Causes of Death

Accurate information about the actual cause of death is critical but often missing or inaccurately entered on hospital and clinic forms. Doctors or nurses completing death certificates may list the way that someone died—such as cardiac arrest—or describe a symptom—such as back pain—but fail to capture the underlying cause of death: HIV, unintentional injury, cardiovascular disease, or any of the myriad conditions or factors which can be fundamentally responsible for a fatality.

This situation presents tremendous challenges for decision-makers and funders determining how to most effectively target limited resources. The lack of available data, especially in middle- and low-income countries, underscores the need for a reliable way to fill in data gaps. IHME researchers are developing new analytic tools to meet this need, such as models that redistribute deaths that were mistakenly assigned to non-fatal diseases or symptoms – not to the actual underlying causes.

Major activities in Causes of Death:

Implement gold-standard data collection for verbal autopsies and validate methods for analyzing them:

IHME is testing and refining a verbal autopsy survey instrument – to be administered with an adult household member of a person who recently died. The objective is to develop a concise questionnaire that collects both symptom and demographic-related information and allows analysis to distribute likely causes of death across a given population. IHME is presently implementing verbal autopsies at sites in India, Tanzania, and the Philippines and will soon begin in Mexico. After refining this method, the Causes of Death research team will apply IHME’s Symptom Pattern Method to the collected data to assess population-level causes of death and validate the methods results. The method has been refined and improved by applying it to verbal autopsy data from Thailand. This is a breakthrough approach that will leverage information collected in-person to assess population-level causes of death.

Apply hospital-based cause of death methods to select national datasets:

In some instances, complete hospital death records are the only guide to deaths overall in a community. However, those who die in the hospital, or even those who are diagnosed there, are not necessarily representative of the population as a whole. Estimating population-level causes of death from hospital records therefore requires methods that incorporate other known characteristics of a population. IHME plans to use a method developed as part of the PHMRC Study and validated against data in Mexico at a subnational level to conduct the initial analysis using this technique. The research team will seek subnational data for other countries where we can apply similar techniques to check and improve the method. More information about this method can be found in the research article published in PLoS Medicine by Murray et al (2007).

Produce country-level estimates of causes of death by age and sex:

The Causes of Death team is producing country-level estimates of 24 clusters of causes of death by age and sex from 1990 to 2005. While IHME’s Mortality team is generating estimates of the rates of mortality worldwide, the Causes of Death team is working to provide information about the underlying reasons for those deaths. Researchers are drawing from vital registration data, disease registries, and other centrally collected data from national governments.

Produce regional estimates of causes of death for the Global Burden of Disease Study:

The GBD Study divides the world into 21 regions, which taken together are globally comprehensive. Using some of the methods developed to create estimates of the cause-clusters by country, IHME will produce estimates for 240 causes.

Top of Page