Fig. 1

Variable importance of baseline variables in the cluster analysis. The corresponding domain of each variable is indicated following the framework of The International Classification of Functioning, Disability, and Health. The variable importance was derived from the mean square error with higher values indicating higher importance. Two clusters were determined as an optimal number of clusters on the basis of silhouette width (0.48). Stability of the clusters was assessed using the Jaccard similarity through resampling of the data 500 times. The Jaccard similarity was 0.97, which indicates stable clusters. BBS, Berg Balance Scale; BI, Barthel Index; FMA, Fugl-Meyer Assessment; HADS, Hospital Anxiety and Depression Scale; NIHSS, National Institutes of Health Stroke Scale; IQR, interquartile range; LE, lower extremity; MoCA, Montreal Cognitive Assessment; mRS, modified Rankin Scale; ROM, passive joint motion; SD, standard deviation; TUG, timed up-and-go test; UE, upper extremity