Physiologic–Inflammatory–Nutrition (TRIAD-TB) Score at 72 Hours Predicts 30-Day Mortality and Length of Stay in Pulmonary Tuberculosis: A Prospective Cohort Study

<b>Background and Objectives</b>: Ward-level risk in pulmonary tuberculosis (TB) is often estimated from static admission data. This study evaluated a pragmatic composite-TRIAD-TB-integrating physiology (SpO₂, respiratory rate), inflammation (systemic immune-inflammation index [SII], C-reactive protein [CRP]), and nutrition (BMI, albumin), augmented by 72 h changes in CRP and albumin, to predict 30-day mortality and hospital length of stay (LOS). <b>Methods</b>: A pooled prospective cohort of 126 HIV-negative adults without chronic systemic immunosuppression hospitalized with culture-confirmed pulmonary TB at two Romanian tertiary centers was analyzed. TRIAD-TB combined z-scored admission markers and 72 h deltas. The primary outcome was 30-day all-cause mortality; secondary outcomes included LOS. Associations were estimated using Firth logistic regression (mortality) and quasi-Poisson regression (LOS). Discrimination and overall performance were summarized by AUC and Brier score; internal performance used bootstrap optimism-correction. <b>Results</b>: Across TRIAD-TB tertiles, 30-day mortality increased from 2.4% to 16.7%, and mean LOS rose from 24.7 ± 5.8 to 32.1 ± 7.3 days. Each SD increase in TRIAD-TB was associated with higher odds of death (adjusted OR 2.4, 95% CI 1.3-4.8; <i>p</i> = 0.006) and longer hospitalization (adjusted IRR 1.19, 95% CI 1.09-1.30; <i>p</i> < 0.001). The full model discriminated mortality well (AUC 0.84; Brier 0.067) and explained 21.8% of LOS deviance. Early dynamics were informative: higher CRP ratio (72 h/0 h) and albumin decline tracked with adverse outcomes. An admission-only "mini-TRIAD" retained strong discrimination (AUC 0.79). <b>Conclusions</b>: A dynamic composite leveraging routine vitals and laboratory tests-plus 72 h trajectories-accurately stratified short-term risk in hospitalized pulmonary TB, while an admission-only "mini-TRIAD" retained strong discrimination. Together, these tools may support early escalation, targeted monitoring, and capacity planning. TRIAD-TB may support early escalation, targeted monitoring, and capacity planning; however, because it was derived in HIV-negative, non-immunosuppressed adults in an Eastern European setting, TRIAD-TB and the admission-only "mini-TRIAD" require external validation, including in cohorts with substantial HIV co-infection and different comorbidity profiles, before any broader implementation.