Estimating the Age at HIV Infection Retroactively in Limited Resource Settings: <i>A Case Study of Tanzania</i>

Abstract

Estimation of HIV infection time is a crucial step in HIV/AIDS management as it can help to make informed decisions on the best intervention strategies for controlling new infections, and for taking care of the infected individuals. This study demonstrates three approaches for estimating the age at HIV infection in limited resource settings. Using HIV testing history data collected from a sample of 88 HIV positive women in Kilimanjaro region-Tanzania, we developed a model for estimating the most likely age at which HIV infection occurs for women under reproductive age. The sampled data were collected from typical poor resource settings where access to data is very challenging and the gap between last HIV negative test and first HIV positive test is wide. Formulation of the proposed model involved three steps. Through Modified Midpoint approach, we first determined the midpoint of the age at last negative HIV test and the age at first positive HIV test for each subject. Then, the average time at risk prior to infection, taken over all individuals was subtracted from each midpoint value to obtain the distribution of their estimated age at HIV infection (T). In the second step, survival analysis techniques were used to obtain the Kaplan Meier plots and Nelson Aalen cumulative hazards estimates in which the median age for HIV infection and the most risky age were estimated. The plots of Kaplan Meir survival curves for women with different marital status and levels of education helped to assess whether their age at infection were significantly different. In the third step, we used bootstrap estimation procedures to generate 200 samples of random data and obtain the bootstrap median age at HIV infection and its confidence intervals. The estimated median age at HIV infection from survival analysis approach was 28 years while from bootstrap estimation procedures was 27 years. Likewise, the Nelson Aalen cumulative hazards plot indicated that the most risky age for HIV infection is between 18-40 years while the most risky age from bootstrap estimation was 25 to 27 years. The confidence intervals obtained through bootstrap estimation approach was narrower than that obtained from the survival analysis approach, implying that the bootstrap approach gives more precise estimates. Generally, the study findings provide useful information towards the attainment of the 90-90-90 global HIV/AIDS target as it shows where to allocate more resources and establish more focused interventions for HIV/AIDS management and control.