A severity-of-illness score in patients with tuberculosis requiring intensive care

Background We previously retrospectively validated a 6-point severity-of-illness score aimed at identifying patients at risk of dying of tuberculosis (TB) in the intensive care unit (ICU). Parameters included septic shock, HIV infection with a CD4 count <200 cells/µL, renal dysfunction, a ratio of partial pressure of arterial oxygen to fraction of inspired oxygen (P/F) <200 mmHg, a chest radiograph demonstrating diffuse parenchymal infiltrates, and no TB treatment on admission. Objectives To prospectively validate the severity-of-illness scoring system in patients with TB requiring intensive care, and to refine and simplify the score in order to expand its clinical utility. Methods We performed a prospective observational study with a planned post hoc retrospective analysis, enrolling all adult patients with confirmed TB admitted to the medical ICU of a tertiary hospital in Cape Town, South Africa, from 1 February 2015 to 31 July 2018. The admission data of all adult patients with TB requiring admission to the ICU were used to calculate the 6-point severity-of-illness score and a refined 4-point score (based on the planned post hoc analysis). Descriptive statistics and χ2 or Fisher's exact tests (where indicated) were performed on dichotomous categorical variables, and t-tests on continuous data. Patients were categorised as hospital survivors or non-survivors. Results Forty-one of 78 patients (52.6%) died. The 6-point scores of non-survivors were higher than those of survivors (mean (standard deviation (SD)) 3.5 (1.3) v. 2.7 (1.2); p=0.01). A score ≥3 v. <3 was associated with increased mortality (64.0% v. 32.1%; odds ratio (OR) 3.75; 95% confidence interval (CI) 1.25 - 10.01; p=0.01). Post hoc, a P/F ratio <200 mmHg and no TB treatment on admission failed to predict mortality, whereas any immunosuppression did. A revised 4-point score (septic shock, any immunosuppression, acute kidney injury and lack of lobar consolidation) demonstrated higher scores in non-survivors than survivors (mean (SD) 2.8 (1.1) v. 1.6 (1.1); p<0.001). A score ≥3 v. ≤2 was associated with increased mortality (78.4% v. 29.3%; OR 8.76; 95% CI 3.12 - 24.59; p<0.001). Conclusions The 6-point severity-of-illness score identified patients at increased risk of death. We were able to derive and retrospectively validate a simplified 4-point score with superior predictive power.