overall preterm birth rate <37 weeks had a low rate of very preterm birth or extremely preterm birth (e.g. Canada, Belgium and Spain), whereas others had a low rate of preterm birth <37 weeks and a high rate of very preterm birth or extremely preterm birth (e.g. Latvia and Northern Ireland, Table 1).

Table 1. Preterm birth rates by country and preterm birth category, Europe and North America, 2004
CountryLive births*Preterm birth category (weeks)
<3732–3628–3124–27
Rate per 100 live birthsRankRate per 100 live birthsRankRate per 100 live birthsRankRate per 100 live birthsRank
  1. *All live births with non-missing gestational age

  2. **Preterm birth rates for the USA are based on the clinical estimate of gestation

  3. ***All P-values for Pearson correlation coefficient <0.001

  4. Data sources: PERISTAT report (Europe), Statistics Canada, and the National Centre for Health Statistics (USA).

Austria78 93411.39289.98280.98270.3723
Belgium75 8497.79226.85220.65100.2711
Canada204 8718.15246.96230.72140.3018
Czech Republic97 6717.00165.92160.77190.2914
Denmark64 4676.93145.87130.77210.269
England and Wales640 3877.53206.17180.91250.3824
Estonia13 9535.9154.8850.6140.3825
Finland57 4535.5634.6840.6030.236
Germany646 6268.86267.54260.88230.3826
Greece104 3496.0275.1760.6480.202
Hungary95 0188.56257.13240.94260.5027
Ireland62 0405.5224.5230.67120.2916
Italy539 0666.84135.87140.72150.225
Latvia20 3545.7544.5120.89240.3220
Lithuania29 4805.3414.3810.6370.2712
Luxembourg53706.0065.74110.2410.021
Malta38877.23186.48210.5420.213
The Netherlands178 5337.38196.31190.74180.268
Northern Ireland22 3626.59105.4590.74170.3722
Norway57 1117.09176.05170.71130.2917
Poland356 6516.82125.73100.73160.3119
Portugal109 1476.75115.82120.6590.2813
Scotland52 8177.65216.43200.84220.3521
Slovak Republic52 3886.2985.3570.67110.247
Slovenia17 8466.98155.88150.77200.2915
Spain402 5908.04237.20250.6150.214
Sweden100 0866.3195.3980.6360.2610
USA**3 550 02610.75279.02271.01280.5028
Correlation coefficient *** Ref 0.99 0.61 0.53 

Preterm birth rates <37 weeks of gestation were highly correlated with preterm birth rates between 32 and 36 weeks (correlation coefficient 0.99). A lesser degree of correlation was observed between overall preterm birth rates <37 weeks and preterm birth rates between 28 and 31 weeks and 24 and 27 weeks of gestation (correlation coefficients 0.61 and 0.53, respectively, Table 1). These differences in correlation were expected, as the majority of preterm births at <37 weeks occurred at 32–36 weeks of gestation. Mild/moderate preterm birth rates ranged from 4.4 to 10.0 per 100 live births, while live births at 24–31 weeks only constituted about 1.5% of all live births.

Stillbirth and neonatal death rates at ≥32 and ≥37 weeks of gestation are presented in Table 2. Countries with relatively low mild/moderate preterm birth rates (32–36 weeks of gestation) had on average higher rates of stillbirth at 32 weeks of gestation or later (correlation coefficient −0.30, weighted correlation coefficient −0.72; Figure 1A), and at 37 weeks of gestation or later (correlation coefficient −0.35, weighted correlation coefficient −0.72, Figure 1B). A similar inverse correlation was observed between mild/moderate preterm birth and neonatal death at ≥32 and ≥37 weeks of gestation (correlation coefficients −0.41 and −0.43, weighted correlation coefficients −0.19 and −0.29, respectively, Figure 2A,B).

Table 2. Stillbirth and neonatal death rates at ≥32 and ≥37 weeks of gestation, Europe and North America, 2004
CountryStillbirth rate (per 1000 total births)Neonatal death rate (per 1000 live births)
≥32 weeks≥37 weeks≥32 weeks≥37 weeks
Austria1.911.031.030.69
Belgium2.281.071.170.73
Canada2.231.341.070.72
Czech Republic1.961.200.860.54
Denmark3.152.382.021.63
England and Wales3.222.001.270.95
Estonia2.612.062.392.06
Finland1.690.810.980.74
Germany1.941.120.400.24
Greece3.041.67n/an/a
Hungary2.841.28n/an/a
Ireland3.271.961.250.84
Italy3.022.10n/an/a
Latvia3.832.403.282.71
Lithuania3.051.932.361.86
Luxembourg1.310.591.120.59
Malta3.111.942.331.39
Netherlands3.332.111.531.12
Northern Ireland2.761.481.220.77
Norway2.571.751.030.81
Poland3.001.662.101.23
Portugal2.031.001.100.75
Scotland3.642.091.090.84
Slovak Republic1.140.531.140.53
Slovenia2.611.391.080.66
Spain2.111.18n/an/a
Sweden2.741.921.310.87
USA2.011.081.270.80
image

Figure 1.  Correlation between stillbirth rates at ≥32 weeks of gestation (A) and ≥37 weeks of gestation (B) and preterm birth rates (32–36 weeks of gestation), Europe and North America, 2004. Each circle represents a country with size proportionate to the number of total births.

Download figure to PowerPoint

image

Figure 2.  Correlation between neonatal death rates at ≥32 weeks of gestation (A) and ≥37 weeks of gestation (B) and preterm birth rates (32–36 weeks of gestation), Europe and North America, 2004. Each circle represents a country with size proportionate to the number of live births.

Download figure to PowerPoint

Random-intercept Poisson regression models revealed that a 1% increase in the mild/moderate preterm birth rate corresponded to a 9% (95% CI 8–10) lower stillbirth rate at ≥32 weeks of gestation, and a 14% (95% CI 13–15) lower stillbirth rate at ≥37 weeks of gestation (Table 3). A weaker association was observed between the preterm birth rates and neonatal death rates; a 1% increase in the preterm birth rate between 32 and 36 weeks was associated with a 3% (95% CI 1–4) decline in neonatal deaths at ≥32 weeks of gestation and a 7% (95% CI 5–8) decline in neonatal death rates at ≥37 weeks of gestation. Adjustment for older maternal age, parity and multiple gestations did not substantially alter the association between preterm birth and stillbirth. However, the association between preterm birth and stillbirth became weaker; a 1% increase in the preterm birth rate between 32 and 36 weeks was associated with a 6% (95% CI 4–8) decline in stillbirth rates at ≥32 and a 12% (95% CI 9–15) decline in stillbirth at ≥37 weeks (Table 3). The association with neonatal death became stronger after the adjustment: a 1% increase in the preterm birth rate between 32 and 36 weeks was associated with a 12% (95% CI 9–15) decline in neonatal death rates at ≥32 weeks and an 18% (95% CI 14–22) decline in neonatal deaths at ≥37 weeks of gestation (Table 3).

Table 3. Crude and adjusted rate ratios and 95% confidence intervals expressing the change in stillbirth rates and neonatal death rates for a 1% increase in preterm birth rates between 32–36 weeks, Canada, Europe and the United States, 2004
OutcomeCrude rate ratio95% Confidence intervalsAdjusted rate ratio*95% Confidence intervals
  1. Crude rate ratios and adjusted rate ratios were obtained from random-intercept Poisson regression models

  2. *Adjusted for multiple births, older maternal age (≥35 years of age), and primiparity.

  3. Analysis of stillbirth based on data from Austria, Belgium, Canada, the Czech Republic, Denmark, England and Wales, Estonia, Finland, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, the Netherlands, Northern Ireland, Norway, Poland, Portugal, Scotland, Slovak Republic, Slovenia, Spain Sweden and USA. Analyses of neonatal deaths included the same countries except for Spain, Hungary, Italy and Greece.

Stillbirth rate
≥32 weeks0.910.900.920.940.920.96
≥37 weeks0.860.850.870.880.850.91
Neonatal death rate
≥32 weeks0.970.960.990.880.850.91
≥37 weeks0.930.920.950.820.780.86

Sensitivity analyses examining the effect of missing gestational age (and exclusion of countries with missing gestational age for more than 1% of live births) slightly weakened the association between preterm birth rate and stillbirth and neonatal death rates. For instance, a 1% increase in mild/moderate preterm birth was associated with a 3% (95% CI 1–5) decline in stillbirths at ≥32 weeks of gestation and with an 8% (95% CI 5–11) decline in stillbirth at ≥37 weeks of gestation when all infants with missing gestational age were assumed to be ≥32 and ≥37 weeks of gestation, respectively. Sensitivity analyses excluding the USA yielded attenuated results, although the direction of the relationships and the statistical significance remained unchanged. Analyses contrasting preterm birth and gestational age-specific perinatal death rates between Canada and England, Wales and Northern Ireland showed that preterm birth rates for <37 weeks of gestation in both countries were similar (7.9 per 100 live births in Canada and 7.6 per 100 live births in England, Wales and Northern Ireland). Crude rates of stillbirth (5.8 per 1000 total births in Canada and 5.7 per 1000 total births in England, Wales and Northern Ireland) and neonatal death (3.6 and 3.7 per 1000 live births, respectively) were also similar. However, the preterm birth rate between 32 and 36 weeks of gestation was significantly higher in Canada (6.7 versus 6.2 per 100 live births, P < 0.001), and the very preterm birth rate was significantly lower (1.2 versus 1.4 per 100 live births, P < 0.001) in Canada compared with England, Wales and Northern Ireland (Figure 3A). Both stillbirth and neonatal death rates at ≥32 weeks were considerably lower in Canada compared with England, Wales and Northern Ireland (stillbirth rates 2.4 versus 3.2 per 1000 total births, respectively, P < 0.001; neonatal death rates 1.0 versus 1.3 per 1000 live births, respectively, P < 0.001, Figure 3B).

image

Figure 3.  Preterm birth rates at <32 and 32–36 weeks of gestation (A) and stillbirth and neonatal death rates at ≥32 weeks of gestation (B), Canada (2003) and England, Wales and Northern Ireland (2004).

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. Disclosure of interests
  10. Contribution to authorship
  11. Details of ethics approval
  12. Funding
  13. References
  14. Commentary on ‘The ecological design’

Our study showed substantial variation in preterm birth rates among industrialised countries in Europe and North America. The majority of preterm births occurred at 32–36 weeks of gestation and countries with relatively high mild/moderate preterm birth rates had on average lower stillbirth and neonatal death rates at 32 weeks of gestation or later. This significant inverse association persisted after adjustment for multiple births, older maternal age and primiparity.

Changes in obstetric practice during the last two decades have profoundly influenced preterm birth rates. The increase in iatrogenic early delivery (i.e. following labour induction and/or caesarean delivery) carried out for fetal or maternal indications has been the primary reason for the upward temporal trend in preterm birth.7,8,11,18 This trend has been paralleled by a temporal decline in stillbirth and neonatal death rates, providing evidence that these obstetric interventions are saving compromised fetuses that would have otherwise died in utero or during the neonatal period.11,16 Our study adds to this evidence by demonstrating that countries with higher rates of mild/moderate preterm birth had on average lower rates of stillbirth and neonatal death at 32 weeks of gestation and beyond. It should be noted, however, that the implications of our study are restricted to medically indicated early delivery. Modern obstetrics is predicated on selective, carefully timed labour induction or caesarean delivery given a fetal or medical indication, with a balancing of risks and benefits associated with early delivery versus expectant management.

Improvements in neonatal care during the last few decades have favoured iatrogenic delivery of the compromised fetus by reducing the neonatal risks of preterm birth, especially close to term. The temporal increase in iatrogenic preterm birth has occurred predominantly at late preterm gestation (34–36 weeks). For example, in the USA, late preterm births increased by 20% from 6.8 to 8.1 per 100 live births between 1990 and 2007, with most of this increase occurring at 36 weeks of gestation (from 3.4 to 4.4 per 100 births).17 Whereas a high rate of preterm birth has been traditionally recognised as an indicator of poor perinatal health in any population, this increase in medically indicated preterm birth for fetal compromise has added a new dimension to the preterm birth index. Contemporary rates of preterm birth also appear to indicate access and availability of obstetric intervention. Whereas high rates of extremely preterm birth and very preterm birth remain indicators of poor population perinatal health, high rates of preterm birth at late preterm gestation may be indicative of the availability of high-quality obstetric and neonatal care including careful fetal surveillance and prompt obstetric intervention where indicated. The overall preterm birth rate before 37 weeks of gestation has therefore become a less meaningful measure of fetal–infant and maternal health in recent years.

Our study has some limitations, including the use of data from publicly available reports based on national vital statistics and birth registries. European data were available in aggregate form (gestational age 24–27, 28–31 and 32–36 weeks of gestation)4 and we were therefore limited in our choice of gestational age categories. In most countries, data on gestational age were likely based on the ‘best obstetrical estimate’, which combines clinical and ultrasound data, but some countries may have favoured alternative estimates.4 To avoid differences in the registration of live births and stillbirths at the borderline of viability, we excluded live births at <24 weeks of gestation. Preterm births at <24 weeks of gestation comprise only a minimal fraction of live births (0.2% or less) and their exclusion is likely to have had a negligible effect on preterm birth rates. Although between-country differences in birth registration35 may have affected our study to a small extent, a substantial bias through under-reporting of stillbirth and neonatal death at ≥32 and ≥37 weeks of gestation is unlikely. European data were available only for year 2004 and we could not assess temporal trends in preterm birth and stillbirth rates. Similarly, information on iatrogenic preterm birth rates was not available for countries in Europe. It is possible that the prevalence of neonatal morbidity from iatrogenic preterm birth was higher in countries with higher rates of preterm birth. However, we could not quantify the association between preterm birth and neonatal morbidity or the long-term sequelae of preterm birth because of the lack of such information by country.

While the rates of spontaneous preterm birth have remained stable over time in most countries, the rates of iatrogenic delivery at 32–36 weeks of gestation have increased; for example, a 68% increase was observed the USA (from 1.9 to 3.2 per 100 live births between 1995/1996 and 2004/2005);36 and a 44% increase has been observed in Scotland (from 0.9 to 1.3 per 100 singleton live births between 1980–1984 and 2000–2004).9 Even in Sweden, where the low rate of preterm birth has remained unchanged since the 1990s, a small but significant temporal rise in iatrogenic delivery at 34–36 weeks of gestation occurred between 1992 and 2001.37 We therefore assumed that the international differences in preterm birth rates at 32–36 weeks of gestation are most likely to be the result of the differences in iatrogenic intervention at late preterm gestation. The inter-country variation, however, could have also arisen through other factors, including differences in the ascertainment of gestational age. Finally, our analyses were based on an ecological design (preterm birth and stillbirth rates by country) and not individual level analyses. Such a design was necessary to avoid confounding by indication when dealing with the effect of iatrogenic delivery, as the indication for the iatrogenic preterm delivery typically implies a high risk status for the pregnancy. However, it is possible that our study was confounded by between-country differences in factors such as prenatal diagnosis of congenital anomalies and pregnancy termination.

Conclusion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. Disclosure of interests
  10. Contribution to authorship
  11. Details of ethics approval
  12. Funding
  13. References
  14. Commentary on ‘The ecological design’

The increase in preterm birth in the industrialised countries has caused public health concern and calls for preventive measures, including calls for a reduction of late preterm births.15 However, our study demonstrates that countries with higher rates of mild/moderate preterm birth had relatively lower rates of stillbirth and neonatal death at and after 32 weeks of gestation. This inverse association is probably a result of an increase in medically indicated late preterm deliveries carried out to prevent perinatal death. Our study suggests that a high preterm birth rate no longer simply reflects poor perinatal health in a population, but rather represents a heterogeneous index that reflects both poor maternal and fetal health and also availability of high-quality obstetric and neonatal care. Preventive strategies to reduce iatrogenic preterm birth, especially at late preterm gestation, have to carefully consider the underlying obstetric risks and the need for medical intervention among high-risk pregnancies.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. Disclosure of interests
  10. Contribution to authorship
  11. Details of ethics approval
  12. Funding
  13. References
  14. Commentary on ‘The ecological design’

We are grateful to Dr Shiliang Liu, Public Health Agency of Canada, for providing the Canadian data. This study was presented and discussed at the April 2011 meeting of the Fetal and Infant Health Study Group, Canadian Perinatal Surveillance System.

Contribution to authorship

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. Disclosure of interests
  10. Contribution to authorship
  11. Details of ethics approval
  12. Funding
  13. References
  14. Commentary on ‘The ecological design’

All authors contributed to the study conception and design. SL and KSJ contributed to the data analysis. SL drafted the manuscript; KSJ, YS and BB contributed to the composition of the manuscript and approved the final version. YS and BB critically reviewed the manuscript and provided subject-specific advice.

Funding

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. Disclosure of interests
  10. Contribution to authorship
  11. Details of ethics approval
  12. Funding
  13. References
  14. Commentary on ‘The ecological design’

SL is supported by a postdoctoral fellowship award from the Michael Smith Foundation for Health Research (MSFHR) and KSJ’s work is supported by the Child and Family Research Institute and the CIHR Peter Lougheed New Investigator award (PLS-56343).

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. Disclosure of interests
  10. Contribution to authorship
  11. Details of ethics approval
  12. Funding
  13. References
  14. Commentary on ‘The ecological design’
  • 1
    Institute of Medicine. Preterm Birth: Cases Consequences and Prevention. Washington DC: National Academy Press, 2007.
  • 2
    Mattison DR, Damus K, Fiore E, Petrini J, Alter C. Preterm delivery – a public health perspective. Pediatr Perinat Epidemiol2001;15 (Suppl 2):717.
  • 3
    MacDorman MF, Mathews TJ. Behind International Rankings of Infant of Infant Mortality: How the United States Compares with Europe. NCHS data brief No. 23. Hyattsville, MD: CDC, National Center for Health Statistics, 2009.
  • 4
    European Perinatal Health Report. Data from 2004. Euro-PERISTAT project. [http://www.europeristat.com/bm.doc/european-perinatal-health-report.pdf]. Accessed 12 December 2011.
  • 5
    Priest L. Canada’s reputation for low infant mortality takes stunning decline; Once at No. 6 in world ranking, ‘shockingly high’ death rate now puts Canada at No. 24. The Globe and Mail. Toronto. 22 May 2010.
  • 6
    Paneth NS. The problem of low birthweight. Future Child1995;5:1934.
  • 7
    Joseph KS, Kramer MS, Marcoux S, Ohlsson A, Wen SW, Allen A, et al.Determinants of preterm birth rates in Canada from 1981 through 1983 and from 1992 through 1994. N Engl J Med1998;339:14349.
  • 8
    Barros FC, Velez MP. Temporal trends of preterm birth subtypes and neonatal outcomes. Obstet Gynecol2006;107:103541.
  • 9
    Norman JE, Morris C, Chalmers J. The effect of changing patterns of obstetric care in Scotland (1980–2004) on rates of preterm birth and its neonatal consequences: perinatal database study. PLoS Med2009;6:e1000153.
  • 10
    Martin JA, Hamilton BE, Sutton PD, Ventura SJ, Mathews TJ, Osterman MJK. Births: Final Data for 2008. National Vital Statistics Reports. Vol. 59, No. 1. Hyattsville, MD: National Center for Health Statistics, 2010.
  • 11
    Ananth CV, Joseph KS, Oyelese Y, Demissie K, Vintzileos AM. Trends in preterm birth and perinatal mortality among singletons: United States, 1989 through 2000. Obstet Gynecol2005;105:108491.
  • 12
    Langhoff-Roos J, Kesmodel U, Jacobsson B, Rasmussen S, Vogel I. Spontaneous preterm delivery in primiparous women at low risk in Denmark: population based study. BMJ2006;332:9379.
  • 13
    Preterm birth: crisis and opportunity. Lancet2006;368:339.
  • 14
    Park A. Why the U.S. gets a D on preterm birth rates. Time, in partnership with CNN: Heath and Science November 2009. [http://www.time.com/time/health/article/0,8599,1940140,00.html]. Accessed 12 December 2011.
  • 15
    Healthy People 2020. Healthy People 2020, maternal infant and child health. Summary of objectives. [http://healthypeople.gov/2020/topicsobjectives2020/pdfs/MaternalChildHealth.pdf]. Accessed 12 December 2011.
  • 16
    Joseph KS, Demissie K, Kramer MS. Trends in obstetric intervention, stillbirth and preterm birth. Semin Perinatol2002;26:2509.
  • 17
    Martin JA, Kirmeyer S, Osterman M, Shepherd RA. Born a bit too early: recent trends in late preterm births. NCHS Data Brief2009;24:18.
  • 18
    Bettegowda VR, Dias T, Davidoff MJ, Damus K, Callaghan WM, Petrini JR. The relationship between caesarean delivery and gestational age among US singleton births. Clin Perinatol2008;35:30923.
  • 19
    Tough SC, Newburn-Cook C, Johnston DW, Svenson LW, Rose S, Belik J. Delayed childbearing and its impact on population rate changes in lower birth weight, multiple birth, and preterm delivery. Pediatrics2002;109:399403.
  • 20
    Fell DB, Joseph KS, Dodds L, Allen AC, Jangaard K, Van den HM. Changes in maternal characteristics in Nova Scotia, Canada from 1988 to 2001. Can J Public Health2005;96:2348.
  • 21
    Blondel B, Kaminski M. Trends in the occurrence, determinants, and consequences of multiple births. Semin Perinatol2002;26:23949.
  • 22
    Blickstein I, Keith LG. Aging, twinning, and perinatal outcomes. Fertil Steril2003;79:6612.
  • 23
    Kramer MS, Demissie K, Yang H, Platt RW, Sauvé R, Liston R. The contribution of mild and moderate preterm birth to infant mortality. Fetal and Infant Health Study Group of the Canadian Perinatal Surveillance System. JAMA2000;16:8439.
  • 24
    Consortium on Safe Labor, Hibbard JU, Wilkins I, Sun L, Gregory K, Haberman S, et al.Respiratory morbidity in late preterm births. JAMA2010;304:41925.
  • 25
    Leone A, Ersfeld P, Adams M, Meyer Schiffer P, Bucher H, Arlettaz R. Neonatal morbidity in singleton late preterm infants compared with full-term infants. Acta Paediatr2012;101:e610.
  • 26
    Mohan SS, Jain L. Late preterm birth: preventable prematurity?Clin Perinatol2011;38:54755.
  • 27
    Cheng Y, Kaimal A, Bruckner T, Hallaron D, Caughey A. Perinatal morbidity associated with late preterm deliveries compared with deliveries between 37 and 40 weeks of gestation. BJOG2011;118:144654.
  • 28
    Teune MJ, Bakhuizen S, Bannerman CG, Opmeer BC, van Kaam AH, van Wassenaer AG, et al.A systematic review of severe morbidity in infants born late preterm. Am J Obstet Gynecol2011;205:e19.
  • 29
    Public Health Agency of Canada. Canadian Perinatal Health Report, 2008 Edition. Ottawa, ON: Public Health Agency of Canada, 2008.
  • 30
    Centers for Disease Control and Prevention. Vital Statistics Data available online. [http://www.cdc.gov/nchs/data_access/Vitalstatsonline.htm]. Accessed 30 March 2011.
  • 31
    Mustafa G, David RJ. Comparative accuracy of clinical estimate versus menstrual gestational age in computerized birth certificates. Public Health Rep2001;116:1521.
  • 32
    Ananth CV. Menstrual versus clinical estimate of gestational age dating in the United States: temporal trends and variability in indices of perinatal outcomes. Paediatr Perinat Epidemiol2007;21 (Suppl 2):2230.
  • 33
    Joseph KS, Huang L, Liu S, Ananth CV, Allen AC, Sauve R, et al.Reconciling the high rates of preterm and postterm birth in the United States. Obstet Gynecol2007;109:81322.
  • 34
    Confidential Enquiry into Maternal and Child Health. Perinatal Mortality, 2005: England, Wales and Northern Ireland. London: CEMACH, 2007.
  • 35
    Joseph KS, Liu S, Rouleau J, Lisonkova S, Hutcheon JA, Sauve R, et al.International comparisons of perinatal and infant mortality: influence of definition-based versus pragmatic birth registration. BMJ2012;344:e746, doi:10.1136/bmj.e746 .
  • 36
    Lisonkova S, Hutcheon JA, Joseph KS. Temporal trends in neonatal outcomes following iatrogenic preterm delivery. BMC Pregnancy Childbirth2011;11:39.
  • 37
    Morken NH, Kallen K, Hagberg H, Jacobsson B. Preterm birth in Sweden 1973–2001: rate, subgroups, and effect of changing patterns in multiple births, maternal age, and smoking. Acta Obstet Gynecol Scand2005;84:55865.

Commentary on ‘The ecological design’

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. Disclosure of interests
  10. Contribution to authorship
  11. Details of ethics approval
  12. Funding
  13. References
  14. Commentary on ‘The ecological design’

Lisonkova et al. show an association between a country’s preterm birth rate, stillbirth rate and neonatal death rate based on country-level data. This type of study, which presents associations between country-level data rather than individual-level observations, is an ecological study. When the research goal is to make inferences about individuals, ecological studies—which do not link observations at the individual-level—are prone to special biases not present in individual-level (e.g. case–control or observational cohort) studies of the same populations (Robinson, Am Soc Rev 1950;15:351–7; Goodman, Am Soc Rev 1953;18:663–4; Greenland et al., Am J Epidemiol 1994;139:747–60; Freedman, Encyclopedia of Social Science Research Methods, Los Angeles, CA, USA: Sage Publications; 2004, p. 293). The ecological fallacy occurs when one falsely assumes that relationships at the group level hold at the individual level.

Robinson was perhaps the first to outline the limitations of using ecological data to infer individual-level relations. He presented research on state-level data from each of the 48 US states in 1930. Comparing the percentage of the population who are literate (in English) to the percentage of the population who are foreign-born, he found a positive correlation of 0.53. This indicated that states with more foreign-born residents had higher literacy rates, which was true. The ecological fallacy occurs when one draws the conclusion that foreign-born individuals are more likely to be literate in English; at the individual level, the correlation was −0.11, with individuals born outside the USA less likely to be literate in English. The ecological correlation of 0.53 occurred because foreign-born individuals were more likely to immigrate to states in which native-born citizens were highly literate.

Ecological studies are practically appealing (Morgenstern, Modern Epidemiology. 3rd edn. New York: Wolters Kluwer/Lippincott Williams & Wilkins, 2008; p.511–31) and can sometimes yield important insights about individuals (Greenland, Int J Epidemiol 2001;30:1343–50). Furthermore, some (Wen et al., J Clin Epidemiol 1999;52:7–12) have suggested that ecological studies may minimise confounding by indication, which occurs when the treatment under study is preferentially allocated to those who need it most and who are therefore at a higher risk of the outcome (Joffe, Pharmacoepidemiol Drug Safety 2000;9:37–41). However, in the absence of individual-level data, determining whether aggregate-level associations accurately reflect individual-level relations is notoriously difficult (Greenland et al., Am J Epidemiol 1994;139:747–60). The assumptions required for ecological studies to minimise confounding by indication are often heroic, and rarely justified (Joffe, Pharmacoepidemiol Drug Safety 2000;9:37–41; Naylor, J Clin Epidemiol 1999;52:1–5). Even if the assumptions needed to minimise confounding by indication are met, confounding in general is more difficult to control in an ecological study (Morgenstern, Modern Epidemiology 3rd edn, New York: Wolters Kluwer/Lippincott Williams & Wilkins; 2008, p.511–31). Hence, confirmation of individual-level trends identified using ecological data alone would, at the very least, require additional research based on individual-level data. Ideally, such inferences should be confirmed using multi-level data that combine the best of both worlds (Greenland, Int J Epidemiol 2001;30:1343–50; Diez-Roux, Am J Public Health, 1998;88:216–22; Subramanian et al., Int J Epidemiol 2009;38:342–60). When making decisions about individuals, ecological results should be viewed with caution until such confirmation is made.

Disclosure of interest

The authors have no conflicts of interest to disclose.

AH Herring,a AI Naimib

aDepartment of Biostatistics and Carolina Population Center and bDepartment of Epidemiology, The University of North Carolina, Chapel Hill, NC, USA