A model for predicting the aesthetics of postoperative scar in pediatric surgery: an original study

BACKGROUND: A postoperative scar is a visible and inevitable result of surgical intervention. A scar on the child’s body can worsen the quality of his/her life as it can be manifested with physical ailment (pain, itching, skin flaking), disharmony in psychological and age adaptation as well as with the feeling of dissatisfaction with one’s own body. The basic question which we had been put before our research was: “Is it possible to predict the aesthetic result of postoperative scars in children before surgery, taking into account previously known predictors?” AIM: To create a prognostic model for predicting aesthetic outcomes in postoperative scars in pediatric surgery using machine learning modalities. METHODS: 219 children with postoperative scars on their body were enrolled in the study in 2022–2024. The decision tree modality in the SPSS Statistic 23 program was chosen as a forecasting algorithm. Models were built using CHAID, Exhaustive CHAID and CRT. The desired prognosis criterion was based on the final score in the SPASS scale questionnaire sheet. Predictors were: child’s age, scar location, scar length, and color gradient. The color gradient was determined in Photoshop CS6 digital graphics editor based on photographs in RGB color coordinate system. RESULTS: A workable predictive model using CHAID modality has been built. The model included 32 nodes, 25 of which were terminal nodes. AUC parameter in the ROC curve was 0.924. Asymptotic confidential interval (95%) was in the range of 0.890–0.922. The value of performed “cross-validation” was acceptable. Predictive modeling was done using nine terminal nodes with high values of “index” and “response” parameters. The length of the postoperative scar is a priority factor in determining its aesthetic perception. CONCLUSION: In this trial, a predictive model of postoperative scar aesthetics in pediatric surgery based on the decision tree modality has been developed. This is the first attempt to predict outcomes of this kind in pediatric surgery. The obtained results will be useful for practical application in pediatric surgery when planning surgical interventions.