To characterize expert assessments of ChatGPT-generated answers about fast-track total knee arthroplasty (TKA), describe ratings for conventional TKA questions, and explore differences between the groups.
Approach:
Question development: A modified Delphi panel selected 50 conventional TKA questions from an initial set of 91. Clinicians experienced in fast-track TKA separately developed 20 fast-track questions.
Response generation and evaluation: In June 2026, ChatGPT generated unedited Polish answers in separate conversations. Three high-volume arthroplasty surgeons independently rated accuracy, completeness, safety, and readability on 5-point scales, and recorded whether specialist correction was nee…
Key Findings:
The supplied context describes the study design and evaluation measures but does not report results or rating scores.
Interpretation:
The study assessed expert judgments of ChatGPT responses to two differently developed sets of TKA questions; the comparison was not designed to establish equivalence or noninferiority.
Limitations:
The conventional and fast-track question sets used different selection processes and may have differed in wording, scope, and difficulty.
Although subgroup labels were withheld from raters, fast-track topics could reveal group assignment.
The provided context does not include the statistical-analysis details or study results.
