Methodology to Determine Cause of Death for Stillbirths and Neonatal Deaths Using Automated Case Reports and a Cause-of-Death Panel

Background: Review of data from multiple sources is often necessary to determine cause of death for stillbirths and neonatal deaths, especially in low- to middle-income countries (LMICs) where available data may vary. The minimally invasive tissue sampling (MITS) procedure provides granular histologic and microbiologic data that clinical reports and verbal autopsies cannot provide. Expert panel evaluation of data from individual deaths can be resource-intensive but remains essential to accurately infer causes of death. Methods: The Project to Understand and Research Preterms and Stillbirths in South Asia (PURPOSe) study uses review panels to evaluate causes of death in 2 LMICs. To make the process manageable, a subset of the study variables was selected with professional input and organized into case reports. Case reports include clinical information, laboratory results, fetal or neonatal organ histology and polymerase chain reaction results from tissue obtained by MITS. Panelists evaluated the complete case report forms and then determined the cause of death based on available data. Results: Computerized case reports averaged 2 to 3 pages. Approximately 6 to 8 cases were reviewed and discussed per 1-hour panel meeting. All panelists were provided the same information; missing data were noted. This limited bias between panelists and across meetings. Study teams notably took ownership of data quality. Conclusions: Standardized case reports for cause-of-death determination panel evaluation improve the efficiency of the review process, clarify available information, and limit bias across panelists, time, and location.

Automated summary

The Project to Understand and Research Preterms and Stillbirths in South Asia (PURPOSe) study utilizes review panels to determine causes of death in LMICs, with computerized case reports averaging 2 to 3 pages and 6 to 8 cases reviewed per 1-hour meeting, reducing bias across panelists and meetings.