Aims and Objectives The digitalisation of patient data and the adoption of electronic health records (EHRs) has changed the patient-doctor encounter. A growing body of evidence suggests EHRs reduce efficiency and negatively impact cognitive load, ultimately reducing the number of patients seen by clinicians and increasing the chances of introducing medical error. This feasibility study aims to investigate the effects of an artificial intelligence co-pilot called OSLER on consultation time and cognitive load.

Method and Design Physicians undertook simulated clinical scenarios (with and without OSLER) using an unfamiliar EHR system, OpenEMR. The AI co-pilot was trained to autonomously navigate OpenEMR on voice command and provide context-specific AI tools (notes summarisation, speech-to-intent, speech-to-text, and automated letter composition). Measures of time taken (mins : secs), number information technology (IT) support calls, quality of documentation (using a visual analogue scale), and cognitive load (NASA Task Load Index) were recorded and analysed.

Results and Conclusion Total of 19 physicians undertook the simulated clinical scenarios. OSLER reduced consultation time by an average of 51% (15:02 ± 04:28 vs 07:25 ± 02:20, p < 0.001). There was significant reduction in the mean number of information technology support calls made with Osler (2 ± 1 vs 0.3 ± 0.48, p < 0.001). Quality of documentation improved with OSLER 4.3 ± 0.3 vs 2.7 ± 0.9 (p < 0.001). The mean NASA TLX score using OSLER was 2.9 ± 1.1 (95% CI 2.3 – 3.5 vs without OSLER 6.9 ± 0.8 (95% CI 6.5 – 7.4), p = 0.001.

This study demonstrates that the implementation an AI co-pilot, significantly reduces consultation time, IT support calls, and cognitive load while improving the quality of documentation for clinicians using OpenEMR. Our findings highlight the potential of AI and natural language processing tools to streamline and automate EHR documentation within the emergency department.