The authors' point about #LLM's increasingly used to generate, manage, and analyze #EHR data is critical. Anyone else think we need a checklist on large language models and AI/ML use in epidemiology articles? For #reproducibility, LLMs are invisible to current guidance as far as I can tell, except for the journals that ask if used in manuscript. However, if an analyst codes while interactively chatting with an LLM, should that transcript be part of the manuscript for the sake of transparency and reproducibility? Perhaps a repository-linked LLM session would be an easy way to keep track of how LLMs have shaped an analysis. For example, at what point(s) might LLMs suggest/alter/misinterpret the analytical approach? Did the analyst ask why an approach was chosen and or what it's assumptions are? We could add a parallel column to all current checklists that asks: "Did you use an LLM or other AI/ML coding or writing assistant in completing this item?" (y/n/unsure). If yes or unsure, how?". This could be an easy add on to any existing checklist or future checklist. Would impact at least half of the statistical analyses published over the next 10 years, if not today. #epidemiology #RWD #RWE #AI #ML.
National initiative for Hepatitis C elimination, Senior Advisor to Dr Francis Collins former NIH Director, Senior Advisor, National Institute for Biomedical Imaging and Bioengineering, GenAI and Biomedical Research
🚀 **Publication Announcement** 🚀 We are delighted to announce the publication of an important Task Force report in *Value in Health* titled "Assessing Real-World Data From Electronic Health Records for Health Technology Assessment: The SUITABILITY Checklist: A Good Practices Report of an ISPOR Task Force." Key Highlights from the Report: 1. Framework for EHR Data Assessment: The report introduces a comprehensive framework for evaluating the suitability of electronic health record (EHR) data in health technology assessments (HTAs). This includes a two-component approach: data delineation and data fitness for purpose. 2. ISPOR SUITABILITY Checklist: A practical tool to guide analysts in assessing EHR data for HTAs. The checklist covers critical aspects such as data characteristics, provenance, governance, reliability, and relevance. 3. Challenges and Recommendations: The report addresses the inherent challenges of using EHR data, including data completeness, accuracy, and the transformation of unstructured information. It provides actionable recommendations for HTA agencies and policymakers to improve the quality and usability of EHR-derived data. 4. Future Directions: Discusses the potential impact of advanced technologies like large language models and generative artificial intelligence on the field of HTA, highlighting the need for continued innovation and adaptation. Co-Chairs: Rachael Fleurence, Ph.D., MSc and Scott Ramsey, MD, PhD Contributing Authors: This report is a collaborative effort by leading experts in the field, including Seamus Kent, PhD, Blythe Adamson, PhD, MPH, James Tcheng, MD, Ran Balicer, MD, PhD, MPH, Joseph Ross, MD, MHS, Kevin Haynes, PharmD, MSCE, Patrick Muller, PhD, Jon Campbell, MS, PhD, Elsa Bouée-Benhamiche, PharmD, RPh, and Sebastian Garcia Marti, MD, MSc. We extend our gratitude to the authors for their significant contributions and dedication. Their work is pivotal in advancing the methodology and practice of HTA using real-world data. We thank ISPOR—The Professional Society for Health Economics and Outcomes Research and especially Elizabeth Molsen-David for their support throughout the process. Read the full report here: https://lnkd.in/eaKdyDfG We invite the scientific community and healthcare stakeholders to engage with these insights and consider their implications for improving healthcare decision-making. #HealthEconomics #HTA #Healthcare #ValueInHealth #EHR #RealWorldData #Research #ISPOR
