(1489) Eye Tracking for Usability of Identifying Airways in Electronic Health Records

Introduction: Effective identification of patients with a history of a difficult or critical airway (DA/CA) improves situational awareness. The electronic health record (EHR) may contain this information but the usability of EHR-based CA/DA decision support remains unknown. Eye tracking can augment EHR usability testing by unobtrusively analyzing search patterns and fixations. In this feasibility study we hypothesized that eye tracking metrics, combined with other usability metrics, will significantly differ between two wireframe EHR-based CA/DA decision support tools.

Methods: In a prior study, we identified the two most common EHR-based CA/DA decision support tools: Flag and Banner. We developed wireframe EHR simulations of these decision support tools. Participants were seated in front of a monitor mounting a Tobii Pro Spark eye tracking device. They completed two simulation tasks, one with each tool, to prepare for a simulated intubation. In each task we calculated fixations (gaze lasting ≥250 ms) and saccades (eye movements >35°/sec) along with traditional usability measures. We report median values and differences tested with Wilcoxon rank-sum.

Results: We developed an open-source Python-based software suite to run EHR usability experiments and analyze eye tracking metrics. Nine participants began the simulation and 8 (88%) successfully completed the task. Participants spent 63% less overall time on the banner task (11 vs 31 sec, p<.05) and visited 8.5 fewer pages (3 vs 12, p=.15) compared to the Flag wireframe. On the banner task there were 86% fewer fixations (3.5 vs 21.5, p<.05) and 60% less total fixation time (11 vs 26 sec, p=.19). Although the banner task resulted in 80% fewer saccades, these differences were not significant (p=.19). Average fixation duration and saccade velocity were not significantly different. Participants in the flag wireframe spent 40% more time looking at screen regions with unnecessary information (51% vs 31%, p<.05).

Conclusions: We demonstrate the development of a software package to deliver and analyze eye tracking usability data within an EHR simulation. We also demonstrate significant differences in eye tracking metrics, as well as traditional usability metrics such as completion time, between two EHR-based CA/DA decision support tools.