Background: Mental disorders can adversely affect HIV treatment outcomes and survival. Data are scarce on premature deaths in people with mental disorders in HIV-positive populations, particularly in low-income and middle-income countries. In this study, we quantified excess mortality associated with mental disorders in HIV-positive people in South Africa, adjusting for HIV treatment outcomes. Methods: For this cohort study, we analysed routinely collected data on HIV-positive adults receiving antiretroviral therapy (ART) in Cape Town, South Africa between Jan 1, 2004, to Dec 31, 2017. Data from three ART programmes were linked with routine medical records on mental health treatment from Jan 1, 2010, to Dec 31, 2017, and mortality surveillance data from the South African National Population Register up to Dec 31, 2017. People living with HIV aged 15 years or older who initiated ART at a programme site were eligible for analysis. We followed up patients from ART initiation or Jan 1, 2010, whichever occurred later, to transfer, death, or Dec 31, 2017. Patients were considered as having a history of mental illness if they had ever received psychiatric medication or been hospitalised for a mental disorder. We calculated adjusted hazard ratios (aHRs) with 95% CIs for associations between history of mental illness, mortality, and HIV treatment outcomes (retention in care with viral load suppression [VLS; viral load <1000 copies per mL], retention in care with non-suppressed viral load [NVL; viral load ≥1000 copies per mL], and loss to follow-up [LTFU; >180 days late for a clinic visit at closure of the database]) using Cox proportional hazard regression and multistate models. Results: 58 664 patients were followed up for a median of 4·3 years (IQR 2·1–6·4), 2927 (5·0%) of whom had a history of mental illness. After adjustment for age, sex, treatment programme, and year of ART initiation, history of mental illness was associated with increased risk of mortality from all causes (aHR 2·98 [95% CI 2·69–3·30]), natural causes (3·00 [2·69–3·36]), and unnatural causes (2·10 [1·27–3·49]), compared with no history of mental illness. Risk of all-cause mortality in people with a history of mental illness remained increased in multivariable analysis adjusted for age, sex, treatment programme, year of ART initiation, CD4 count and WHO clinical stage at ART initiation, retention in HIV care with or without VLS, and LTFU (2·73 [2·46–3·02]). In our multistate model, adjusted for age, sex, year of ART initiation, cumulative time with NVL, and WHO clinical stage and CD4 cell count at ART initiation, rates of excess all-cause mortality in people with history of mental illness were greatest in patients retained in care with VLS (aHR 3·43 [95% CI 2·83–4·15]), followed by patients retained in care with NVL (2·74 [2·32–3·24]), and smallest in those LTFU (2·12 [1·78–2·53]). History of mental illness was also associated with increased risk of HIV viral rebound (transitioning from VLS to NVL; 1·50 [1·32–1·69]) and LTFU in people with VLS (1·19 [1·06–1·34]). Interpretation: Mental illness was associated with substantial excess mortality in HIV-positive adults in Cape Town. Excess mortality among people with a history of mental illness occurred independently of HIV treatment success. Interventions to reduce excess mortality should address the complex physical and mental health-care needs of people living with HIV and mental illness. Funding: National Institutes of Health, Swiss National Science Foundation, South African Medical Research Council.
