Reporting guidelines for chatbot health advice studies: explanation and elaboration for the Chatbot Assessment Reporting Tool (CHART)
Huo B., Thirunavukarasu AJ., Collins GS., Chartash D., Flanagin A., Iorio A., Cacciamani G., Chen X., Liu N., Mathur P., Chan AW., Laine C., Pacella D., Berkwits M., Antoniou SA., Camaradou JC., Canfield C., Mittelman M., Feeney T., Loder E., Agha R., Saha A., Mayol J., Sunjaya A., Harvey H., Ng JY., McKechnie T., Lee Y., Verma N., Stiglic G., McCradden M., Ramji K., Boudreau V., Ortenzi M., Meerpohl J., Vandvik PO., Agoritsas T., Samuel D., Frankish H., Anderson M., Yao X., Loeb S., Lokker C., Liu X., Guallar E., Guyatt G.
The Chatbot Assessment Reporting Tool (CHART) reporting guideline promotes transparent and comprehensive reporting of studies evaluating the performance of generative artificial intelligence (AI)driven chatbots for the purposes of summarising clinical evidence and providing health advice, referred to here as chatbot health advice (CHA) studies. CHART is the product of an international, multi-phase, consensus based initiative involving various stakeholders and comprises a 12-item checklist with 39 subitems. The checklist includes items on open science, title and abstract, introduction, model identification, model details, prompt engineering, query strategy, performance definition and evaluation, statistical analysis, results, discussion, with an accompanying flow diagram. Each item includes distinct subitems. This explanation and elaboration article discusses each subitem and provides a detailed rationale for its inclusion in the CHART checklist.