AUTHOR=Bai Zhongfei , Zhang Jiaqi , Tang Chaozheng , Wang Lejun , Xia Weili , Qi Qi , Lu Jiani , Fang Yuan , Fong Kenneth N. K. , Niu Wenxin TITLE=Return-to-Work Predictions for Chinese Patients With Occupational Upper Extremity Injury: A Prospective Cohort Study JOURNAL=Frontiers in Medicine VOLUME=Volume 9 - 2022 YEAR=2022 URL=https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2022.805230 DOI=10.3389/fmed.2022.805230 ISSN=2296-858X ABSTRACT=Objective: This study aimed to establish predictive models for return-to-work (RTW) in patients after traumatic upper extremity injury and select relevant factors using machine learning algorithms. Methods: A total of 163 patients after traumatic upper extremity injury were enrolled. Patient data were obtained immediately before discharge, and they were followed up for one year. K-nearest neighbors, logistic regression, supportive vector machine and decision tree algorithms were used to create predictive models for RTW. Results: In total, 107 (65.6%) of them successfully returned to work at one-year follow-up. The decision tree model had a lower F1-score compared with any of other models (t values: 7.93 to 8.67, p values < 0.001), and the others had comparable F1-scores. Furthermore, the logistic regression and supportive vector machine models were significantly superior to the k-nearest neighbors and decision tree models (t values: 6.64 to 13.71, p values < 0.001). Compared with supportive vector machine, logistical regression selected two essential factors only, namely, patient’s expectation to RTW and carrying strength to waist, suggesting its superior efficiency in the prediction for RTW. Conclusion: Our study demonstrated that high predictability for RTW can be achieved by using machine learning models, which is helpful to the development of individualized vocational rehabilitation plan and relevant policy making.