BackgroundTuberculosis (TB) is a preventable and curable disease, but mortality remains high among those who develop sepsis and critical illness from TB.MethodsThis was a population-based, multicentre retrospective cohort study of patients admitted to all 15 publicly funded Hong Kong adult intensive care units (ICUs) between 1 April 2008 and 31 March 2019. 940 adult critically ill patients with at least one positive Mycobacterium tuberculosis (MTB) culture were identified out of 133 858 ICU admissions. Generalised linear modelling was used to determine the impact of delay in TB treatment on hospital mortality. Trend of annual Acute Physiology and Chronic Health Evaluation (APACHE) IV-adjusted standardised mortality ratio (SMR) over the 11-year period was analysed by Mann-Kendall’s trend test.ResultsICU and hospital mortality were 24.7% (232/940) and 41.1% (386/940), respectively. Of those who died in the ICU, 22.8% (53/232) never received antituberculosis drugs. SMR for ICU patients with TB remained unchanged over the study period (Kendall’s τb=0.37, p=0.876). After adjustment for age, Charlson comorbidity index, APACHE IV, albumin, vasopressors, mechanical ventilation and renal replacement therapy, delayed TB treatment was directly associated with hospital mortality. In 302/940 (32.1%) of patients, TB could only be established from MTB cultures alone as Ziehl-Neelsen staining or PCR was either not performed or negative. Among this group, only 31.1% (94/302) had concurrent MTB PCR performed.ConclusionsSurvival of ICU patients with TB has not improved over the last decade and mortality remains high. Delay in TB treatment was associated with higher hospital mortality. Use of MTB PCR may improve diagnostic yield and facilitate early treatment.