Retrospective validation of the STUMBL score in a Level 1 trauma centre

•This is the first study to our knowledge investigating the clinical utility of the STUMBL score (a risk prediction tool for chest trauma) in an Australian population.•The optimal STUMBL score for predicting complications in this study was 17.5 (AUROC of 0.76) which is similar to the original validation cohort, but contrasts with other studies.•This study identifies that in our population, utilising a STUMBL score of > 17 would help to support decision making for ICU admission from ED with an AUROC of 0.82, sensitivity 80.6 % and specificity 73.9 %.•This study suggests improvements in the STUMBL scores predictive power could be improved through incorporation of additional risk factors which have shown promise in the literature, mainly Glasgow coma score on admission, presence of chest tube, pulmonary contusion or frailty. Chest trauma is a common presentation to major trauma centres. Risk assessment tools have proven useful to support decision making in this group and the STUMBL (STUdy of the Management of BLunt chest wall trauma) score is one such measure that has been increasingly utilised. The aim of this study was to retrospectively validate the STUMBL score in an Australian population of patients admitted following chest trauma. Methods: A single-centre retrospective validation study was undertaken using information from all patients with an Emergency Department (ED) attendance for isolated blunt chest trauma at a major trauma centre in Australia from 2018. The performance of the STUMBL score was measured including the cut-off score which best predicted 1) the discharge disposition from ED (ward or intensive care unit [ICU]), 2) the development of pulmonary complications, 3) an extended length of stay (LOS) (7 days or more) and 4) any complication (pulmonary, extended LOS, in hospital mortality). The performance measures included sensitivity, specificity, negative and positive predictive values as well discrimination and calibration. Results: There were 300 patients admitted between 1st January 2018 and 31st December 2018 with a median age of 60 years (IQR 44–75) and 65 % were male. The risk prediction cut-off score for our patient cohort ranged from 18.5 for LOS 7 days or more to 11.5 for ward admission from ED. The positive predictive value (PPV) ranged from 56.7 % for ward admission from ED to 21.1 % for pulmonary complications. The negative predictive value (NPV) and sensitivity was highest for ICU admission from ED (96.5 % and 80.6 %) and the spe