Predicting Rehabilitation Potential with the K-Nearest Neighbors Algorithm: A Comparison with the Current Clinical Assessment Protocol

M. Zhu, W. Chen, J. Hirdes, and P. Stolee (Canada)

Keywords

Bayes’ theorem; Binary prediction; Diagnostic likelihood ratio; interRAI minimum data set; Machine learning; Mod elling and simulation methodologies.

Abstract

Using data from eight Community Care Access Centres (CCACs) in Ontario, we demonstrate that an automatic, data-driven, machine learning algorithm such as the K nearest neighbors (KNN) algorithm can predict rehabil itation potential more effectively than the current Clini cal Assessment Protocol (CAP). Implications for clinical decision-making and computerized health information sys tems are discussed.

Important Links:

• DOI:
• From Proceeding (530) Modelling and Simulation - 2006

Go Back