Our results suggest that stillbirth prevalence at the community level is in general less than 1% in more developed parts of the world and could exceed 3% in less developed regions, but we were not able to provide overall estimates of stillbirth prevalence for different regions of the world due to significant heterogeneity across sub-regions. Facility based studies show higher rates, which could be due to referral bias.
Meta-regression analysis explained a considerable proportion (52%) of the observed heterogeneity in these data. Not surprisingly, development status of the setting in which the study was conducted was shown to be a strong predictor of stillbirth prevalence. Perhaps less expected was that the quality of a study is another significant predictor, independent of development status, with prevalence rates being lower in studies of higher quality. All other study-level variables we tested for possible influence on stillbirth rates did not show a significant relationship. The remaining variation could be due to other unmeasured variables that could not be investigated in this analysis. For example, with the information available to us, we could not investigate the influence of characteristics such as age and parity, both of which are important predictors of stillbirth [96-98].
Meta-analytical methods including meta-regression has increasingly been used in summarizing outcomes and explaining between-study variability in investigations of treatment effects or associations [99-101], but its use in prevalence studies is relatively infrequent, with existing literature largely limited to the area of mental health [102,103].
The meta-regression techniques were helpful in explaining a significant portion of the observed variation in stillbirth rates. We believe it is timely to use this approach more widely in the estimation of maternal and perinatal health indicators associated with internationally set goals and targets. The need for global estimates of such indicators is greater than ever in the context of international development goals including the Millennium Development Goals (MDGs) . More empirical evidence should improve the selection, implementation and interpretation of indicators used to monitor the progress towards achievement of the MDGs as well as addressing the increased demand for reliable estimates.
The empirical evidence we provide regarding the significant influence of the development status of the study setting on stillbirth prevalence has implications for policy and programmatic actions. The significantly higher rates in less developed country settings and the highest rate observed in Western Africa could largely be due to inadequacies in accessing appropriate maternal health care during both antenatal period and delivery. The reported skilled attendance at birth in this region is also very low, corroborating these findings .
The independent effect of the quality of primary studies on the rates deserves attention as well. For effect-size studies the perceived quality of a published article is known to be related to its likelihood of being included in a meta-analysis , although the extent to which this is also true for prevalence studies is less well established. It has also been demonstrated that reporting of observational studies including cross-sectional designs are not in accordance with the desirable standards . Our findings contribute to this literature by demonstrating the influence of quality on the outcome of a prevalence study. More carefully conducted and reported studies are needed if researchers want their findings to be useful for the scientific community as well as to have an influence on policy decisions.
Our study has several limitations. First, our analysis focuses on a subgroup of studies selected from a larger systematic review. The search strategy for the larger review, however comprehensive, did not specifically target stillbirths, and therefore, some relevant studies may have been missed. The trade-off in deciding to limit our investigation to prevalence studies having one-year duration reduced the number of studies included in the analysis. We took this decision because the durations of the remaining studies varied widely and studies of longer duration may have counted multiple pregnancies in the same woman. Since stillbirth may be a recurrent event [108-110], we aimed to avoid including repeating stillbirths in the analysis. Finally, as discussed above, we demonstrated that two important variables influence the stillbirth rates, but the influence of other factors, particularly those measured at the individual level, remains to be investigated.
The results of this systematic review show significant variation in stillbirth rates in different parts of the world and that, even in the settings with the highest standards of maternal and perinatal care, around five out of 1000 newborns will be stillborns.
Implications for policy and practice
While these findings do not have direct implications for clinical practice, they highlight the relative frequency of stillbirth as an indicator of the quality of service delivery. Even in developed countries the fact that stillbirths constitute close to 1% of all births should alert policy-makers to initiate audit procedures to identify avoidable cases and take action.
Implications for research
We urge epidemiology community to address the methodological standards as well as reporting of prevalence studies. The application of meta-analytical techniques including meta-regression in summarizing prevalence rates needs further research. The standards for data collection and reporting should be addressed through international consensus.