Abstract
<title>Abstract</title> <p>Background Mild traumatic brain injury (MTBI) is a leading cause of emergency department visits among adults aged ≥ 75 years. Current guidelines recommend liberal use of cranial computed tomography (CT), although clinically significant intracranial haemorrhage (ICH) remains relatively uncommon. Improved risk stratification may optimize imaging in this population. Objective To identify predictors of traumatic ICH after MTBI in older adults and derive a pragmatic bedside clinical prediction score. Methods We conducted a retrospective single-centre cohort study at Beaujon University Hospital between October 2019 and November 2021. Patients aged ≥ 75 years who underwent cranial CT following MTBI were included. Independent predictors of ICH were identified using multivariable logistic regression. Model performance was assessed by receiver operating characteristic (ROC) analysis. A weighted clinical prediction score (FILM-G) was derived from regression coefficients and internally validated using 1,000 bootstrap resamples. Results Among 1,234 screened patients, 566 met inclusion criteria. ICH was identified in 59 patients (10.4%), whereas neurosurgical intervention was required in only two cases. Independent predictors of ICH included loss of consciousness (adjusted odds ratio [aOR] 3.16, 95% CI 1.47–6.81), GCS < 14 (aOR 6.16, 95% CI 2.53–15.01), focal neurological deficit (aOR 5.06, 95% CI 1.72–14.94), and frontal, parietal, temporal, and occipital impact locations. Combining neurological findings, impact location, and male sex yielded the FILM-G model, which demonstrated acceptable discrimination (AUC 0.77, 95% CI 0.71–0.83) with limited overfitting after bootstrap validation. Conclusions Neurological abnormalities and cutaneous impact location are strong predictors of ICH in older adults with MTBI. The FILM-G score provides a simple bedside tool that may help optimize CT utilization while maintaining patient safety. External validation is warranted before clinical implementation.</p>