Return-to-Work Predictions for Chinese Patients With Occupational Upper Extremity Injury: A Prospective Cohort Study

Abstract

Objective: We created predictive models using machine learning algorithms for return-to-work (RTW) in patients with traumatic upper extremity injuries.

Methods: Data were obtained immediately before patient discharge and patients were followed up for 1 year. K-nearest neighbor, logistic regression, support vector machine, and decision tree algorithms were used to create our predictive models for RTW.

Results: In total, 163 patients with traumatic upper extremity injury were enrolled, and 107/163 (65.6%) had successfully returned to work at 1-year of follow-up. The decision tree model had a lower F1-score than any of the other models (t values: 7.93-8.67, p < 0.001), while the others had comparable F1-scores. Furthermore, the logistic regression and support vector machine models were significantly superior to the k-nearest neighbors and decision tree models in the area under the receiver operating characteristic curve (t values: 6.64-13.71, p < 0.001). Compared with the support vector machine, logistical regression selected only two essential factors, namely, the patient's expectation of RTW and carrying strength at the waist, suggesting its superior efficiency in the prediction of RTW.

Conclusion: Our study demonstrated that high predictability for RTW can be achieved through use of machine learning models, which is helpful development of individualized vocational rehabilitation strategies and relevant policymaking.

Keywords: machine learning; occupational health; return-to-work; support vector machine; upper extremity injury; vocational rehabilitation.

Figure 1

Comparison on F1-scores of the four models. The left histograms show the distribution of the F1-score, and the right bar chart shows a direct comparison on the F1-scores of kNN (0.816 ± 0.041), Log (0.820 ± 0.044), SVM (0.823 ± 0.044) and DT (0.774 ± 0.059). The error bars represent one standard deviation of uncertainty. kNN, k-nearest neighbors; Log, logistic regression; SVM, support vector machine; DT, decision tree.

Figure 2

Comparison on the areas under the ROC of the kNN (0.723 ± 0.064), Log (0.766 ± 0.054), SVM (0.766 ± 0.053) and DT (0.665 ± 0.070) and the effects of the number of trainings on F1-scores. The error bars in (A) represent one standard deviation of uncertainty. The shaded regions in (A,B) represent one standard error of the mean. kNN, k-nearest neighbors; Log, logistic regression; SVM, support vector machine; DT, decision tree; ROC, receiver operating characteristic curve.