Examining the Impact of Interface Design and Nurse Characteristics on Satisfaction With Machine Learning Decision Support Explanations.

Abstract

Machine learning has the potential to drive the next generation of clinical decision support systems by identifying patterns in health data to enhance efficiency and safety. Explanatory information is intended to help clinicians understand the outputs of these complex systems. No studies have evaluated associations between display design strategies or nurse characteristics and nurse satisfaction with machine learning explanatory information. This gap leaves much unknown about designing explanatory displays that meet nurses' information needs, supporting effective use and adoption in practice settings. To address this, we aimed to describe nurses' satisfaction with explanatory information displays for machine learning clinical decision support, examine associations between the format and complexity of explanatory information and nurse satisfaction, and investigate the influence of nurse characteristics, such as numeracy and graphical literacy, on satisfaction. Our findings indicate that local feature-based explanatory information may not satisfy nurses' information needs, and that nurse age, artificial intelligence training level, and numeracy influence preferences. We found no significant effects of the format or complexity of explanatory displays on satisfaction. These insights into the usability of machine learning clinical decision support for nurses can inform the design of more effective displays.