DATASET: Data-Driven Clinical Decision Trees and Generative Topographic Mapping to Guide Stroke Rehabilitation Program Allocation

A simulated and expert-annotated dataset of 700 stroke patients characterized by 49 demographic, clinical, and functional variables, we modeled real-world heterogeneity across seven rehabilitation programs ranging from home-based to high-intensity hospital-based care. Each patient of the dataset was profiled by forty-nine parameters in total. The dataset was subsequently reviewed and annotated in a structured, multi-stage approach, by an expert panel of four attending physicians.The .ppt file in the directory is the in-house tree that it is currently used at the institution (Hospital del Mar, Barcelona, Spain), and it is based on national stroke rehabilitation guidelines. The in-house decision tree utilizes as predictor variables the NIHSS, MoCA score, mRS, patient age, caregiver presence, and the Charlson Index.