Using data from hospital and clinic files, research databases, and autopsy, we describe causes and outcomes of admissions, and assess investigation for TB among adults with advanced HIV who were hospitalized after enrolment to the ‘TB Fast Track’ trial in South Africa (2013–2015; see Grant AD, et al. Algorithm-guided empirical tuberculosis treatment for people with advanced HIV (TB Fast Track): an open-label, cluster-randomised trial. Lancet HIV. 2020 Jan;7(1):e27-e37.) All participants were asked by study staff at follow-up visits if they had attended hospital at any point after enrolment. Study staff also looked for evidence of hospitalisations when abstracting from clinic files. For each participant with a recorded hospitalisation and for all participants who were recorded as having died after enrolment, the study team manually searched the admission records of the surrounding public hospitals and reviewed clinical notes when they could be located. The National Health Laboratory Service (NHLS) database was additionally searched and the results of investigations recorded. Data were abstracted by a member of the study team (doctor or professional nurse) using standardized paper forms. Ascertainment of death was done as part of the main trial. For each recorded admission, the most likely cause of admission was assigned by a research physician using all the available information. This included information that was not available to clinicians assessing the individual in real-time, such as information from primary health care (PHC) clinic files, results of laboratory tests conducted in other facilities, and findings from the parent trial or mortality sub-study (e.g., urine LF-LAM results or autopsy findings). Research-assigned causes were classified into mutually exclusive categories with criteria pre-specified for likelihood of diagnosis (possible, probable, or definite) using a standardised tool. System-based categories were used (renal/liver failure, haematological, neurological, and gastrointestinal disorders) when information was insufficient to assign a more specific diagnosis. The parent trial received ethical approval from the research ethics committees of the University of the Witwatersrand, the London School of Hygiene & Tropical Medicine (LSHTM), the South African Medicine Controls Council, and Provincial and District Departments of Health. This analysis was approved by LSHTM. All participants gave informed consent, which included permission for the study team to extract information from clinical and laboratory records. No identifiers were extracted from hospital files.

Diagnosis, Hospitalisation, Mortality, Tuberculosis, HIV