INTRODUCTION: Artificial intelligence (AI) has been shown to positively impact medicine and healthcare through improving physician efficiency and accuracy, streamlining workflows, and improving patient health literacy. ChatGPT has become a valuable, free, and easily accessible tool both for healthcare providers and for the public. Endometrial biopsies are common outpatient gynecologic procedures as part of the evaluation of abnormal uterine bleeding. With increased accessibility of patients viewing results independently on online charts and portals, patients often view pathology reports directly and before a healthcare provider can interpret results. This study is designed to evaluate faculty and trainee satisfaction with artificial intelligence-derived, patient-focused interpretations of benign endometrial biopsy pathology reports. OBJECTIVE: To examine the perceived utility of AI-derived explanations of benign endometrial biopsy results among academic OBGYN faculty physicians and trainees. To compare the AI tool’s perceived utility between trainee and faculty physicians. METHODS: 26 physicians who perform endometrial biopsies within the Department of Obstetrics and Gynecology were included in the study. 69% of participants were residents, 12% were fellows, and 19% were faculty physicians. No identifiable patient or participant information was collected. A brief ChatGPT-derived explanation of various benign endometrial biopsy results was provided to participants. A 15-question survey was then administered detailing providers’ anticipated use and application of the tools. Survey answers were recorded using a 5-point Likert scale, with the top two positive responses (“agree” and “strongly agree”) compared with the remaining three responses (“neutral,” “disagree,” and “strongly disagree”). Descriptive statistics and Fisher’s exact test were used to calculate differences in survey answers. RESULTS: 92% of all participants felt that the ChatGPT explanations of benign endometrial biopsy results were accurate. 81% of all respondents felt that these tools would save time, while only 65% anticipated that the explanations would increase patient health literacy. Trainees were more likely than faculty members to feel that AI explanations would decrease pathology-related phone calls and visits (71% vs 20%; p=0.05). 81% of trainees vs 40% of faculty (p=0.1) anticipated that patients would feel comfortable with these descriptions explaining their biopsy results. Similarly, 81% of trainees vs 40% of attending physicians (p=0.1) predicted higher patient satisfaction with AI-derived explanations than traditional verbal or written explanations. All respondents reported reservations about using the tool in certain situations. 85% reported that they would not use these descriptions for patients who require further workup, and 62% reported that they would not use these for patients with perceived low health literacy. CONCLUSIONS: Most faculty and trainees view artificial intelligence-derived explanations of endometrial biopsy results as accurate and easy to use. Trainees were more likely than faculty to predict patient comfort with these tools, improved patient satisfaction, and their utility in streamlining related phone calls and visits.
Garfinkel et al. (Fri,) studied this question.