JMIR Publications

ABSTRACT

Background:

The integration of artificial intelligence (AI) into clinical practice is transforming the field of medicine and medical education. AI-based systems aim to improve the efficacy of clinical tasks, such as disease diagnosis and treatment delivery. In addition, it is critical to use AI-based systems responsibly to ensure effectiveness, mitigate bias, and maintain safe clinical practices. However, the utilization of AI-based systems to enhance clinical decision making in medical education remains understudied. In the context of identifying patients with heart failure (HF), AI-based HF screening tools encounter various challenges. These include precise detection of the subtle progression of HF, and the ability to present information in a manner that healthcare providers and learners can use to support their clinical decision-making. Although some research exists on AI-enabled clinical decision support systems, further investigations are needed to explore effective educational approaches for using AI to augment clinical decision-making.

Objective:

This study aims to explore the use of a web-based educational tool to augment medical students’ ability to recognize patients with HF within a prototype clinical context: preoperative evaluation for major surgery. Our study measures the effectiveness of interventions incorporating machine learning algorithms and HF guidelines implemented in the tool to improve the accuracy of recognizing patients with HF before surgery.

Methods:

This randomized trial with a 3x2 factorial design consists of three interventions and changing the order of two sets of surgical cases in pre- and post-tests. The study participation involves two components: 1) a 30-minute e-learning module on using the educational tool, and key information in the intervention, followed by a 5-question quiz on intervention knowledge, and 2) a 60-minute review of 20 surgical cases, which were extracted from electronic health records and adjudicated by a consensus panel of heart failure experts. In the study analysis, the initial ten cases were part of the pre-test phase, where no intervention was applied, and the remaining ten cases were included in the post-test phase, after intervention was introduced. Three interventions were evaluated: two machine learning-based interventions and one HF guideline-based intervention.

Results:

The subject enrollment commenced in September 2022 and will end in December 2023, with the goal of recruiting 75 medical students in Years 3 and 4 with clinical experience. This study will measure the accuracy of students’ HF recognition before and after educational intervention access.

Conclusions:

This AI- and HF guideline-based educational tool can offer medical students opportunities to enhance their ability to recognize complicated diseases like HF using EHR data in a safe and secure environment.