Machine Learning Algorithm Identifies the Importance of Environmental Factors for Hospital Discharge to Home of Stroke Patients using Wheelchair after Discharge

Abstract

Background and purposePhysical environmental factors are generally likely to become barriers for discharge to home of wheelchair users, compared with non-wheelchair users. However, the importance of environmental factors has not been investigated adequately. Application of machine learning technology might efficiently identify the most influential factors, although it is not easy to interpret and integrate various information including individual and environmental factors in clinical stroke rehabilitation. This study aimed to identify the influential factors affecting home discharge in the stroke patients who use a wheelchair after discharge by using machine learning technology. MethodsThis study used the rehabilitation database of our facility, which includes all stroke patients admitted into the convalescence rehabilitation ward. The chi-squared automatic interaction detection (CHAID) algorithm was used to develop a model to classify wheelchair-using stroke patients discharged to home or not-to-home. ResultsAmong the variables, including basic information, motor functional factor, activities of daily living ability factor, and environmental factors, the CHAID model identified house renovation and the existence of sloping roads around the house as the first and second discriminators for home discharge. ConclusionsOur present results could scientifically clarify that the clinician need to focus on the physical environmental factors for achieving home discharge in the patients who use a wheelchair after discharge.