Verbal autopsy for stillbirth and neonatal deaths – comparing population cause specific mortality fraction using two methods

摘要

Background: Every year 3.2 million infants are stillborn and 3.6 million die within the first month. Up to 98% of these deaths occur in countries with inadequate or non-existent vital registration systems, where cause of death data are sparse and mostly derived from verbal autopsies (VA). It has been advocated that VA are included in routine national statistics. This thesis proposes and compares the strengths and limitations of methodologies to collect and interpret VA data for stillbirths and neonatal deaths. Methods: Data were derived from three research areas in Malawi, Nepal and Mumbai. The development of classifications, diagnostic algorithms and questionnaires for VA, suitable for physician review interpretation is described. A probabilistic method to analyse all age deaths (InterVA) was adapted for stillbirths and neonatal deaths. Cause specific mortality fractions were compared using physicians’ review and InterVA. Results: Neonatal mortality rate in Malawi was 25/1000 livebirths (LB), in Nepal 31/1000 LB and in Mumbai 16/1000 LB. A total of 922 VA including both live and stillbirths were analysed to establish causes of death. Stillbirths accounted for 44-54% of deaths. Of neonatal deaths, in Malawi the majority were attributed to severe infections according to physician review (55%) and InterVA (46%); in Nepal (43%) and Mumbai (61%) perinatal asphyxia was most common according to InterVA. In Nepal however, physician review ascribed the majority of neonatal deaths to severe infections (50%). Kappa statistics for individual agreement comparing both methods was 0.60 (CI 0.567-0.702) in Malawi, 0.62(CI 0.59- 0.65) in Nepal and 0.48(0.40 - 0.50) in Mumbai.udDiscussion: Different VA interpretation methods exist, however standardised procedures are necessary for international comparison. The role of physicianudreview in interpreting VA is changing while computerised methods are becoming more widespread. The modified InterVA model provides a rapid and consistent method to establish causes of stillbirths and neonatal deaths, however it requires further refinements and ultimately a validation study using a comparison other than physician review.