Skip to main content

Table 6 Comparing population cause distributions estimated from facility deaths to observed population cause distributions in the study countries

From: Using health facility deaths to estimate population causes of neonatal and child mortality in four African countries

Age group
Analysis method
CountryCSMF accuracy of population causes of death estimated from facility deaths compared to observed population causes
N deaths% facility deathsProjection A* Estimate (95% CI)Projection B** Estimate (95% CI)Projection E*** Estimate (95% CI)
Neonates
EAVACameroon16434%0.85 (0.76, 0.90)0.84 (0.72, 0.89)0.87 (0.75, 0.91)
Malawi32055%0.89 (0.84, 0.94)0.91 (0.85, 0.94)0.93 (0.85, 0.95)
Niger45319%0.79 (0.67, 0.86)0.75 (0.59, 0.84)0.84 (0.66, 0.88)
Nigeria72227%0.91 (0.83, 0.93)0.92 (0.82, 0.93)0.93 (0.79, 0.93)
Average  0.860.860.89
PCVACameroon16434%0.87 (0.73, 0.90)0.87 (0.74, 0.92)0.89 (0.74, 0.92)
Malawi32055%0.94 (0.88, 0.96)0.95 (0.89, 0.97)0.94 (0.88, 0.96)
Niger45319%0.83 (0.70, 0.89)0.87 (0.69, 0.91)0.83 (0.67, 0.87)
Nigeria72227%0.86 (0.77, 0.90)0.88 (0.76, 0.93)0.85 (0.74, 0.90)
Average  0.880.890.88
Children
EAVACameroon63529%0.92 (0.85, 0.94)0.91 (0.84, 0.94)0.91 (0.85, 0.93)
Malawi69150%0.94 (0.90, 0.96)0.95 (0.90, 0.96)0.94 (0.89, 0.95)
Niger61919%0.90 (0.77, 0.92)0.90 (0.40, 0.92)0.92 (0.80, 0.93)
Nigeria205522%0.95 (0.89, 0.96)0.94 (0.88, 0.95)0.95 (0.89, 0.96)
Average  0.930.930.93
PCVACameroon63529%0.91 (0.83, 0.93)0.90 (0.83, 0.93)0.91 (0.83, 0.93)
Malawi69150%0.92 (0.88, 0.94)0.92 (0.88, 0.94)0.92 (0.87, 0.94)
Niger61919%0.94 (0.81, 0.94)0.92 (0.79, 0.94)0.92 (0.81, 0.94)
Nigeria205522%0.96 (0.89, 0.96)0.97 (0.89, 0.97)0.95 (0.89, 0.96)
Average  0.930.930.93
  1. Observed and estimated population cause distributions are from verbal autopsies
  2. CSMF cause-specific mortality fraction, EAVA expert algorithm verbal autopsy, PCVA physician-coded verbal autopsy
  3. *Population causes estimated by substituting observed causes from facility deaths for community deaths
  4. **Population causes estimated by predicting community causes with multinomial logistic regression of facility deaths based on age at death, mother’s education, and receiving any ANC (only for neonates, yes/no)
  5. ***Population causes estimated with random forest, using same predictors as multinomial projection B, plus birthplace