Speak and Save
Where vital records don't exist, verbal autopsies yield clues to major causes of death
When death comes to newborn children in Bangladesh, its causes often are distressingly familiar and highly preventable: respiratory infection, measles, tetanus, diarrhea, premature birth. Such deaths are also distressingly common: Bangladesh had a neonatal mortality rate of 36 deaths per 1,000 live births in 2004.
Arriving at that sobering statistic is difficult. Bangladesh numbers among the nations in which “vital statistics systems are not adequately developed to capture deaths,” says Henry D. Kalter, MD, MPH ’87, an associate in International Health (IH). “A lot of people die without even having seen a doctor, or if they saw a doctor, they did not die in a medical facility. The death is not captured [statistically].”
In fact, only one-third of the nearly 56 million deaths worldwide in 2004 had a medical certification of death, WHO reports.
That lack of data on deaths and their causes denies governments and researchers evidence that can help focus resources and set priorities for interventions.
The answer? Verbal autopsies.
With a structured questionnaire and an open-ended recounting of facts surrounding the death, trained lay researchers interview surviving family members and elicit sufficient information for physicians to determine and record a cause of death.
“It’s a crude tool, but it’s the best we have available,” says Rolf D.W. Klemm, DrPH ’02, MPH ’85, an associate scientist in IH. “By knowing a proximate cause of death, however crude the method, we can advocate for earlier detection and treatment.” Klemm has used verbal autopsies in his own research on how vitamin A supplementation may be able to reduce infant mortality by 15 percent.
The raw transcripts of verbal autopsies, replete with heart-wrenching details, also provide a richness of detail that is invaluable to researchers, says Alain Labrique, PhD ’07, MHS ’99, MS, an IH assistant professor. “Verbal autopsies help us understand and identify broader nuances in repeated trajectories to mortality,” says Labrique. “The verbal autopsy offers a much richer picture about contextual factors and the pathway to mortality than a death certificate.”
Yet even expertly handled verbal autopsy interviews are less authoritative than a clinical autopsy and death certificate. One key problem is that it can be difficult for evaluators to discern between causes of death that have similar symptoms or multiple causes. For instance, the symptoms of malaria may overlap with those of pneumonia and other diseases.
“Every verbal autopsy is only as strong as its design,” says Keith West, DrPH ’87, MPH ’79, the George G. Graham Professor of Infant and Child Nutrition.
The earliest verbal autopsies usually consisted only of open-ended survivor interviews. Eventually, more rigid close-ended questionnaires—which reduce variability and lend themselves to coding—were devised. Today’s verbal autopsy is a hybrid of the two approaches, balancing greater rigor in responses and the advantages of electronic data gathering with the wealth of detail gleaned only by allowing survivors to tell their story about the death.
Do verbal autopsies work? Numerous studies have examined the results, and the answer is a strong but qualified yes.
“Verbal autopsy gives a reasonably good performance on causes of death,” observes Kalter. Verbal autopsies obtain stronger results in assigning causes to types of death that are more readily identifiable from a verbal interview and questionnaire (neonatal tetanus, for instance); results are acceptable but less authoritative with potentially murkier causes of death such as pneumonia and birth asphyxia.
The verbal autopsy has become in-dispensable to any comprehensive discussion of mortality in regions where vital statistics are lacking, says Parul Christian, DrPH ’96, MPH ’92, MSc, an IH professor. Researchers trying to reduce unnecessary deaths in those regions need information on causes of death as well as whether specific interventions are making a difference. “As a researcher, you have to rely on this sort of instrument,” says Christian.
Indeed, thousands of verbal autopsies compiled over more than a decade by researchers involved in the School’s JiVitA Maternal and Child Health and Nutrition Research Project in Bangladesh have informed the findings of micronutrient supplementation trials for pregnant mothers and newborns. They also have generated new data and hypotheses, such as recent findings suggesting that hepatitis E (preventable with a recently developed vaccine) is responsible for almost 10 percent of pregnancy-related deaths in the country.
Improvements to the public health infrastructure in Bangladesh and other countries that would make the verbal autopsy obsolete are still decades away, says Shegufta Shefa Sikder, MHS ’10, a PhD candidate conducting JiVitA research. “Verbal autopsies help to allow us to identify the major causes of mortality and prioritize research efforts,” she says.
Given the ongoing need for verbal autopsies, researchers are focusing on ways to enhance it. New technology is allowing interviewers to collect more data electronically, while innovative statistical methods can improve the validity of diagnoses. And Kalter is leading a movement to add a “social autopsy” component—questions that systematically explore social, behavioral and health care infrastructure factors. “We’re trying to extend what the verbal autopsy can tell us,” he says. “Verbal autopsy is used to determine the causes of death that health programs should focus on, while social autopsy helps understand the best ways to implement the interventions against these causes.”
As a researcher at the beginning of her career, Sikder is enthusiastic about the possibilities with the addition of the social autopsy: “It can let us trace the complex pathway to mortality.”