Assessing the Potential Role of Artificial Intelligence in Medication-Related Osteonecrosis of the Jaw Information Sharing

Abstract

BackgroundArtificial Intelligence, by answering questions about disease prevention strategies, can contribute to making diseases more treatable in their early stages. PurposeThis study aims to evaluate the quality of patient information by assessing the responses of the Chat Generative Pre-trained Transformer (ChatGPT, Open Al, USA) artificial intelligence model to questions related to medication-related osteonecrosis of the jaw (MRONJ). Study design, setting, sampleThe study was prospective cross-sectional design. The study was conducted within the Department of Oral and Maxillofacial Surgery. The study's questions were prepared by an experienced oral and maxillofacial surgeon and directed to the artificial intelligence platform. The responses were evaluated by oral and maxillofacial surgeons using the Global Quality Scale (GQS). Predictor variableThe predictor variable is question type. A total of 120 questions were categorized into six groups, which encompassed general information about MRONJ (Group 1), queries from patients about to initiate medication therapy (Group 2), questions from patients currently undergoing medication treatment (Group 3), inquiries from patients who had completed medication usage (Group 4), general treatment-related information (Group 5), and case scenarios (Group 6). Main Outcome VariablesThe main variable is the GQS score. The GQS rates the quality of information and its utility for the patients. The scores are as follows: Score 1: Poor quality, Score 2: Generally poor quality, Score 3: Moderate quality, Score 4: Good quality, Score 5: Excellent quality. CovariatesNot applicable. AnalysesKruskal-Wallis and Mann-Whitney-U tests were applied for intragroup and intergroup analyses. The statistical significance level was determined as p<0.05 and p<0.01. ResultsThe average score for all questions was calculated to be 3.9±0.8, which is above the "moderate quality" threshold. Group 1 had a mean score of 3.4±1.1; Group 2 had 4.1±0.7; Group 3 had 3.8±0.8; Group 4 had 4.3±0.6; Group 5 had 4.2±0.7; and Group 6 had 4.1±0.5. The variations in mean scores among these groups did not exhibit statistical significance (p>0.05). Conclusions and RelevanceThe artificial intelligence model has generated responses of moderate quality to questions about MRONJ. The use of the artificial intelligence platform may assist in patients gaining a fundamental understanding of MRONJ.