The Birmingham mandible and midface rules: final data analysis

Heuristics and predictor rules are a mainstay in the field of emergency medicine where the breadth of knowledge required of frontline staff can be significant. The Birmingham mandible and midface (BruMM) rules were previously developed to streamline decision making concerning plain film radiography in suspected facial fractures. Frontline clinicians within a large NHS foundation trust were asked to apply the BruMM rules to any suspected mandibular or zygomatic fracture, prompting them to consider plain film radiographs as the investigation of choice. Radiographs were then reviewed by members of the BruMM Rules Study Group to determine the presence or absence of fractures. Discriminant analysis was performed whereby binomial fracture outcomes were modelled using cross-validated least absolute shrinkage and selection operator (LASSO) regression. Custom analysis was then performed using a forward stepwise procedure, maximising the Jaccard index of a predictor against the target outcome. A total of 116 patients were recruited, of whom 28 (24.1%) demonstrated a fracture. Combining the results from the LASSO regression analysis and forward stepwise Jaccard procedure yielded a predictor rule with three mandibular and six zygomatic predictors. This composite BruMM-rules score demonstrated maximal sensitivity and specificity of 96.4% and 56.5%, respectively.Implementation of the BruMM rules in the current cohort would have avoided 49 plain film radiograph studies (43.4%), translating into potential cost savings of £8,356.86-£15,440.76 per annum.