Background
As the global response to HIV intensifies toward the goal of ending the epidemic by 2030, accurate estimates of mortality among people with HIV are needed to appropriately allocate limited resources. Standard patient monitoring approaches frequently lead to misclassification of patient outcomes, including vital status. In South Africa, the national HIV electronic health record system, TIER.net, fails to link patient-level outcomes across clinics, further complicating estimation of true patient outcomes. To address this, we used a multistage sampling strategy to systematically capture and verify vital status of people with HIV in KwaZulu-Natal Province (KZN).
Methods
We employed a stratified, multi-stage sampling approach to obtain representative estimates of vital status across all 11 districts in KZN. Clinics were selected from among all government-funded clinics providing HIV treatment services in KZN using probability proportional to size sampling. Random sampling was then employed to select up to 125 patients who had accessed HIV care in the prior two years but were 90+ days late for their most recent scheduled clinic visit using TIER.net. Trained teams conducted phone-based and in-person tracing to verify patient vital status. Inverse-probability weighting was used to obtain a provincial estimate of mortality among people with HIV with a current clinical lapse.
Results
Across 36 sampled clinics, a total of 4,482 individuals were randomly selected from 33,581 individuals who had at least one clinic visit in the prior two years but were 90+ days late for their most recent scheduled visit. Of the 33,581 individuals eligible for sampling,~4% had a documented death in TIER.net. Of sampled individuals, 3,156 (70.4%) were successfully traced; 2,737 (61.1%) were alive and 419 (9.3%) had died. Clinic-level estimates of mortality ranged from 5.0% to 19.3%. After weighting, an estimated 11.0% of the 33,581 individuals seeking HIV care in KZN who were late for their most recent scheduled clinic visit had died.
Conclusions
These findings indicate that relying solely on national health record data substantially underestimates mortality in people with HIV, emphasizing the importance of tracking efforts to obtain accurate estimates of HIV-related deaths. Moreover, our results underscore the potential of sampling-based approaches to refine mortality and other HIV care estimates, thereby informing strategies that bolster re-engagement pathways and mitigate overall mortality.
Table or Figure
![]()