A Novel Artificial Intelligence–assisted Risk Assessment Model for Preventing Complications in Esthetic Surgery

Prevention of complications to reduce morbidity and mortality, and improve patient satisfaction is of paramount importance to plastic surgeons. This study aimed to evaluate the predictive risk factors for complications and to validate a novel risk assessment model, using artificial intelligence. METHODS: A retrospective review of esthetic surgery procedures performed by the author between 2015 and 2020 was conducted. The Pearson correlation test was used to analyze the risk factors and complications. Differences in the mean risk scores among the three risk groups were tested using one-way analysis of variance. Risk scoring was validated using a machine learning process with a support vector machine in a Google Colaboratory environment. RESULTS: Of the 372 patients, 28 (7.5%) experienced complications. The Pearson correlation coefficients between the risk score and body mass index (BMI: 0.99), age (0.97), and Caprini score of 5 or more (0.98) were statistically significant (P < 0.01). The correlations between the risk scores and sex (−0.16, P = 0.58), smoking habit (−0.16, P = 0.58), or combined procedures (−0.16, P = 0.58) were not significant. Necrosis was significantly correlated with dehiscence (0.92, P = 0.003) and seroma (0.77, P = 0.041). The accuracy of the predictive model was 100% for the training sample and 97.3% for the test sample. CONCLUSIONS: Body mass index, age, and the Caprini score were risk factors for complications following esthetic surgery. The proposed risk assessment system is a valid tool for improving eligibility and preventing complications.