Abstract Body (Do not enter title and authors here): Background
Heart disease is a leading cause of death for American Indians, who face barriers to early diagnosis and accessing cardiac care, particularly in rural areas. To address limited cardiac services for an Arizona tribe, we used task-shifting to train Community Health Representatives (CHRs) to obtain point-of-care screening echocardiography with artificial intelligence-assisted interpretation (POCSE-AI).
Research Question
Evaluate if a competency-based curriculum is an effective method of training CHRs to independently utilize POCSE-AI to detect heart disease.
Methods
A 14-image POCSE-AI protocol was created based on the most frequently reported cardiac diagnoses in the relevant population. A novel, strategically sequenced curriculum was developed including asynchronous online didactics, two-day hands-on workshop, and a six-month deliberate practice phase with expert mentorship at community-based screening events. CHRs advanced through the curriculum by completing targeted, mixed-methods competency assessments, including multiple choice questions and assessment of image acquisition skills and diagnostic image quality. Final assessments included a professional skills exam via simulated patient encounter and submission of 5 qualifying studies. Qualifying studies were defined as complete, independently obtained POCSE-AI protocols with AI-confirmation of diagnostic image quality for clinically significant echocardiography parameters (Figure 1).
Results
Ten CHRs began the curriculum in July 2024. CHRs were mostly female (80%) with median age 52 years (IQR 50-60) and median experience as CHRs 8 years (IQR 4-16); none had prior training in diagnostic ultrasound. All CHRs completed introductory online modules with median quiz scores of 100% (IQR 96-100). At end of workshop, 4 CHRs independently obtained the POCSE-AI protocol per bedside expert assessment, but none had image quality sufficient to meet criteria for AI-confirmed qualifying study. During deliberate practice phase, CHRs variably obtained POCSE-studies (<5 to 29 per CHR). To date, 7 CHRs have submitted qualifying exams and 2 CHRs have passed the final assessments.
Conclusion
Preliminary findings demonstrate a competency-based model can train dedicated novice learners, facilitate task-shifting and deploy use of POCSE-AI in underserved rural communities. Future work will include strategies to improve curriculum engagement and identification of trainees most likely to benefit from limited training resources.