DOI: 10.1177/08927790261473184 ISSN: 0892-7790

Physician-Supervised ChatGPT-Assisted Versus Physician-Led Informed Consent for Reducing Preoperative Anxiety: A Randomized Controlled Trial

Yavuz Mert Aydın, Tevfik Erdem Özarslan, Merve Sena Baytar, Necmettin Aydın Mungan
Show PDF Cite

Purpose:

To evaluate whether ChatGPT-assisted informed consent (IC) is noninferior to conventional physician-led IC in reducing preoperative anxiety (PA) among patients undergoing a urologic procedure.

Materials and Methods:

In this prospective, open-label randomized controlled trial, 180 adult patients scheduled for retrograde intrarenal surgery/ureteroscopy or holmium laser enucleation of the prostate were randomized 1:1 to either artificial intelligence (AI)-assisted or physician-led IC. The primary outcome was postintervention anxiety changes, measured using the State-Trait Anxiety Inventory (STAI-1). Secondary outcomes included IC duration and postoperative quality of recovery (QoR), assessed using the QoR-15. Analysis of covariance was performed to adjust for baseline anxiety differences. Noninferiority was defined as an upper 95% confidence interval (CI) margin of +5 STAI points.

Results:

After applying the exclusion criteria, a total of 161 patients were included (AI-assisted: n = 83; control: n = 78). Baseline anxiety scores were higher in the AI-assisted group ( p < 0.01). STAI-1 scores decreased significantly in both groups following the IC process ( p < 0.01), with no significant between-group difference in ΔSTAI-1 ( p = 0.533). After adjustment for baseline STAI-1 and STAI-2 scores, postintervention anxiety did not differ between groups (mean difference, 0.26; 95% CI: −1.45 to 1.97), confirming noninferiority. The median IC duration was 10 minutes in both groups, although the between-group difference was significant ( p = 0.020). The QoR-15 score favored the control group ( p = 0.019).

Conclusions:

ChatGPT-assisted IC is noninferior to conventional physician-led IC in reducing PA without prolonging the IC process. These findings suggest a potential role for large language models as complementary tools within physician-led IC.

More from our Archive