Phone-based followed by onsite verbal autopsy for perinatal deaths in rural North India: Feasibility and concordance
Neelanja Chaturvedi, Ankit Raina, Minakshi Sharma, Premananda Mohanty, Anuj Kumar Pandey, Sutapa B. Neogi, Deepak ChawlaAbstract
Background:
Globally, 2.4 million neonatal deaths (NDs) occur annually within 1 st month of life, with India bearing a significant proportion. Perinatal deaths in India are common, yet cause-of-death data are sparse in rural areas. Verbal autopsy (VA), a World Health Organization-endorsed tool, is widely used, but the comparative feasibility of phone versus onsite VA is unclear. Reliable mortality data are crucial for determining causes of deaths and to reduce perinatal mortality. The aim of this study was to compare the feasibility and concordance of key indicators between paired phone and onsite VAs for stillbirths (SBs) and NDs.
Material and Methods:
This prospective study evaluated the feasibility and data concordance of phone-based versus onsite VAs. Deaths were reported by Accredited Social Health Activists to Foundation for Survival of Women and Children in collaboration with the government. Data measures included maternal demographics, pregnancy and birth details, causes of death, and adverse outcomes. Health service utilization, maternal factors, and interview process metrics were also captured. In a 3 months period, 70 SBs and 64 NDs were investigated. Phone interviews showed agreement with onsite data for delivery type, baby’s sex, tetanus toxoid completion, ultrasound, and caste.
Results:
Agreement was lower for >3 antenatal coverage visits and gestation age on phone interviews. Most reported causes of SB were prematurity (23 phone; 22 onsite) and fetal distress (18 phone; 17 onsite). For ND, concordance was observed for asphyxia (27 each) and low birth weight (7 each). However, Phone VA required more contact attempts (190 calls for 70 SB and 204 calls for 64 ND) and completion days. Interview time was shorter by phone, but transcription took longer; total effort was lower on phone than onsite (SB: 8972 vs. 10,528 min; ND 9798 vs. 10,515 min).
Conclusion:
In this rural program, phone VA was feasible but had markedly lower response, higher completion days and higher transcription effort than onsite VA; concordance for several key indicators was acceptable. The data show good concordance but low feasibility in terms of reaching respondents and completion of interviews. A hybrid model – phone screening plus selective onsite validation – may optimize effort and data quality. VAs can be used for improving mortality surveillance within health systems while reducing travel time and costs in rural areas.