The number of studies on the characteristics of patients with stroke who would benefit from robot-assisted upper limb rehabilitation is limited, and there are no clear criteria for determining which individuals should receive such treatment. The current study aimed to develop a clinical prediction rule using machine learning to identify the characteristics of patients with stroke who can the achieve minimal clinically important difference of the Fugl-Meyer Upper Extremity Evaluation (FMA-UE) after single-joint hybrid assistive limb (HAL-SJ) rehabilitation.

This study included 71 patients with subacute stroke who received HAL-SJ rehabilitation. The chi-square automatic interaction detector (CHAID) model was applied to predict improvement in upper limb motor function. Based the analysis using CHAID, age, sex, days from stroke onset to the initiation of HAL-SJ rehabilitation, and upper limb motor and cognitive functions were used as independent variables. Improvement in upper limb motor function was determined based on the minimal clinically important difference of the FMA-UE, which was used as a dependent variable.

According to the CHAID model, the FMA-UE score during the initiation of HAL-SJ rehabilitation was the most significant predictive factor for patients who are likely to respond to the intervention. Interestingly, this therapy was more effective in patients with moderate upper limb motor dysfunction and early initiation of HAL-SJ rehabilitation. The accuracy of the CHAID model was 0.89 (95% confidence interval: 0.81–0.96).

We developed a clinical prediction rule for identifying the characteristics of patients with stroke whose upper limb motor function can improve with HAL-SJ rehabilitation.