Design and Evaluation of a Machine Learning-Guided Intervention Strategy for Managing Dental Anxiety and Pain in Children

Abstract

The present study aimed to develop and evaluate a machine learning-guided strategy for allocating non-pharmacological interventions and to compare its clinical outcomes with a fixed standard intervention for managing dental anxiety and pain in children. The study was conducted in two phases. In the first phase, data from 300 children aged 6–12 years, including demographic characteristics, dental history, baseline anxiety, expected pain, and responses to non-pharmacological interventions, were collected. Four algorithms, including logistic regression, support vector machine, random forest, and XGBoost, were trained and evaluated. In the second phase, 80 independent children were randomly assigned to an algorithm-guided intervention group or a control group. In the experimental group, the intervention type was selected according to the output of the machine learning model, whereas the control group received fixed visual-auditory distraction. Dental anxiety was assessed using the CFSS-DS, experienced pain using the Wong-Baker FACES Pain Rating Scale, and heart rate at different stages of treatment. The first-phase results showed that XGBoost had the best predictive performance, with an accuracy of 0.91 and an ROC-AUC of 0.95. In the second phase, the algorithm-guided group showed greater reductions in dental anxiety and experienced pain than the fixed-intervention control group. Children in the algorithm-guided group also demonstrated lower heart-rate responses, better behavioral cooperation, and greater parental satisfaction. These findings support the comparative clinical performance of algorithm-guided intervention allocation versus a fixed visual-auditory distraction strategy. Because the control condition used a single fixed intervention and individual counterfactual treatment effects were not estimated, the results should not be interpreted as isolating the causal effect of personalization itself.