Clinical predictors of mortality and functional outcomes following moderate-to-severe traumatic brain injury: A systematic review and meta-analysis

Background Early reliable prognosis following moderate to severe traumatic brain injury (TBI) can guide clinical decision-making and family counselling. This systematic review and meta-analysis aimed to review predictors of mortality and functional outcomes and assess existing prognostic models. Objective To identify predictors of mortality and functional outcomes and to determine the discriminatory ability of existing prognostic models in moderate-to-severe TBI patients. Methods PubMed and Europe PMC were searched from inception to 10 September 2026 along with the ClinicalTrials.gov registry for original prognostic studies of patients with moderate-to-severe TBI. Restricted maximum-likelihood random-effects models with modified Hartung–Knapp–Sidik–Jonkman confidence intervals were used to synthesize adjusted predictor estimates and model-discrimination measures separately. Risk of bias was assessed using QUIPS or PROBAST, as appropriate. Results Twenty-three studies were included, 10 of which were unique studies and led to seven quantitative syntheses. Higher age was associated with increased mortality (OR per year 1.046, 95% CI 1.034–1.059; I²=32%) and higher admission Glasgow Coma Scale scores predicted lower mortality (OR per point 0.811, 95% CI 0.786–0.838; I²=0%). There was no clear association between GCS and unfavorable functional outcome (OR 0.881, 95% CI 0.678–1.144) or decreasing platelet count with mortality (OR 1.105, 95% CI 0.795–1.538). Any abnormal pupillary findings resulted in a very rough estimate. The pooled discrimination was 0.805 for IMPACT and 0.822 for CRASH. Overall, six studies had low, 13 moderate, and four high risk of bias. Conclusions Age and admission GCS were common predictors of mortality. The general discrimination of IMPACT and CRASH was good with some uncertainty. The prognosis should be based on clinical, neurological, radiological, and laboratory data.