Two distinct subtypes of severe pulmonary tuberculosis for short-term risk stratification: a cluster analysis

Severe tuberculosis (STB) is clinically heterogeneous, but whether distinct subtypes with different short-term clinical outcomes exist is unknown. This retrospective study included 500 STB patients. K-means clustering using six routine variables (age, BMI, albumin, lymphocyte, neutrophil, CRP) identified subtypes. Clinical outcomes were compared between clusters. Two subtypes emerged: Cluster 1 (hyperinflammatory, n = 228) and Cluster 2 (hypoinflammatory, n = 272). Patients in Cluster 1 were older than those in Cluster 2, with lower BMI and albumin, lymphopenia, and higher neutrophils and CRP. When compared with Cluster 2, Cluster 1 had significantly higher rates of non-invasive ventilation (11.8% vs. 1.8%), invasive mechanical ventilation (8.8% vs. 2.2%), ICU admission (6.1% vs. 1.1%), and longer hospital stay (14.0 vs. 11.0 days, all P < 0.01), with no statistically significant difference in in-hospital mortality (Cluster 1: 2/228, 0.9%; Cluster 2: 1/272, 0.4%; P = 1.000), though the very low number of deaths substantially limits the interpretability of this comparison. Multivariate logistic regression adjusting for age, BMI, autoimmune disease, and COPD confirmed that Cluster 2 remained at significantly lower risk for NIV (OR = 0.138, 95% CI 0.037–0.421, P = 0.001), IMV (OR = 0.149, 95% CI 0.050–0.391, P < 0.001), and ICU admission (OR = 0.078, 95% CI 0.017–0.267, P < 0.001), with Cluster 1 as the reference group. Six routine variables stratify STB into hyperinflammatory and hypoinflammatory clusters. The hyperinflammatory cluster requires more respiratory support and ICU care. This offers a pragmatic risk stratification tool for resource-limited settings.