Long-term prediction of traumatic brain injury outcome using the CBI-M framework.

BACKGROUND

Long-term functional and vocational outcomes after traumatic brain injury (TBI) remain difficult to predict, limiting prognostic counselling and rehabilitation planning. We aimed to evaluate the Clinical-Biomarker-Imaging-Modifier (CBI-M) framework for predicting functional and employment outcomes up to 10 years following TBI.

METHODS

Retrospective cohort study of 2529 patients with TBI admitted for inpatient rehabilitation (1984-2024), contributing 5424 observations across 1, 2, 5 and 10 years post-injury. Prediction models used logistic regression with temporal validation, incorporating 21 CBI-M variables spanning clinical (post-traumatic amnesia (PTA), Glasgow Coma Scale (GCS)), imaging (9 CT variables) and modifier pillars (age, sex, pre-injury employment, mental health, extracranial injuries). Outcomes were Glasgow Outcome Scale-Extended (GOSE) functional outcomes (favourable outcome, return to participation, full recovery) and employment outcomes (any productive work, competitive work).

RESULTS

Mean age was 38.7 years (SD=18.5); 705 (27.9%) were female. Models demonstrated good discrimination (mean area under the curve (AUC) 0.720; range 0.635-0.863), with 60% achieving AUC ≥0.70. Employment outcomes were slightly more predictable than GOSE outcomes (mean AUC 0.740 vs 0.70). PTA dominated prediction across both outcome types (GOSE: 28.8% variable importance; employment: 24%) versus GCS (GOSE: 3.9%; employment: 3.3%). Pre-injury employment was the second most important predictor for vocational outcomes (16.9%) after PTA. The clinical pillar was the dominant per-feature contributor to both GOSE (69.9%) and employment (64.5%) predictions, while the modifier pillar contributed most in absolute terms (41.6% and 50.8%, respectively).

CONCLUSIONS

The CBI-M framework predicts functional and vocational outcomes up to 10 years post-TBI. PTA substantially outperforms GCS for long-term prognostication in TBI survivors, and vocational outcomes also depend on pre-injury social factors.