Prospective Validation of Four Prognostic Models for In-Hospital Mortality Prediction in Adult Traumatic Brain Injury at an Indian Tertiary Care Center

Abstract Traumatic brain injury (TBI) carries a disproportionate burden of mortality and long-term disability in India, where road trauma is the principal driver. Several prognostic instruments are available—the indigenous NIMHANS (National Institute of Mental Health and Neuro Sciences) Prediction Model and the Madras Head Injury Prognostic Scale (MHIPS), the Western-derived Edinburgh Prognostic Model, and the anatomical Injury Severity Score (ISS)—yet no Indian study has evaluated all four side-by-side in a single prospective cohort. We compared the diagnostic accuracy of these four scales for predicting in-hospital mortality at a tertiary care center in Western India. A prospective observational cohort study with diagnostic accuracy analysis was conducted over 18 months, enrolling 50 consecutive adult patients (≥ 18 years) admitted with TBI. All four scales were applied at admission using original published variables and cut-points. The primary outcome was in-hospital mortality. Sensitivity, specificity, positive predictive value, and negative predictive value were calculated from 2 × 2 contingency tables for each model at its published cut-point. Association between a composite prognostic index (Points of Survival score, dichotomized at the cohort median) and outcome was assessed by Pearson's chi-square test (p < 0.05). All analyses used IBM SPSS version 26. The cohort comprised 29 men and 21 women (mean age 38.3 ± 13.4 years); 50% sustained concurrent systemic injuries. In-hospital mortality was 22% (11/50). The NIMHANS model demonstrated the highest sensitivity (90.9%; 95% confidence interval [CI]: 62.3–98.4%); Edinburgh demonstrated the highest specificity (92.3%; 95% CI: 79.7–97.3%); ISS provided balanced performance (sensitivity 81.8%, specificity 84.6%); MHIPS demonstrated the lowest sensitivity in this mixed-severity cohort (63.6%/84.6%). The composite index was strongly associated with mortality (chi-square = 9.44, p = 0.002). This is the first prospective head-to-head comparison of NIMHANS, MHIPS, Edinburgh, and ISS in a single Indian TBI cohort. The NIMHANS model offers the most sensitive bedside screening tool; Edinburgh provides the strongest specificity. We propose a pragmatic two-step framework—NIMHANS for initial screening and Edinburgh for confirmatory risk stratification—for Indian neurosurgical practice, with multicenter validation as the immediate next step.