Systematic review: The use of large language models as medical chatbots in digestive diseases.

BACKGROUND: Interest in large language models (LLMs), such as OpenAI's ChatGPT, across multiple specialties has grown as a source of patient-facing medical advice and provider-facing clinical decision support. The accuracy of LLM responses for gastroenterology and hepatology-related questions is unknown. AIMS: To evaluate the accuracy and potential safety implications for LLMs for the diagnosis, management and treatment of questions related to gastroenterology and hepatology. METHODS: We conducted a systematic literature search including Cochrane Library, Google Scholar, Ovid Embase, Ovid MEDLINE, PubMed, Scopus and the Web of Science Core Collection to identify relevant articles published from inception until January 28, 2024, using a combination of keywords and controlled vocabulary for LLMs and gastroenterology or hepatology. Accuracy was defined as the percentage of entirely correct answers. RESULTS: Among the 1671 reports screened, we identified 33 full-text articles on using LLMs in gastroenterology and hepatology and included 18 in the final analysis. The accuracy of question-responding varied across different model versions. For example, accuracy ranged from 6.4% to 45.5% with ChatGPT-3.5 and was between 40% and 91.4% with ChatGPT-4. In addition, the absence of standardised methodology and reporting metrics for studies involving LLMs places all the studies at a high risk of bias and does not allow for the generalisation of single-study results. CONCLUSIONS: Current general-purpose LLMs have unacceptably low accuracy on clinical gastroenterology and hepatology tasks, which may lead to adverse patient safety events through incorrect information or triage recommendations, which might overburden healthcare systems or delay necessary care.

Infection prevention and control and related practices in African neonatal units: The Pan-African neonatal care assessment study (PANCAS).

BACKGROUND: The burden of neonatal mortality is primarily borne by low- and middle-income countries (LMICs), including deaths due to healthcare-associated infections (HAIs). Few studies have assessed infection prevention and control (IP&C) practices in African units caring for small and/or sick newborns aimed to reduce HAIs. METHODS: We performed a mixed-methods study composed of a survey and virtual tour to assess IP&C and related practices. We created a survey composed of multiple-choice and open-ended questions delivered to site respondents via Zoom or video equivalent. Respondents provided a virtual tour of their unit via video and the study team used a checklist to evaluate specific practices. RESULTS: We recruited 45 units caring for small and sick newborns in 20 African countries. Opportunities to optimize hand hygiene, Water, Sanitation and Hygiene (WASH) practices, Kangaroo Mother Care, and IP&C training were noted. The virtual tour offered further understanding of IP&C challenges unique to individual sites. All respondents expressed the need for additional space, equipment, supplies, education, and IP&C staff and emphasized that attention to maternal comfort was important to IP&C success. DISCUSSION: This study identified opportunities to improve IP&C practices using low-cost measures including further education and peer support through learning collaboratives. Virtual tours can be used to provide site-specific assessment and feedback from peers, IP&C specialists and environmental engineering experts.

Underrepresentation of blind and deaf participants in the All of Us Research Program.

Blind and deaf individuals comprise large populations that often experience health disparities, with those from marginalized gender, racial, ethnic and low-socioeconomic communities commonly experiencing compounded health inequities. Including these populations in precision medicine research is critical for scientific benefits to accrue to them. We assessed representation of blind and deaf people in the All of Us Research Program (AoURP) 2018-2023 cohort of participants who provided electronic health records and compared it with the Centers for Disease Control and Prevention 2018 national estimates by key demographic characteristics and intersections thereof. Blind and deaf AoURP participants are considerably underrepresented in the cohort, especially among working-age adults (younger than age 65 years), as well as Asian and multi-racial participants. Analyses show compounded underrepresentation at the intersection of multiple marginalization (that is, racial or ethnic minoritized group, female sex, low education and low income), most substantively for working-age blind participants identifying as Black or African American female with education levels lower than high school (representing one-fifth of their national prevalence). Underrepresentation raises concerns about the generalizability of findings in studies that use these data and limited benefits for the already underserved blind and deaf populations.