CROI 2020 Abstract eBook

Abstract eBook

Poster Abstracts

Results: Analyses included 17342 patients; 1403 (8.1%) had died and 8817 (50.8%) were LTFU. 1342 of patients LTFU were traced, of whom 46 (3.4%) were found to have died. At 5 years after ART start, estimated cumulative hazard (risk) of dying was 0.26 (95%-CI 0.17-0.39) out of the “LTFU” state and 0.19 (0.18-0.20) out of the “on ART” state, indicating an increased risk of dying for patients LTFU. Male sex, less advanced clinical stages, and starting ART in more recent calendar periods were associated with a greater hazard of LTFU (Figure). Higher mortality was associated with male sex, lower CD4 counts, more advanced clinical stages, and starting ART in earlier calendar periods. These associations were apparent for both patients on ART and LTFU, but differed in their magnitude (Figure). Conclusion: Multistate models are an attractive alternative to common approaches for dealing with LTFU when estimating program-level mortality in ART facilities. They allow us to distinguish between patients LTFU and those remaining in care, while still providing pooled estimates combining the two groups. Progression of patients from starting ART to LTFU and death can thus be described in more detail to inform the design of appropriate models of differentiated care.

Conclusion: People frommany of the major HIV transmission categories had a higher risk of non-HIV-associated mortality compared to those without the relevant risk factor. Interventions for people with HIV should also focus on reducing non-HIV-related causes of death to achieve maximum impact.

875 CD4 COUNT PATTERNS OVER TIME IDENTIFY LONG-TERM HIV CARE TRAJECTORIES IN SOUTH AFRICA Ingrid V. Bassett 1 , Ai Xu 1 , Janet Giddy 2 , Sue Candy 3 , Laura M. Bogart 4 , Andrew Boulle 5 , Lucia Millham 1 , Robert A. Parker 1 , Elena Losina 6 1 Massachusetts General Hospital, Boston, MA, USA, 2 McCord Hospital, Durban, South Africa, 3 National Health Laboratory Service, Johannesburg, South Africa, 4 RAND Corporation, Santa Monica, CA, USA, 5 University of Cape Town, Cape Town, South Africa, 6 Brigham and Women's Hospital, Boston, MA, USA Background: Predicting long-term care engagement at HIV diagnosis would allow targeted interventions for those at high risk of poor outcomes. Our objective was to uncover distinct CD4-based trajectories and determine baseline contextual, clinical and sociobehavioral factors associated with higher risk of being in a worse CD4 trajectory. Methods: We used data from the Sizanani trial (NCT01188941) in which adults (≥18y) were enrolled prior to HIV testing at 4 Durban outpatient sites from Aug 2010-Jan 2013. We ascertained longitudinal CD4 count data over 5y follow up using probabilistic matching with data from the National Health Laboratory Service. We used group-based statistical modeling to identify groups with similar CD4 count trajectories over time and Bayesian information criteria to determine distinct CD4 trajectories. We then evaluated baseline risk factors that predict membership in a specific (worse) trajectory using multinomial logistic regression. We examined year of enrollment, age, gender, whether people lived alone, TB positivity at enrollment, and number of domains of self- identified barriers to care (related to service delivery, financial, personal health perception, logistical, and structural) and accounted for ART initiation within 3 months of diagnosis and mortality. Results: 688 participants had longitudinal data available by NHLS crossmatch; 555 (81%) were women and median baseline CD4 count was 218 (IQR 94-368). Group-based trajectory modeling identified 4 distinct trajectories (Figure); Group 1 (19.5% of sample), with a consistent very low CD4 count that did not increase (red); Group 2 (20.7%), with a very low at baseline but increasing over time CD4 count (green), Group 3 (44.6%) with a medium-low but increasing CD4 count (blue), and Group 4 (15.9%) with a high baseline CD4 count that increased steadily overtime (black). Earlier year of enrollment, younger age, failure to start ART within 3 months, male sex, TB positivity and a greater number of self- identified barriers to care domains predicted membership in groups with poorer outcomes (Groups 1 and 2) compared to Group 4 (reference). Conclusion: One-fifth of people newly-diagnosed with HIV presented with low CD4 counts that failed to rise over time. Factors available in early clinical encounters, including potentially modifiable healthcare barriers, can predict long-term outcomes. Identifying those at high risk for poor care engagement can inform design of differentiated interventions to improve long-term clinical outcomes.

Poster Abstracts

874 INCREASED MORTALITY AMONG PEOPLE AT HIGH RISK FOR HIV IN THE UNITED STATES Fatma Shebl 1 , Julia H. Foote 1 , Krishna P. Reddy 1 , Yiqi Qian 1 , Kenneth Freedberg 1 , Rochelle P. Walensky 1 , Emily P. Hyle 1 1 Massachusetts General Hospital, Boston, MA, USA Background: People with and at risk for HIV have competing risks of mortality independent of their HIV status, such as smoking, injection drug use (IDU), and serious mental illness. We sought to quantify the non-HIV-associated mortality rates among people from the major HIV transmission categories compared to those without the relevant risk factor: men who have sex with men (MSM); high-risk heterosexuals; and people who inject drugs (PWID). Methods: We used the National Health and Nutrition Examination Survey (NHANES) (cycles 2001-14) and the National Health Interview Survey (NHIS 1991) with linked mortality data (through 2015) to examine independent associations of mortality with sexual orientation, low socio-economic status (SES), and IDU among adults (>18y). We considered male respondents to be MSM if they reported a history of male sexual partner or self-identified as gay or bisexual and compared them to heterosexuals (in NHANES). We considered low socio-economic status (SES) as a proxy for the mortality risk experienced by high-risk heterosexuals and characterized low/high SES as poverty income ratio (PIR) <1 or ≥1 to examine associations between SES and mortality (in NHANES). We categorized individuals as ever PWID if they reported ever using heroin and compared them to never IDU (in NHIS). We included all major causes of death but excluded the “other” category to avoid double-counting HIV-associated causes of mortality. We used Cox proportional hazard models to estimate age- and race-adjusted mortality rates and hazard ratios (HR) with 95% confidence intervals (CI). Analyses were stratified by age at risk (≤55y vs >55y). Results: MSM older than 55y had a non-significant higher risk of mortality compared to male heterosexuals (HR, 1.62) (Table). For females of low SES, mortality was higher for both ≤55y and >55y compared to females of high SES (HR, 2.93/3.34), whereas mortality was increased only among males of low SES older than 55y compared to males of high SES (HR, 2.47). Mortality was higher among ever PWID compared to never PWID. This was significant among ever PWID ≤55y (M/F HR, 2.75/4.09).

CROI 2020 327