Abstract
Objective: This exploratory study examined whether the Berg Balance Scale (BBS) and Trunk Impairment Scale (TIS) provide additional predictive information beyond a prespecified basic clinical model for achieving an individual-level performance index equivalent of ≥48 in patients with stroke admitted to a convalescent rehabilitation ward.
Methods: Participants were 150 patients. A basic clinical logistic regression model included age, days from stroke onset to admission, pre-admission modified Rankin Scale, and Functional Independence Measure (FIM) cognitive score. Three additional models were prespecified by adding BBS, TIS, or both to the basic clinical model. Changes in predictive performance relative to the basic clinical model were descriptively evaluated using the area under the receiver operating characteristic curve (AUC), Brier score, calibration slope, and 1,000 bootstrap resamples. A sensitivity analysis incorporated the admission FIM motor score.
Results: The optimism-corrected AUCs for the basic clinical, BBS-added, TIS-added, and BBS-plus-TIS models were 0.733, 0.772, 0.784, and 0.781, respectively; the corresponding optimism-corrected Brier scores were 0.204, 0.187, 0.182, and 0.183. In the sensitivity analysis, models adding BBS, TIS, or both had numerically higher optimism-corrected AUCs and lower optimism-corrected Brier scores than the reference model.
Conclusion: BBS and TIS may provide additional predictive information beyond the basic clinical model. This study did not evaluate superiority among the additional models. External validation is required before clinical application.
