Validation Study of Verbal Autopsy Methods
Author Information
Author(s): Murray Christopher JL, Lopez Alan D, Black Robert, Ahuja Ramesh, Ali Said Mohd, Baqui Abdullah, Dandona Lalit, Dantzer Emily, Das Vinita, Dhingra Usha, Dutta Arup, Fawzi Wafaie, Flaxman Abraham D, Gómez Sara, Hernández Bernardo, Joshi Rohina, Kalter Henry, Kumar Aarti, Kumar Vishwajeet, Lozano Rafael, Lucero Marilla, Mehta Saurabh, Neal Bruce, Ohno Summer Lockett, Prasad Rajendra, Praveen Devarsetty, Premji Zul, Ramírez-Villalobos Dolores, Remolador Hazel, Riley Ian, Romero Minerva, Said Mwanaidi, Sanvictores Diozele, Sazawal Sunil, Tallo Veronica
Primary Institution: Institute for Health Metrics and Evaluation, University of Washington
Hypothesis
Can verbal autopsy methods accurately determine causes of death in populations without complete medical certification?
Conclusion
The study created a robust validation dataset that allows for the evaluation of different verbal autopsy methods and their effectiveness in determining causes of death.
Supporting Evidence
- Over 12,000 verbal autopsies were collected across six sites.
- The study established stringent diagnostic criteria for identifying gold standard deaths.
- 500 test-train datasets were created to ensure independence in validation.
- The dataset can inform future verbal autopsy implementations.
Takeaway
This study collected information on over 12,000 deaths to see how well verbal autopsy methods can figure out what caused those deaths, especially in places where doctors don't always keep good records.
Methodology
Data was collected through verbal autopsies in six sites across four countries, using stringent diagnostic criteria to identify gold standard deaths.
Potential Biases
Potential biases include the quality of medical records and the reliance on hospital diagnoses, which may not be accurate.
Limitations
The study faced challenges in finding sufficient cases to meet gold standard criteria and issues with misclassification for certain causes.
Participant Demographics
The study included deaths from various age groups: 7,836 adults, 2,075 children, 1,629 neonates, and 1,002 stillbirths.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website