Critical information on population health is needed to inform planning, resource allocation, program implementation, monitoring, and evaluation. Since many countries lack complete vital registration systems, one of the key pieces of information about population health is missing – causes of death. Understanding what the most prevalent causes of death are in a given population can help target preventive interventions and provide health services.
Verbal autopsy (VA) is a method of determining individuals’ causes of death and cause-specific mortality fractions in populations without a complete vital registration system. Verbal autopsies consist of a trained interviewer using a questionnaire to collect information about the signs, symptoms, and demographic characteristics of a recently deceased person from an individual familiar with the deceased. A standard VA instrument paired with easy-to-implement and effective analytic methods could help bridge significant gaps in information about causes of death, particularly in resource-poor settings.
The Verbal Autopsy research team’s activities provide data and resources that will be the basis for rapid and effective field assessment of population-level causes of death. The goals of the research team are to:
- Develop methods with the highest possible performance, as assessed by an explicit set of transparent metrics, that can accurately and expeditiously help estimate population-level causes of death
- Analyze community verbal autopsy data
- Create methods for combining verbal autopsy data with hospital records to further improve population-level estimates for causes of death
- Develop high-performing methods to analyze verbal autopsy data
IHME was the lead institution in the Population Health Metrics Research Consortium (PHMRC) project, which aimed to develop methods and instruments to quantitatively assess population health. IHME collaborated with the Harvard School of Public Health, Johns Hopkins Bloomberg School of Public Health, the University of Queensland, the Research Institute for Tropical Medicine (the Philippines), the George Institute (India), King George Medical College (India), the Pemba Public Health Laboratory (Tanzania), and the Muhimbili University of Health and Allied Sciences (Tanzania) to collect over 12,500 gold standard verbal autopsies in multiple cultural and linguistic settings.
In order to validate verbal autopsy analysis methods, data were collected in hospital and clinical environments for decedents with known causes of death. The Consortium developed a list of stringent gold standard definitions for each cause of death, detailing the diagnostic criteria required to qualify for the study. Cases that met the criteria were followed with blinded VA interviews with a relative of the deceased. The VA questionnaire was used to collect information about the symptoms of the deceased, demographic characteristics, possible risk factors (such as tobacco use), and other potentially contributing characteristics.
The most commonly used approach to analyzing VA data is physician review. However, this approach can prove both time-consuming and costly. All of the gold standard cases collected as part of the PHMRC have been physician-reviewed to provide a comparison to the current standard procedures in the field. In parallel with collecting the results of the physician coding, we have been testing a suite of different methods. Three of them were previously in existence: InterVA (a program that incorporates commonly applied physician decision points by coding them into algorithms), the King-Lu method (previously developed by the PHMRC), and the Symptom Pattern Method (also previously developed by the PHMRC). In addition, we have developed a number of other methods that utilize either a tariff or a machine learning approach.
The machine learning methods outperform physician review and offer an analysis solution that is more cost effective and less time consuming than physician review, which could in turn result in more timely and accessible cause of death data in resource-poor settings.
- Analyze community verbal autopsy data
Multiple projects, including the PHMRC and the Improving Methods to Measure Comparable Mortality by Cause project, have collected community verbal autopsies in Mexico, the Philippines, Tanzania, India, Bangladesh, and Papua New Guinea. The research team is applying the new machine learning analysis methods to these datasets to produce population-level cause of death estimates for these sites.
- Co-hosted the Global Congress on Verbal Autopsy: State of the Science
The Global Congress on Verbal Autopsy: State of the Science brought together eminent researchers and practitioners from around the globe. It was co-hosted by IHME, the University of Queensland, and the journal Population Health Metrics. The conference convened key researchers and those who work with VA data to discuss important aspects of instrument design, analysis methods, and the use of VA in national health information systems.
A great deal of research has been conducted in the past several decades about VA, but some traditional methods of implementation and analysis can be costly, time consuming, and potentially of varying quality. Verbal autopsies are now analyzed using a much wider array of cutting edge techniques, some of which could be less expensive or yield higher quality results than those used traditionally. Conference participants discussed and debated these different approaches with the goal of producing the best possible information for health systems in regions where verbal autopsies are being used.
The Verbal Autopsy research team’s gold standard VA data set, instruments, analysis tools, and protocols for implementing it will be placed in the public domain, enabling policymakers and researchers to help address persistent inequities in health outcomes in both the developed and the developing world.
Related Publications & Presentations
Murray CJL, Lozano R, Flaxman AD, Vahdatpour A, Lopez AD. Robust metrics for assessing the performance of different verbal autopsy cause assignment methods in validation studies. Population Health Metrics. 2011; 9:28.
Hernandez B, Ramirez-Villalobos D, Romero M, Gomez S, Atkinson C, Lozano R. Assessing quality of medical death certification: concordance between gold standard diagnoses and underlying cause of death in selected Mexican hospitals. Population Health Metrics. 2011; 9:38.
Flaxman AD, Vahdatpour A, James SL, Birnbaum JK, Murray CJL, and the Population Health Metrics Research Consortium (PHMRC). Direct estimation of cause-specific mortality fractions from verbal autopsies: multisite validation study using clinical diagnostic gold standards. Population Health Metrics. 2011; 9:35.
Lozano R, Freeman MK, James SL, Campbell B, Lopez AD, Flaxman AD, Murray CJL, the Population Health Metrics Research Consortium (PHMRC). Performance of InterVA for assigning causes of death to verbal autopsies: multisite validation study using clinical diagnostic gold standards. Population Health Metrics.2011; 9:50.
Lozano R, Lopez AD, Atkinson C, Naghavi M, Flaxman AD, Murray CJL, the Population Health Metrics Research Consortium (PHMRC). Performance of physician-certified verbal autopsies: multisite validation study using clinical diagnostic gold standards. Population Health Metrics. 2011; 9:32.
James SL, Flaxman AD, Murray CJL, the Population Health Metrics Research Consortium (PHMRC). Performance of the Tariff Method: validation of a simple additive algorithm for analysis of verbal autopsies. Population Health Metrics. 2011; 9:31.
Flaxman AD, Vahdatpour A, Green S, James SL, Murray CJL, the Population Health Metrics Research Consortium (PHMRC). Random Forests for verbal autopsy analysis: multisite validation study using clinical diagnostic gold standards. Population Health Metrics. 2011; 9:29.
Murray CJL, James SL, Birnbaum JK, Freeman MK, Lozano R, Lopez AD, and the Population Health Metrics Research Consortium (PHMRC). Simplified Symptom Pattern Method for verbal autopsy analysis: multisite validation study using clinical diagnostic gold standards. Population Health Metrics. 2011; 9:30.
Murray CJL, Lopez AD, Black R, Ahuja R, Ali SM, Baqui A, Dandona L, Dantzer E, Das V, Dhingra U, Dutta A, Fawzi W, Flaxman AD, Gomez S, Hernandez B, Joshi R, Kalter H, Kumar A, Kumar V, Lozano R, Lucero M, Mehta S, Neal B, Ohno SL, Prasad R, Praveen D, Premji Z, Ramirez-Villalobos D, Remolador H, Riley I, Romero M, Said M, Sanvictores D, Sazawal S, Tallo V. Population Health Metrics Research Consortium gold standard verbal autopsy validation study: design, implementation, and development of analysis datasets. Population Health Metrics. 2011; 9:27.