Opinion versus practice regarding the use of rehabilitation services in home care: an investigation using machine learning algorithms

• DOI: 10.1186/s12911-015-0203-1
• Publication Date: 2015-10-09T12:09:09Z
• Authors: Lu Cheng; Mu Zhu; Jeffrey W. Poss; John P. Hirdes; Christine Glenny; Paul Stolee
• Keywords: Male; Health Policy; Rehabilitation; Health Informatics; Middle Aged; Home Care Services; 01 natural sciences; Decision Support Techniques; 3. Good health; Machine Learning; Data Mining; Humans; Female; Medical Informatics Applications; Mobility Limitation; 0101 mathematics; Algorithms; Research Article

Resources for home care rehabilitation are limited, and many clients who could benefit do not receive therapy. The interRAI Contact Assessment (CA) is a new screening instrument comprised of subset Home Care (HC) items, designed to be used as preliminary assessment identify which potential should referred full assessment, or services such rehabilitation. We investigated client characteristics most relevant in predicting use the HC assessment. applied two algorithms from machine learning data mining ― LASSO random forest frequency matched service utilization Ontario, Canada. Analyses confirmed importance functional decline mobility variables targeting services, but suggested that other items predictors may less relevant. Six highly ranked related ambulation. Diagnosis cancer was associated with decreased use; however, cognitive status not. Inconsistencies between considered important classifying need those identified this study based on indicate discrepancy theory versus actual practice.