Background
Older adults represent an increasing proportion of trauma admissions, yet existing prognostic models focus almost exclusively on mortality. A comprehensive, multi-outcome model could enable early risk stratification and inform care planning.
Methods
We developed and validated the Older Trauma Outcome Prediction (OTOP) model using data from the Western Australia Trauma Registry (2000–2020) for adults aged 50 years or older admitted after external injury. Predictor variables—demographic, clinical, and injury-related—were identified from a systematic review. The cohort (n=105,972) was randomly divided into training (80%) and validation (20%) sets. Ridge logistic regression was used for binary outcomes (in-hospital mortality, surgery, intensive care unit [ICU] admission, and medical complications), multinomial logistic regression for discharge destination, and quasi-Poisson regression for hospital length of stay. Model performance was evaluated using the area under the curve (AUC), calibration, and error metrics.
Results
The mean age was 74.0 years, and 45.3% were men. The mortality model achieved excellent discrimination (AUC, 0.95; SD, 0.004), exceeding that of the Geriatric Trauma Outcome Score (AUC, 0.83). Surgical intervention and ICU admission were predicted with high accuracy (AUCs, 0.97 and 0.96, respectively). Prediction of length of stay showed moderate precision (mean absolute error, 4.9 days), while the medical-complication model had limited discrimination (AUC, 0.60). The discharge-destination model effectively distinguished home, rehabilitation, residential care, and death.
Conclusions
OTOP is the first validated multi-outcome prognostic model for older trauma patients, offering superior mortality prediction and clinically actionable estimates across multiple inpatient outcomes.