Prediction of Unfavorable Functional Status at Hospital Discharge in Older Adults with Traumatic Brain Injury : Model Development and Internal Evaluation

Background: Traumatic brain injury (TBI) in older adults is an increasing public health burden. An early risk assessment of poor functional outcome can help guide prognosis and communication. The objective of the present study was to develop and internally validate a prediction model for poor functional status at discharge from the hospital for older adults with TBI. Methods: A retrospective cohort study included 3,193 patients aged ≥65 years with TBI. The Glasgow Outcome Scale (GOS) was used to evaluate the functional status of the patient at hospital discharge and was dichotomized as favorable (GOS 4-5) or unfavorable (GOS 1-3). Logistic regression was used to analyze candidate predictors; then, backward selection with the Akaike Information Criterion was used for multivariable analysis. The area under the receiver operating characteristic curve (AUC) was used to measure discrimination, and the calibration intercept, calibration slope, and Brier score were used to measure calibration. Optimism-corrected performance was estimated using 1,000 bootstrapped resamples in the internal validation. Bootstrap inclusion frequencies were used to evaluate the predictor-selection stability. Results: Among 3,193 patients, 224 (7.0%) had poor functional outcomes. Warfarin use, motor weakness, hypotension, category of GCS, pupillary response, basilar skull fracture, acute subdural hematoma, contusion/intracerebral hematoma, intraventricular hemorrhage, and basal cistern obliteration were predictor variables retained in the final model. The model achieved an apparent AUC of 0.964 (95% CI, 0.948-0.979). After 1,000 bootstrap resamples, the optimism-corrected AUC was 0.962 with a calibration intercept of -0.001 and the calibration slope of 0.959. Conclusion: The model demonstrated strong discrimination with limited optimism after bootstrap internal validation. External validation in independent populations is required before clinical implementation.