Publication date: 
January 28, 2014

Findings show the preferred method for gathering and analyzing cause of death information in low-resource settings

New research from the Institute for Health Metrics and Evaluation (IHME) shows that verbal autopsy (VA) is a powerful and valid method for data gathering in low-resource settings and points to a newly developed tool as the analytic method of choice.

VA is a type of household survey that helps identify causes of death in a population. Trained interviewers use a questionnaire to gather information about the signs, symptoms, and demographic characteristics of a recently deceased person from a relative or other individual who knew them. In many low-income countries that lack vital registration systems, causes of death are not officially recorded. VA is often the only way to obtain key information about population health in these settings.

Verbal autopsy has great potential for countries like this where millions of people die every year from unknown causes. With huge gaps in data on what is killing people, it is nearly impossible for governments to respond appropriately and to plan effectively.

When VA was first devised, physicians reviewed the questionnaires individually, assessed the symptoms reported, and assigned a cause of death to each one, a laborious and expensive practice. In an effort to streamline this process, computerized methods have been developed to analyze data collected through VA surveys and assign causes of death.

“One of the impediments to wider use of verbal autopsy in routine data collection in low-resource settings has been the amount of time required of a scarce resource – namely physicians – to read the autopsies,” said IHME Director Dr. Christopher Murray. “That major impediment to generating better cause of death data can be removed.” 

The InterVA method, one of the earliest techniques, helped modernize verbal autopsy tools by pioneering a way of transferring physician review to an automated, computerized process. The question remained, though, about whether such a method could provide both the speed of a computer and the accuracy of a physician.

In the new paper, “Using verbal autopsy to measure causes of death: the comparative performance of existing methods,” published this month in BMC Medicine, researchers investigated five automated VA methods for assigning cause of death: InterVA-4, Random Forest, Simplified Symptom Pattern, Tariff method, and King-Lu.

The Tariff method proved through a series of analyses to perform consistently better than physician review, and the authors recommend it for wide use in mortality surveillance systems with poor cause of death certification practices.

“In our lifetime we will never find enough doctors with enough time to certify the deaths of the 30 million people who die every year without a cause of death,” said Dr. Alan Lopez, Laureate Professor at the University of Melbourne. “But we have the cost-effective means right now to help make important policy decisions that will improve health in communities that very much need it.”

This week in Dubai at the Arab Health Congress, Dr. Ali Mokdad, Director of Middle Eastern Initiatives for IHME, is releasing a new toolkit for making the Tariff method more easily accessible worldwide.

“Countries worldwide are undergoing tremendous change in health patterns, and many of them still have a long way to go to make their health information systems strong enough to respond to these new health challenges,” Dr. Mokdad said. “To be able to act first instead of reacting when it’s too late, they need these verbal autopsy tools.”

You can find out more about verbal autopsy at