Estimating the risk of mortality attributable to recent late HIV diagnosis following admission to the intensive care unit: A single-centre observational cohort study.

OBJECTIVES: Despite improvements in survival of people with HIV admitted to the intensive care unit (ICU), late diagnosis continues to contribute to in-ICU mortality. We quantify the population attributable fraction (PAF) of in-ICU mortality for recent late diagnosis among people with HIV admitted to a London ICU. METHODS: Index ICU admissions among people with HIV were considered from 2000 to 2019. Recent late diagnosis was a CD4 T-cell count < 350 cells/µL and/or AIDS-defining illness at/within 6 months prior to ICU admission. Univariate comparisons were conducted using Wilcoxon rank-sum/Cochran-Armitage/χ2 /Fisher's exact tests. We used Poisson regression (robust standard errors) to estimate unadjusted/adjusted (age, sex, calendar year of ICU admission) risk ratios (RRs) and regression standardization to estimate the PAF. RESULTS: In all, 207 index admissions were included [median (interquartile range) age: 46 (38-53) years; 72% male]; 58 (28%) had a recent late diagnosis, all of whom had a CD4 count < 350 cells/µL, and 95% had advanced HIV (CD4 count < 200 cells/µL and/or AIDS at admission) as compared with 57% of those who did not have a recent late diagnosis (p < 0.001). In-ICU mortality was 27% (55/207); 38% versus 22% in those who did and did not have a recent late diagnosis, respectively (p = 0.02). Recent late diagnosis was independently associated with increased in-ICU mortality risk (adjusted RR = 1.75) (95% confidence interval: 1.05-2.91), with 17.08% (16.04-18.12%) of deaths being attributable to this. CONCLUSIONS: There is a need for improved public health efforts focused on HIV testing and reporting of late diagnosis to better understand potentially missed opportunities for earlier HIV diagnosis in healthcare services.