Background Unstable distal clavicle fractures (UDCFs) exhibit a significant propensity for bone nonunion and are commonly managed through open reduction and internal fixation(ORIF). A notable drawback of this approach is the considerable incidence of fixation failure, including 3-month nonunion (also referred to as delayed union), and subsequent bone nonunion. Accurate prediction of union outcomes post-UDCF treatment could provide substantial evidence to inform and optimize intraoperative and postoperative strategies.
Methods Data were gathered from patients with UDCFs at Nanjing Luhe People's Hospital, China, from January 2015 to May 2023. The adverse outcome, characterized by the non-disappearance of the fracture line and uncontinuous callus formation, was defined as 3-month nonunion. To predict the risk of 3-month nonunion, five machine learning models were constructed, employing the area under the precision-recall curve (AUPRC) as the principal evaluation criterion. Subsequently, The Shapley additive explanations (SHAP) was further used to interpret the best model after comparing the models' performances.
Results In this investigation, a cohort of 248 patients was enrolled, among which 76 individuals (30.6%, median age: 55 years, males: 52) encountered adverse outcomes. The integration of variables such as the coracoclavicular distance(CCD), anesthesia methodology, high-density lipoprotein (HDL) levels, and blood loss revealed that the CatBoost model was markedly more effective compared to four other predictive models, achieving an AUPRC of 0.801 (95% CI: 0.592-0.918). The SHAP values identified the CCD as the paramount predictor for assessing the risk of 3-month outcomes in patients with UDCFs within the Chinese demographic.
Conclusion The refined model incorporated four readily accessible variables, wherein the CCD, HDL levels, and blood loss were associated with an elevated risk of 3-month nonunion. Moreover, the CatBoost model has demonstrated a commendable capability in enhancing the prediction of 3-month outcomes for Chinese patients with UDCFs, effectively distinguishing individuals at high risk for delayed nonunion. This enhancement is pivotal for reinforcing intraoperative and postoperative management, thereby ameliorating patient prognoses.