ファインデックス’s consolidated subsidiary has published the results of a clinical trial of “CocktailAI,” an AI product for generating medical documents. In a study using 6 simulated ophthalmology patient records, the group in which physicians edited AI-generated drafts outperformed manual preparation, while reducing preparation time. The paper was published online in the international journal PLOS Digital Health.
ファインデックス announced that a paper summarizing the results of a clinical trial jointly conducted by its consolidated subsidiary フィッティングクラウド and Kyoto University Hospital was published online in the international academic journal “PLOS Digital Health” on 27 October 2026. The study evaluated “CocktailAI,” a text-generation solution for medical documents that uses generative AI to produce drafts of discharge summaries and referral letters from electronic medical record entries. The solution is designed to address the risk of hallucinations and factual errors by displaying the source medical-record entries alongside the generated text.
The clinical trial was a pilot randomized controlled trial using 6 simulated ophthalmology patient records. It involved 21 ophthalmologists and compared the time required to prepare, and the quality of, discharge summaries and referral letters across 3 groups: physicians editing AI-generated drafts, physicians preparing the documents manually, and a group without human review. The group that edited AI-generated drafts outperformed manual preparation on many quality measures and reduced discharge-summary preparation time by an average of approximately 59.4 seconds. Although the AI-only output was prepared in the shortest time, its quality did not meet the required standard.
The company said it plans to expand its lineup, including integration with the in-hospital document creation system “DocuMaker” and the standalone version “CocktailBar.”