Uncovering complex correlations between multidimensional factors and breastfeeding duration using XGBoost.

Evidence-based medicine is the practice of integrating clinical judgment with current best evidence, and is essential for making informed clinical decisions and improving patient care.[–] In 2010, 75 randomized controlled trials (RCTs) and 11 systematic reviews were published daily, with no anticipated plateau in the projected growth of publications.[] Nearly 10 years later, the number of RCTs and systematic reviews published each day were estimated to be approximately 140 and 80, respectively. [–] This volume of publications far exceeds the available time of clinicians and medical trainees. For trainees in particular, it can be overwhelming to become familiar with landmark, practice-setting RCTs while concurrently assimilating emerging evidence into their clinical workflows.[,,] According to a study of American Emergency Medicine resident physicians, keeping up with medical literature was reported to be one of their most significant day-to-day challenges.[] While clinicians commonly access resources such as UpToDate or MedScape to find summaries of the existing literature, their content may exclude recently published trials and the cumbersome nature of their interfaces is not always user-friendly.[–] While systematic reviews can help keep clinicians up to date, they commonly lag behind the most recent RCTs by one year, are time-consuming to review, and are typically outdated by the time of publication.[,] Evidently, novel strategies to collect, summarize, and share clinical research are needed. Artificial intelligence (AI) has the potential to modernize medical practice, including medical education, clinical decision-making, and healthcare research.[–] In medical education, specific applications of AI-based large language models, such as ChatGPT, include summarizing medical research, generating realistic patient simulations, personalizing learner experiences, and enhancing medical textbooks.[,] Additionally, the integration of AI tools into medical education programs can improve students’ digital and AI literacy skills.[] A cross-sectional study of graduates from an international medical school showed that 63% of medical trainees planned to use AI during residency to explore new medical topics and research.[] Accessible resources that inform clinicians of new research are needed. Our objective was to evaluate the accuracy and usefulness of a large language model to generate a newsletter summarizing results from RCTs. This study represents a prospective implementation and evaluation of an LLM-assisted knowledge-translation system. We prospectively created and evaluated a newsletter summarizing RCTs relevant to general internal medicine using a large language model. We selected the following journals to identify RCTs: New England Journal of Medicine, Annals of Internal Medicine, Journal of the American Medical Association (JAMA), JAMA Internal Medicine, and The Lancet. These were selected because they commonly publish RCTs relevant to general internal medicine. This focused selection was intended to prioritize high-relevance trials for early implementation of the PaperScrape tool. The primary evaluation outcome was the accuracy of LLM-generated summaries compared with source abstracts. Secondary outcomes included subscriber growth and user-reported usefulness.