Abstract:

Objective To analyze the co-occurrence influencing factors of severe visual impairment and low back pain and establish a predictive model，providing ideas for the prevention and control measures of severe visual impairment and low back pain. Methods A stratified cluster sampling method was employed to select 1 701 primary and secondary school students in the urban area of Qingdao from September to November 2023. A co-morbidity screening for severe visual impairment and low back pain and a questionnaire survey on influencing factors were carried out. Based on the machine learning algorithm Boruta combined with XGBoost，the co-morbidity influencing factors were analyzed and a predictive model was developed，which was visualized by an interactive Nomogram. Statistical analysis was performed using R statistical software version 4.3.3. Results The co-morbidity rate of severe visual impairment and low back pain among primary and secondary school students in the urban area of Qingdao was 15.34%（261 cases）. The main influencing factors selected were grade，types of fruits consumed，days with more than 60 minutes of moderate-to-vigorous-intensity physical activity，moderate-to-vigorous-intensity physical activity for more than 60 minutes on weekends，places for activities during breaks，one fist distance during reading and writing，one foot distance during reading and writing，one inch distance during reading and writing，reading or using electronic screens while lying down，distance of more than 3 meters between eyes and TV screen，daily sleep time，and daily indoor sedentary time，etc.（12 influencing factors，all P<0.05）. The predictive model demonstrated an AUC of 0.782（95%CI：0.750-0.813） in the training set and 0.770（95%CI：0.710-0.829） in the test set，with no statistically significant difference between them（DeLong test P>0.05）. The threshold probabilities for achieving clinical utility were identified as 0.54 in the training set and 0.61 in the test set. Conclusion A machine learning-based predictive model for the co-morbidity of severe visual impairment and low back pain among primary and secondary school students has been developed. The prediction results are presented through visualization tools，which can provide a basis for formulating prevention and control policies for severe visual impairment and low back pain.

Key words: Severe visual impairment, Lower back pain, Machine Learning, Comorbidity risk, Influencing factors, Predictive model