Can Machine Learning Identify Patients Who are Appropriate for Outpatient Open Reduction and Internal Fixation of Distal Radius Fractures?

PURPOSE

This study aimed to identify which patients were "unsafe" for outpatient surgery patients and determine the most predictive demographic and clinical factors contributing to postoperative risk following open reduction internal fixation for distal radius fractures.

METHODS

Adult patients (aged ≥18 years) who presented with distal radius fracture and underwent open reduction internal fixation were identified using the American College of Surgeons National Surgical Quality Improvement Program database for years 2016 to 2021. Patients who were deemed "unsafe" therefore contraindicated for outpatient open reduction internal fixation of distal radius fracture if they required admission (length of stay of one or more days) or experienced any complication or required readmission within 7 days of the index operation. The model with optimal performance was determined according to area under the curve on the receiver operating characteristic curve and overall accuracy. Additional model metrics were also evaluated, and predictive factors (ie, features) that were most important to model derivation were identified.

RESULTS

A total of 2,020 eligible patients underwent open reduction and internal fixation for distal radius fractures. The majority (78.6%) were women, with a mean age of 57.5 ± 16.0 years. Of these patients, 21.5% experienced short-term adverse events. Gradient boosting was the optimal model for predicting patients who were "unsafe" for outpatient surgery, with key features including International Classification of Diseases, 10th Revision code, preoperative white blood cell count, age, body mass index, and Hispanic ethnicity.

CONCLUSIONS

Using machine learning techniques, a predictive model was developed, which demonstrated good discrimination and excellent performance in predicting which patients were "unsafe" for outpatient operative fixation of distal radius fracture. Findings of this study highlight the predictive value of artificial intelligence and machine learning for the purposes of preoperative risk stratification as well as its potential to better inform shared decision making and guide personalized fracture care.

LEVEL OF EVIDENCE/TYPE OF STUDY: Prognostic IV.