Archives of Pathology & Laboratory Medicine
Context: Drug history is important clinical information for the correct histologic diagnosis of a wide variety of nonneoplastic diseases, such as hepatitis, gastritis, and colitis. As pathologists, however, keeping a strong knowledge of all drugs (and variations of drug names) can be challenging. Using online databases to find relevant drug information one at a time is very time-consuming.
Design: Sixty-seven drugs were collected from the medical records of 5 randomly selected patients. ChatGPT (https://chat.openai.com/chat) was instructed to identify (1) the main clinical indications, and (2) the major side effects of each drug, using the prompt “List the major indications and side effects, excluding vitamins and over-the-counter drugs, of the following drugs: cyclobenzaprine, …’.” We then had a list of 36 drugs with their main indications and side effects. We compared the results from ChatGPT with those listed in the UpToDate drug information section. The MedlinePlus website was also used as a confirmatory reference.
Results: ChatGPT immediately responded by listing information (indications and side effects) for 36 drugs. On the other hand, we could only get the same information via UpToDate one drug at a time, and it took almost 37 minutes to obtain the same information. Most of the indications and side effects were similar to those found on UpToDate (Table).
Conclusions: Utilizing ChatGPT by pathologists is not only a fast and efficient way to obtain relevant information on multiple drugs simultaneously, but it also has comparable efficacy to well-known specialized drug websites.
Elzieny M, Qu Z. Using an artificial intelligence (AI) tool (ChatGPT) to identify the main indications and side effects of drugs in patient medical records. Arch Pathol Lab Med. 2023 Sep;147(9):e140. doi:10.5858/arpa.2023-0258-AB