ChatGPT provides safe responses to post-operative concerns following total joint arthroplasty

Abstract

Introduction: Postoperative triage is a considerable financial and time burden following total joint arthroplasty (TJA). An ideal post-operative triage method would be equally as safe as triage nurses, but more cost-effective and time-efficient. Therefore, this study aimed to see if ChatGPT could serve as a safe and effective method of postoperative triage in responding to patient concerns following TJA. Methods: A list of ten potential concerns from post-operative arthroplasty patients was compiled ranging from “How much walking can I do?” to “My leg is throbbing and red, and I’m having trouble breathing. What should I do?”. These concerns were presented to trained orthopaedic triage nurses (n=3) and ChatGPT (Version 3.5). The generated responses were rated blindly by three fellowship-trained orthopaedic surgeons specializing in adult reconstruction using Likert scales ranging from 1-5 for safety and completeness (1 being highly unsafe/incomplete; 5 being highly safe/complete). The average scores were calculated for each prompt and are presented as mean±standard deviation. Results: There was no significant difference in average safety scores between ChatGPT responses (4.2±0.5) and those of triage nurses (4.3±0.4) (P=0.68). Similarly, there was no significant difference in completeness scores between ChatGPT responses (3.9±0.8) and those of triage nurses (4.0±0.6) (P=0.67). Furthermore, there were no significant differences in safety or completeness between ChatGPT and triage nurses for any of the individual prompts (all P values >0.05). Conclusions: Postoperative triage is a critical component of perioperative care but represents a substantial time and cost burden. This preliminary study highlights the promising opportunity for future use of AI-enabled chatbots to provide instantaneous, safe, and complete responses to postoperative patient concerns.