Key points are not available for this paper at this time.
Poststroke spasticity reduces arm function and leads to low levels of independence. This study suggested applying machine learning from routinely available data to support the clinical management of poststroke spasticity. One hundred seventy-two patients with acute first-ever stroke were included in this prospective cohort study. Twenty clinical information and rehabilitation assessments were obtained to train various machine learning algorithms for predicting 6-mo poststroke spasticity defined by a modified Ashworth scale score ≥1. Factors significantly relevant were also defined. The study results indicated that multivariate adaptive regression spline (area under the curve value: 0.916; 95% confidence interval: 0.906-0.923), adaptive boosting (area under the curve: 0.962; 95% confidence interval: 0.952-0.973), random forest (area under the curve: 0.975; 95% confidence interval: 0.968-0.981), support vector machine (area under the curve: 0.980; 95% confidence interval: 0.970-0.989), and outperformed the traditional logistic model (area under the curve: 0.897; 95% confidence interval: 0.884-0.910) ( P < 0.05). Among all of the algorithms, the random forest and support vector machine models outperformed the others ( P < 0.05). Fugl-Meyer Assessment score, days in hospital, age, stroke location, and paretic side were the most important features. These findings suggest that machine learning algorithms can help augment clinical decision-making processes for the assessment of poststroke spasticity occurrence, which may enhance the efficacy of management for patients with poststroke spasticity in the future.
Chen et al. (Tue,) studied this question.