BACKGROUND: Stigma is a major barrier to ending HIV as a public health threat. We present an analytical framework for quantifying the effects of HIV-related stigma on the treatment cascade using biomarker data from a Population-based HIV Impact Assessment (PHIA) in Tanzania. METHODS: We first reviewed HIV-related stigma items from 15 PHIA surveys in sub-Saharan Africa. Using nationally representative data of 1,831 diagnosed and undiagnosed PLHIV aged 15 and older in Tanzania, we applied modified Poisson regression models to examine associations of stigma with the treatment cascade, adjusting for HIV knowledge and demographics. RESULTS: We identified 41 unique stigma-related items in 13 of the 15 PHIA surveys. In Tanzania, PLHIV who expressed any stigma driver (stigmatizing attitude, discriminatory attitude, or shame) were 27% less likely to know their HIV status (adjusted prevalence ratio [aPR] 0.73; 95%CI [0.65-0.83], p < 0.001), while those expressing all three were almost never aware of their status (aPR < 0.01; 95%CI [0-0.01], p < 0.001). Stigma drivers were not significantly associated with ART use among diagnosed PLHIV or viral load suppression (VLS) among those on ART. Diagnosed PLHIV who felt the need to hide their status when seeking non-HIV healthcare were 9% less likely to be on ART (aPR 0.91; 95%CI [0.85-0.98], p = 0.013), and those on ART were 10% less likely to achieve VLS (aPR 0.90; 95%CI [0.81-0.99], p = 0.047). CONCLUSIONS: Stigma likely prevented many undiagnosed PLHIV in Tanzania from knowing their status. Fear of healthcare discrimination due to anticipated stigma undermines ART uptake among diagnosed PLHIV and viral suppression among those on ART. PHIA surveys have untapped potential to quantify the effects of HIV-related stigma and inform interventions to end HIV as a public health threat.

INTRODUCTION: The World Health Organization recommended the initiation of antiretroviral therapy (ART) for people living with HIV (PLHIV) regardless of CD4 cell counts. Tanzania adopted this recommendation known as test-and-treat policy in 2016. However, programmatic implementation of this policy has not been assessed since its initiation. The objective of the study was to assess the impact of this policy in Tanzania. METHODS: This was a cross-sectional study among PLHIV aged 15 years and older using routinely collected program data. The dependent variable was interruption in treatment (IIT), defined as no clinical contact for at least 90 days after the last clinical appointment. The main independent variable was test-and-treat policy status which categorized PLHIV into the before and after groups. Co-variates were age, sex, facility type, clinical stage, CD4 count, ART duration, and body mass index. The associations were assessed using the generalized estimating equation with inverse probability weighting. RESULTS: The study involved 33,979 PLHIV-14,442 (42.5%) and 19,537 (57.5%) were in the before and after the policy groups, respectively. Among those who experienced IIT, 4,219 (29%) and 7,322 (38%) were in the before and after the policy groups respectively. Multivariable analysis showed PLHIV after the policy was instated had twice [AOR 2.03; 95%CI 1.74-2.38] the odds of experiencing IIT than those before the policy was adopted. Additionally, higher odds of experiencing IIT were observed among younger adults, males, and those with advanced HIV disease. CONCLUSION: Demographic and clinical status variables were associated with IIT, as well as the test-and-treat policy. To achieve epidemic control, programmatic adjustments on continuity of treatment may are needed to complement the programmatic implementation of the policy.

INTRODUCTION: Population-based biomarker surveys are the gold standard for estimating HIV prevalence but are susceptible to substantial non-participation (up to 30%). Analytical missing data methods, including inverse-probability weighting (IPW) and multiple imputation (MI), are biased when data are missing-not-at-random, for example when people living with HIV more frequently decline participation. Heckman-type selection models can, under certain assumptions, recover unbiased prevalence estimates in such scenarios. METHODS: We pooled data from 142,706 participants aged 15-49 years from nationally representative cross-sectional Population-based HIV Impact Assessments in seven countries in sub-Saharan Africa, conducted between 2015 and 2018 in Tanzania, Uganda, Malawi, Zambia, Zimbabwe, Lesotho and Eswatini. We compared sex-stratified HIV prevalence estimates from unadjusted, IPW, MI and selection models, controlling for household and individual-level predictors of non-participation, and assessed the sensitivity of selection models to the copula function specifying the correlation between study participation and HIV status. RESULTS: In total, 84.1% of participants provided a blood sample to determine HIV serostatus (range: 76% in Malawi to 95% in Uganda). HIV prevalence estimates from selection models diverged from IPW and MI models by up to 5% in Lesotho, without substantial precision loss. In Tanzania, the IPW model yielded lower HIV prevalence estimates among males than the best-fitting copula selection model (3.8% vs. 7.9%). CONCLUSIONS: We demonstrate how HIV prevalence estimates from selection models can differ from those obtained under missing-at-random assumptions. Further benefits include exploration of plausible relationships between participation and outcome. While selection models require additional assumptions and careful specification, they are an important tool for triangulating prevalence estimates in surveys with substantial missing data due to non-participation.