Relevance ranking of intensive care nursing narratives

Abstract

Current computer-based patient records provide many capabilities to assist nurses' work in intensive care units, but the possibilities to utilize existing free-text documentation are limited without the appropriate tools. To ease this limitation, we present an adaptation of the Regularized Least-Squares (RLS) algorithm for ranking pieces of nursing notes with respect to their relevance to breathing, blood circulation, and pain. We assessed the ranking results by using Kendall's τb as a measure of association between the output of the RLS algorithm and the desired ranking. The values of τb, were 0.62, 0.69, and 0.44 for breathing, blood circulation, and pain, respectively. These values indicate that a machine learning approach can successfully be used to rank nursing notes, and encourage further research on the use of ranking techniques when developing intelligent tools for the utilization of nursing narratives.