The effect of health-facility admission and skilled birth attendant coverage on maternal survival in India: a case-control analysis

Abstract

Background: Research in areas of low skilled attendant coverage found that maternal mortality is paradoxically higher in women who seek obstetric care. We estimated the effect of health-facility admission on maternal survival, and how this effect varies with skilled attendant coverage across India.

Methods/findings: Using unmatched population-based case-control analysis of national datasets, we compared the effect of health-facility admission at any time (antenatal, intrapartum, postpartum) on maternal deaths (cases) to women reporting pregnancies (controls). Probability of maternal death decreased with increasing skilled attendant coverage, among both women who were and were not admitted to a health-facility, however, the risk of death among women who were admitted was higher (at 50% coverage, OR = 2.32, 95% confidence interval 1.85-2.92) than among those women who were not; while at higher levels of coverage, the effect of health-facility admission was attenuated. In a secondary analysis, the probability of maternal death decreased with increasing coverage among both women admitted for delivery or delivered at home but there was no effect of admission for delivery on mortality risk (50% coverage, OR = 1.0, 0.80-1.25), suggesting that poor quality of obstetric care may have attenuated the benefits of facility-based care. Subpopulation analysis of obstetric hemorrhage cases and report of 'excessive bleeding' in controls showed that the probability of maternal death decreased with increasing skilled attendant coverage; but the effect of health-facility admission was attenuated (at 50% coverage, OR = 1.47, 0.95-1.79), suggesting that some of the effect in the main model can be explained by women arriving at facility with complications underway. Finally, highest risk associated with health-facility admission was clustered in women with education ≤ 8 years.

Conclusions: The effect of health-facility admission did vary by skilled attendant coverage, and this effect appears to be driven partially by reverse causality; however, inequitable access to and possibly poor quality of healthcare for primary and emergency services appears to play a role in maternal survival as well.

Figure 1. Flow diagram of sample population from national Indian datasets: MDS - Million Death Study, DLHS-2 - second round District Level Health Survey .

Figure 2. Conceptual framework of factors associated with maternal survival.

Figure 3. Predicted probability of maternal death by % skilled attendant coverage over health-facility admission.

Datasouce: Indian MDS 2001–2003 and DLHS-2. Regression models adjusted for: fixed effects - receipt of antenatal care, age, , education, place of residence (rural/urban), district level standard of living, and interaction between health-facility admission and skilled attendant coverage; random effects - district cluster, state cluster. Presented with 95% CI. Women (n = 139 321) in districts (n = 593) in states (n = 35); random effects , , , .

Figure 4. Predicted probability of maternal death by % skilled attendant coverage over (4a) delivery in health-facility and (4b) health-facility admission in sub-population of obstetric hemorrhage.

Datasouce: Indian MDS 2001–2003 and DLHS-2. Adjusted for receipt of antenatal care, age, , education, place of residence (rural/urban), district level standard of living, and interaction between skilled attendant coverage and (4a) delivery in a health-facility or (4b) health-facility admission. Presented with 95% CI. (4a) Women (n = 139 417) in districts (n = 593) in states (n = 35); random effects (4b) Women (n = 17 391) in districts (n = 593) in states (n = 35); random effects , , , .

Figure 5. Odds ratio of maternal death by exposure at 50% skilled attendant coverage.

Datasource: MDS and DLHS-2 2001–2003. All regression models adjusted for: fixed effects - receipt of antenatal care, age, , education, place of residence (rural/urban), district level standard of living, and interaction between main exposure and skilled attendant coverage; random effects - district cluster, state cluster. (1) Main exposure: health-facility admission (1a–1d - linear combinations of coefficients after estimation of this regression model) (2) Main exposure: delivery in health-facility (Routine) (3) Main exposure: health-facility admission; main outcome: hemorrhage in subpopulation.