Development and validation of a machine learning model for early detection of adolescent idiopathic scoliosis using census data

To explore and validate the machine learning (ML) risk prediction models based on the survey data of adolescent scoliosis in Laoshan District, Qingdao City. Adolescents who underwent scoliosis screening in 10 primary and secondary schools in Laoshan District, Qingdao City from April to July 2023 were retrospectively selected as research subjects. The included data were randomly sampled at a ratio of 7:3, with 70 % entering the modeling group and 30 % entering the validation group. Modeling group patients were further divided into scoliosis group and non-scoliosis group. Predictive models including logistic regression, random forest (RF), and support vector machine (SVM) were constructed using R software. The area under the receiver operating characteristic curve (AUC) was used to evaluate the discriminative accuracy of the model. Among 510 adolescents (357 adolescents in the modeling group and 153 adolescents in the validation group), there were 32 (8.96 %) cases of scoliosis in the modeling group and 325 cases without scoliosis. Logistic regression had the best evaluation indicators, with optimal model performance, AUC, accuracy, and recall rate were 0.988, 0.902, and 0.895, respectively. Further interpretation by column plots revealed that nutritional status, parental history of scoliosis, use of double-shoulder bags, backpack weight, number of physical education classes per week, weekly sitting time, daily exercise time, physical exercise projects, and reading and writing posture were important factors in predicting adolescent scoliosis by the model. The calibration curve indicated good consistency between the logistic regression risk prediction model and the ideal model. The logistic regression model based on ML effectively predicts adolescent scoliosis using key factors like nutrition, parental history, backpack use, posture, and physical activity, aiding quick and accurate risk assessment by medical staff.