ChatGPT & CTIs: A Study Using Artificial Intelligence in Hospice.

1. Employing the available literature, participants will self-report the ability to contextualize the potential applications of ChatGPT in hospice and comprehend the current ethical and statutory limitations facing its use and development. 2. Analyzing the referenced study, participants will self-report the ability to incorporate ChatGPT within the hospice recertification process and dispute the utility of integrating and advocating for such technology into similar compliance processes. Certificates of Terminal Illness (CTIs) generated with ChatGPT showed similar quality and acceptability by Medicare standards compared to physicians. The highly standardized format of CTIs may have contributed to these positive results. With further improvements, ChatGPT could augment clinical workflow in hospice. ChatGPT is a Large Language Model which has gained significant popularity since its launch in November 2022. Few studies have investigated the use of ChatGPT in clinical workflow, but none have investigated its use in documenting assessments. Meanwhile challenges exist, including ethical concerns and the dynamic regulatory landscape surrounding its use. Within hospice, Certification of Terminal Illnesses (CTIs) are short narrative documents summarizing care and patient trajectory in the last few months. We aimed to compare the quality of CTIs between ChatGPT and physicians. Standardized nursing recertification notes were used for two hospice medical directors to write physician CTIs. The same notes were deidentified and entered into ChatGPT with a uniform prompt to generate ChatGPT CTIs. A mix of randomly selected physician and ChatGPT CTIs were sent to the blinded panel of hospice medical directors. The panel evaluated each CTI on acceptability based on Medicare standards, overall quality (5-point Likert scale), and authorship (physician vs ChatGPT vs not sure). Preliminary results included 50 CTIs (20 Physician, 30 ChatGPT). Seventy-five percent of Physician CTIs were acceptable by Medicare standards, compared to 70% of ChatGPT CTIs (p = 0.703). The average quality of physician CTIs was between fair and good (3.3), which was similar to the average quality of ChatGPT (3.6; p = 0.532). The panel correctly identified the authorship in 65% of physician CTIs, and 53% of ChatGPT CTIs. ChatGPT produced hospice assessment documentations with comparable quality to physicians. This is most likely due to the summative format of the documentation used to construct CTIs and may not be demonstrated for other types of documentation. Nonetheless, these results align with current literature suggesting the use of Artificial intelligence (AI) may augment clinical workflow. Innovative Technologies / Ethical / Legal Aspects of Care [ABSTRACT FROM AUTHOR]