Accuracy of WHO Verbal Autopsy Tool in Determining Major Causes of Neonatal Deaths in India

OBJECTIVES: This study was conducted to evaluate the performance of World Health Organisation (WHO) verbal autopsy tool in determining major causes of neonatal deaths. METHODS: From a tertiary care hospital and a government multispecialty hospital, the attending paediatricians ascertained a clinical cause of death for 371 neonatal deaths. Trained field workers conducted verbal autopsy (VA) interviews. Two independent paediatricians, who had no access to the clinical information, assigned cause of death as per verbal autopsy. Analysis was based on 313 cases in which both clinical diagnosis and VA diagnosis was obtained. FINDINGS: As per the clinical diagnosis, four most common causes of neonatal deaths were sepsis (29.1%), preterm birth (27.8%), birth asphyxia (27.2%), and congenital anomalies (11.5%). Cause specific mortality fractions by VA diagnosis were statistically similar to those obtained by clinical diagnosis except for birth asphyxia (16.3%). Diagnostic accuracy of verbal autopsy diagnosis against clinical diagnosis ranged from 78% to 92% in ascertaining different underlying causes of death. Area under the Receiver-Operator Characteristics curve (95% confidence interval) was 0.75 (0.69-0.80) for sepsis, 0.74 (0.68-0.80) for preterm birth, 0.73 (0.65-0.82) for congenital anomaly and 0.70 (0.64-0.75) for birth asphyxia. Kappa for all four causes was moderate (0.46-0.55). INTERPRETATION: The WHO verbal autopsy tools can provide reasonably good estimates of predominant causes of neonatal deaths in countries where neonatal mortality is high. Caution is required to interpret cause specific mortality fraction (CSMF) for birth asphyxia by VA because it is likely to be an underestimate.