Abstract
Background: Voluntary turnover (VTO) of nursing employees is expensive for hospital systems and is often associated with lower levels of patient satisfaction, as well as adverse patient outcomes such as falls and medication errors.
Purpose: The aim of this study was to establish nurses' electronic medical record (EMR) use patterns and test if they can be used to predict VTO.
Methodology/Approach: The study followed 1,836 hospital nurses via the collection of EMR metadata through two 1-month time periods that were 1 year apart. Machine learning algorithms were then used to derive patterns of EMR utilization using VTO as a key variable for classification. Post hoc analysis of the most predictive variables was conducted.
Results: The predictive model was effective in identifying which nurses would turnover 73.4% of the time and which nurses would not turnover 84.1% of the time.
Practice Applications: The ability to accurately predict nurses' intentions to leave is critical to reducing turnover. Early identification can lead to specific interventions to mitigate factors that are adversely impacting the nursing experience. Post hoc analysis and the key informant interviews indicated that many nurses do not appear to have good EMR navigation skills and spend significant effort in search of patient information.