Why are we still looking through guidelines to get a clinical answer? Naturally, we need a chatbot that was trained on these guidelines so we can simply ask⚡ [Thread] 👇
A quick & dirty search on PubMed revealed no relevant studies among 61 results. A 2022 scoping review did not report any chatbots like that either: doi.org/10.2196/35882 BTW, only 88% of the chatbots studied were reported technically well 😱 2/8
In one study, a chatbot providing management guidelines for patients with beta thalassemia was developed, but it was rule-based, which clearly isn’t the state of the art today: doi.org/10.1155/2022/9… 3/8 
Some other chatbot was created for hidradenitis suppurativa, but it didn’t look quite like the subject judging from the report: doi.org/10.1159/000511… 4/8
Another chatbot was developed for giving wound dressing guidance, but it seemed to be trained on its own knowledge base, not published guidelines; and it wasn’t so good with NLP either: doi.org/10.1016/j.ijme… 5/8
One other study provided a notable use case of incorporating an ontology in the development of a chatbot but had another purpose: doi.org/10.2196/29289 6/8
Another ScR on chatbots in health care is in progress, but its protocol excluded chatbots targeted at HCPs: doi.org/10.2196/40265 7/8
Finally, and probably most importantly, when I google CPG + chatbot, there isn’t anything there…but it could be! I’ve already procured a list of the most searched for guidelines from Google Keyword Planner, so we have training data… 😏 What do you say? 8/8
*88% of the chatbots studied were NOT reported technically well, indeed! A typo 😔 doi.org/10.2196/35882
• • •
Missing some Tweet in this thread? You can try to
force a refresh
