Despite impressive successes with the public health approach in scaling up HIV testing and ART access in the past 20 years, advanced HIV disease remains a persistent problem. People living with HIV (PLHIV) who require admission to hospital are at extremely high risk of death. Tuberculosis (TB) remains the single biggest cause of HIV-related deaths. This thesis considers adults living with HIV admitted to hospital, with the major focus being a cluster randomised trial of enhanced TB diagnostics (CASTLE trial). The thesis consists of the following chapters: first, a systematic review of interventions to reduce mortality among adults living with HIV admitted to hospital in low- and middle-income countries. Ten studies were identified, including two TB diagnostics intervention trials that showed mortality reductions. Secondly, routine data from Queen Elizabeth Central Hospital in Blantyre, Malawi was used to estimate trends in incidence of adult HIV-related admission to hospital and in-hospital deaths. The population incidence of HIV related hospital admissions declined substantially all age and sex groups from 2012 to 2019. In-hospital case fatality for admitted PLHIV remained unchanged, at 23.5%, with no significant reduction in any age-sex group, and no association with ART use at admission. Thirdly, a cluster-randomised trial (Computer Aided Screening for Tuberculosis in Low Resource Environments: CASTLE) using admission day as the unit of randomisation was designed and conducted. Admission days were randomly assigned to: 1) enhanced diagnostics for TB using urine lipoarabinomannan (LAM) (SILVAMP-LAM, Fujicorp, Japan and LF-LAM, Alere/Abbot, USA), digital chest X-ray with computer aided diagnosis (dCXR-CAD) using CAD4TBv6 (Delft, Netherlands) plus usual care; or 2) to usual care alone. The primary outcome was TB treatment initiation during admission. Between 2 September 2020 and 15 February 2022, 415 adults were recruited during 207 admission-days. TB treatment was initiated in 46/208 (22%) in the enhanced TB diagnostics arm and 24/207 (12%) in the usual care arm (risk ratio [RR] 1.92, 95% CI 1.20-3.08). Urine SILVAMP-LAM/LF-LAM plus dCXR-CAD diagnostics identified more hospitalised PLHIV with TB than usual care, but with no evidence of impact on survival, undiagnosed TB, or TB treatment initiation within 24 hours. Unanticipated findings in CASTLE included poor concordance between SILVAMP-LAM and LF-LAM results. The PhD highlights the ongoing high mortality of people living with HIV in hospital, the relative paucity of evidence based interventions for this population, and the persistently high death rate. I show that an enhanced TB diagnostic intervention is feasible and increases the number of people diagnosed with TB. To achieve goals to get to zero AIDS deaths, people with advanced HIV disease particularly those in hospital, deserve greater research attention.