Use of an artificial intelligence driven voice recognition platform for training communication skills in undergraduate primary care

Problem

The utility of virtual/augmented reality and artificial intelligence (AI) for developing communication skills in undergraduates are yet to be appraised. These technologies have potential for enhancing existing methods of education and building capacity for clinical training. In partnership with a provider of a generative AI-driven voice recognition platform for developing communication skills in healthcare professionals, we compared the effectiveness and cost-effectiveness of this platform with conventional methods of communication skills training in primary care undergraduate education.

Approach

During a one week, 3rd year undergraduate medical student primary care attachment (University of Nottingham), we implemented this technology using a randomised crossover-design, comparing half-day AI-based consultation skills training (AICST) with half-day simulation-based consultation skills training (SBCST) with actors. Pre-post test changes in self-reported consultation skills acquisition and self-efficacy were measured using 10-point linear scales, with consultation skills measurements anchored to the Calgary-Cambridge model. Aggregated mean scores for communication skills were used as the primary outcome. Additionally, student satisfaction was measured in post-test surveys. Mean differences in pre-post scores and mean differences between interventions were determined using paired t-tests. The costs of implementing AISCT and SBCST were calculated and compared as cost per student taught.

Findings

Provisional findings are presented: 391 students were enrolled on the attachment. 307 students (79%) completed SBCST pre-session surveys and, of these, 268 (87%) completed post-session surveys. For AICST, 337 (86%) completed pre-session surveys and, of these, 292 (87%) completed post-session surveys. Aggregated mean communication skills scores significantly improved with both methods of training: 6.36 pre-SBCST and 7.86 post-SBCST; 6.70 pre-AICST and 7.84 post-AICST (both p<0.001). The mean difference in communication skills acquisition was greater for SBCST than AICST (1.5 versus 1.14; p=0.045). Student satisfaction scores were generally high for both modalities, but were higher for SBCST than AICST (accessibility: 9.26 versus 8.76; timing: 9.03 versus 8.08; Engaging: 9.11 versus 7.45; well-structured: 9.26 versus 7.93; ILO attainment: 9.28 versus 8.26; all p<0.01). Mean score of students recommending AICST to others was 8.03 versus 9.34 for SBCST (p<0.01).The cost of delivering SBCST per student was £61.75 compared with £33.48 for AICST.

Consequences

Provisional findings suggest AICST was effective in improving communication skill acquisition for undergraduate 3rd year medical students and this was similar to the skills acquisition from SBCST. Furthermore, the cost of using AICST was substantially lower than SBCST. However, student satisfaction was greater for SBCST than AICST. AICST may facilitate enhancement of communication skills training in undergraduate medical students and provide a cost-effective opportunity to build capacity for training within education curricula. However, further research testing the generalisability of findings across other year and professional groups, and methods for improving student satisfaction of AICST is warranted.