Impact of large language model Dialogue System on Perioperative Anxiety and Depression in Patients Undergoing Total Knee Arthroplasty: A Randomized Controlled Trial (Preprint)

BACKGROUND Total knee arthroplasty (TKA) is a widely performed procedure for knee osteoarthritis, yet patient satisfaction remains suboptimal, with preoperative anxiety and depression contributing to poor postoperative outcomes. Large-scale language models (LLMs), such as ChatGPT, have shown potential in healthcare applications but remain underexplored in the context of TKA. This study evaluates the impact of an AI-driven dialogue system on perioperative anxiety and depression in TKA patients. OBJECTIVE This study aimed to evaluate the effect of LLMs dialogue system on improving anxiety and depression, as well as overall psychological well-being, in patients undergoing TKA. The primary objective was to assess the system’s impact on reducing perioperative anxiety and depression levels before and after surgery. Secondary objectives included evaluating patient satisfaction, the feasibility and scalability of the AI system compared to traditional methods, and its potential for broader clinical application in surgical settings. METHODS We expect to conduct a parallel randomized and controlled trial in Tianjin Hospital. This trial content needs to gather 210 participants between the ages of 50-80 and randomly divide them into two groups, namely, the LLM intervention group or the usual care group. ChatGPT 4.0 software was used to conduct a personalized psychological education activity for about 20-30 minutes for the LLM intervention group every day, and the routine nursing group was used as the control group, receiving standard preoperative education. The test results were evaluated by the hospital anxiety and Depression Scale (HADS) at baseline, 48 hours, 24 hours before surgery, and 24 hours, 48 hours, 72 hours after surgery to determine whether artificial intelligence is effective in anti anxiety and depression, and improve patients' surgical mood and health. RESULTS Statistical analyses showed that compared with the usual care group, the reduction of anxiety and depression in the LLM group was significantly greater. We found that both groups of patients experienced significant anxiety and depression before surgery. With the intervention and guidance of artificial intelligence, these symptoms in the LLM group continued to alleviate after surgery. The effect was particularly significant at 48 and 72 hours after surgery (P<.001), while the conventional nursing group showed a slower decline. CONCLUSIONS Experimental data shows that an intelligent psychological support system driven by artificial intelligence can effectively reduce anxiety and depression in TKA patients before and after surgery. Compared with traditional psychological intervention measures, this method breaks through the limitations of high labor costs, high economy, and difficulties in implementing personalized plans, achieving efficient and innovative breakthroughs. CLINICALTRIAL Chinese Clinical Trial Registry ChiCTR2500099927; https://www.chictr.org.cn/bin/project/edit?pid=256631ChatGPT；Total Knee Arthroplasty；Perioperative；Anxiety；Depression.