Using Burial Surveillance Data to Monitor HIV/AIDS Mortality in Developing Countries. An Example from Ethiopia

Georges Reniers, University of Pennsylvania
Tekebash Araya, Addis Ababa University
Eduard Sanders, Ethiopian Health and Nutrition Research Institute

The current state of knowledge of HIV/AIDS mortality does not quite match the attention the pandemic has received over the last two decennia. The major obstacle in obtaining more precise and reliable estimates of HIV/AIDS mortality is the availability of adequate data, vital statistics in particular. In addition to epidemiological models to project mortality from extrapolated surveillance data, AIDS mortality can be estimated using indirect techniques referred to as the brotherhood or orphanhood method. Another fairly standard methodological response in contexts with limited data availability is the use of verbal autopsies. Despite their respective merits, these sources and methods lack the advantage of monitoring mortality on a continuous basis. This contribution examines the validity of a new, relatively cheap and partial substitute for a proper vital registration system: burial surveillance data. The burial surveillance was initiated at all cemeteries of Addis Ababa in February 2001 and currently covers 55 Orthodox, 9 Muslim, 1 Catholic, 1 Jewish and 8 municipal cemeteries. Over 21,000 deaths were registered during the first year, and initial calculations suggest that the surveillance is exhaustive. The surveillance is assisted by cemetery clerks who fill out a relatively simple form to collect data on the name, sex, age, date of burial, address, cause of death, marital status, birth region, ethnicity and religion for each of the individuals buried. Relatives or close friends facilitating administrative procedures before the burial ceremony provide the cemetery clerks with the information. In addition to a discussion of the quality of these data, we illustrate their use to estimate HIV/AIDS mortality. In doing so, we follow two strategies that give us quasi-independent estimates. The first method is based on the lay diagnoses collected at the burial sites, the second uses the age and sex distribution of deaths as an input to create and analyze life tables. To evaluate the predictive value of the lay diagnoses of death, they are compared with two different gold standards. The first comes from a physician review of verbal autopsies that were collected on a sample of the burial surveillance records (N=200). The second gold standard is derived from matched hospital records that were collected as part of a simultaneous surveillance of hospital deaths in Addis Ababa (2546 hospital diagnoses and 1480 outcomes of autopsies over the course of 1 year). Our results indicate that AIDS accounts for around 68% of the deaths in the adult population and that this estimate is not sensitive to the choice of the gold standard. Records of hospital deaths tend to underestimate HIV/AIDS mortality by around 5%. The results are also promising with respect to the use of lay diagnoses to monitor HIV/AIDS mortality, even though a great reluctance exists to talk about HIV/AIDS directly (only 1% of the adult deaths are explicitly ascribed to AIDS). ‘Lung disease’ and ‘common cold’ in particular have become euphemisms for HIV/AIDS in the community. Together they have a specificity of 90% and a sensitivity of around 55%. The sensitivity increases with 20 percentage-points when we also include ‘diarrhea’, ‘TB’, ‘herpes zoster’, ‘mental or nerve problem’, ‘sickness’ and other undetermined causes of death. In that case, the gain in sensitivity is compensated by a loss in the specificity of 10 percentage-points. The second method to use these burial surveillance data to obtain estimates of HIV/AIDS mortality takes the age and sex distribution of deaths as the starting point to create life tables and analyze age patterns of mortality. No definitive results of these analyses are available at this stage. In the concluding remarks, we discuss the strengths and weaknesses of the burial surveillance data to monitor mortality in countries with similar characteristics. The advantages of the system as it exists now is that it taps into an existing infrastructure, that it is relatively cheap and that it could become a permanent source for vital demographic data. In addition to the application that is illustrated in this paper, we suggest other potential uses.