Assessing reliability of the verbal autopsy method for identifying adult HIV status among a living population: Agreement between adult and child reports on the HIV status of adults caring for children in HIV endemic South Africa

Background: Verbal autopsy methods have been used globally to determine HIV related cause of death among individuals in communities with generalized HIV epidemics. Verbal autopsy has the potential to be a valid measurement tool for estimating HIV illness among living adult populations.

Methods: A cross-sectional survey using a representative community sample of matched adult-child dyads (2,477 adults caring for child ≥18 and N=2,477 children 10-17 years) was conducted in two high-HIV prevalence communities in South Africa. Analyses focused on a subset of this larger sample (n=673 adult-child dyads). Adult-child dyads reported on adult HIV symptoms based on verbal autopsy method.

Results: Concordant reporting between adults and children on adult HIV status based on verbal autopsy evaluation was reliable at 72%. Children reliably reported on eight of the ten symptoms used in the verbal autopsy method. Reports for jaundice, oral candidiasis, herpes zoster, TB, wasting, sores, weight loss and acute respiratory illness and all had concurrency agreement above the sixtieth percentile. Children who correctly matched their corresponding adult for verbal autopsy evaluation and subsequent predictive HIV status (n=484) were between 14-17 years (53%), female (53%), from the urban study site (66%) and in grades 7-12 (55%).

Conclusion: Verbal autopsy is a reliable method for assessing adult HIV illness in a generalized epidemic setting with a living population using concurrent adult and child reports of HIV symptoms. Certain characteristics of children including gender, age, education and location are associated with greater accuracy in matching their corresponding adult's answers.