Development and internal validation of a prediction model for in-hospital mortality in trauma patients with high-energy pelvic fractures admitted to a single major trauma centre.

BACKGROUND

Pelvic fractures, although relatively uncommon, are associated with high economic burden, morbidity, and mortality. Mortality is largely driven by severe associated injuries and high-energy mechanisms. While predictors of mortality are well established, their local applicability for the study institution has not been updated using site-specific data since an earlier regional study (2001-2008). As trauma management protocols have evolved, this study utilizes prediction model development and internal validation-adhering to the TRIPOD guidelines-to update the prognostic value of these established predictors within our institutional context by using site-specific data.

METHODS

Data from July 2010 - December 2022 were sourced from a Level I adult Major Trauma Centre's registry. The cohort included patients > =15 years old with Injury Severity Score (ISS)> 12 and pelvic fractures. In-hospital mortality was the outcome of interest. Model development utilized backward elimination for predictor selection into a multivariable logistic regression model, with internal validation via bootstrap methods. Model performance was assessed using the Brier scaled score, and discrimination (c-statistic), and calibration (calibration plot).

RESULTS

Out of 1564 included patients, 118 were non-survivors (mortality rate 7.5%). The optimism-adjusted prediction model identified ISS≥ 50 (OR 7.4), age≥ 65 (OR 6.1), and severe head injury (OR 3.6) as strong predictors of mortality. Additional predictors with ORs between 2-3 included ISS 25-49, shock on admission, direct transport from the scene of injury, and severe comorbidity. The model demonstrated good to excellent discrimination with an optimism-adjusted c-statistic of 0.88.

CONCLUSIONS

This study developed and internally validated a prediction model for in-hospital mortality in major trauma patients with high-energy pelvic fractures using recent, single-center data, identifying key predictors. Notably, the severity of the pelvic fracture itself was not an independent predictor, indicating that pelvic injuries act primarily as markers of overall systemic injury severity. While optimism-adjusted odds from internal validation were attenuated, the findings remain comparable to the broader literature. External validation is recommended to assess the model's transportability and broader applicability.