Verbal autopsy as a tool for diagnosing HIV-related adult deaths in rural Uganda

Int J Epidemiol. 1996 Jun;25(3):679-84. doi: 10.1093/ije/25.3.679.


Background: In general, information on the causes of adult deaths in developing countries is scarce. More specifically, relatively little is known about the effect of HIV-1 associated disease on adult mortality in general populations. In this study we have used a verbal autopsy technique to ascertain whether adult deaths were associated with HIV-1 in a rural population with a prevalence of HIV-1 infection of 8%, and used HIV-1 antibody status to validate the verbal autopsy findings.

Methods: All adult deaths in the population cohort that occurred between December 1990 and November 1993 were identified through a monthly death registration system. Approximately 2 months after death, a relative of the deceased was interviewed by a trained nurse, and questionnaires were assessed by at least two independent clinicians; all were unaware of the HIV serostatus of the deceased.

Results: A total of 155 adult deaths was assessed, i.e. 53% of all recorded adult deaths. Of those assessed half were HIV-1 positive. In all 47% of deaths were classified as HIV-related. The overall specificity and positive predictive value of the verbal autopsy tool were both 92%; in those aged 13-44 years (83 adults) the corresponding values were 85% and 95% respectively. The verbal autopsy estimated HIV-1 attributable mortality fraction was similar to the calculated fraction based on prospective data.

Conclusions: The results of this study suggest that verbal autopsy studies may assist in providing data on HIV-associated mortality in general populations and may be useful as surveillance tools.

PIP: The verbal autopsy technique represents a means for increasing the accuracy of acquired immunodeficiency syndrome (AIDS)-related mortality statistics in sub-Saharan Africa, where many deaths occur at home. A structured interview, adapted to local disease perceptions, is administered by lay personnel to relatives or friends of the deceased. This technique was evaluated in a rural area of Uganda's Masaka district with a human immunodeficiency virus (HIV)-1 seroprevalence rate of 8%. All adult deaths occurring in the 15 index villages from December 1990 to November 1993 were identified through a monthly village-based death registration system. A nurse paid an initial condolence visit to the bereaved family and returned about two months later with the questionnaire. Interviews were conducted with families or friends of 155 (53%) of the 293 adult deaths for which HIV serostatus was known. At least two clinicians unaware of serostatus results assessed each questionnaire and indicated the likely causes of death. There was agreement on whether the death was or was not HIV-related in 141 (91%) of these deaths. The 14 cases with discordant assessments were reviewed by a third clinician. Overall, 73 (47%) of deaths were attributed to HIV-related causes. Using HIV serostatus as a standard of comparison, the overall specificity and predictive value of the verbal autopsy were both 92%. These findings suggest that the method can be relied upon to estimate HIV-associated mortality in populations with a relatively high HIV prevalence.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Adolescent
  • Adult
  • Cause of Death*
  • HIV Infections / diagnosis
  • HIV Infections / mortality*
  • Humans
  • Interviews as Topic / methods*
  • Rural Population
  • Sensitivity and Specificity
  • Seroepidemiologic Studies
  • Surveys and Questionnaires
  • Uganda / epidemiology