Exploring Key Factors Of T2DM In Urban And Rural Populations: A Machine Learning Approach

Type 2 Diabetes Mellitus (T2DM) poses a significant challenge in the global healthy aging process. It is crucial to explore the key factors of T2DM in potential patients from urban and rural areas for differentiated and precise health interventions. PURPOSE: One of the primary objectives of precision health is to identify key factors influencing disease susceptibility within different populations. Our study aims to analyze the key factors and disparities in the incidence of T2DM in urban and rural populations in China using machine learning methods. METHODS: This study, conducted in Fujian Province, China, utilized a multi-stage cluster random sampling method and successfully collected 30,373 adult samples. Participants underwent laboratory testing to determine their T2DM status. Subsequently, data on 57 aspects were gathered through questionnaires and physical fitness tests. Finally, the key factors influencing T2DM, and the urban-rural differences were analyzed using the random forest (RF) model and a multinomial logistic regression (MLR). Subsequently, a T2DM prediction model was constructed respectively. RESULTS: The detection rate of T2DM in older adults (Age > 65) was 27.3 %, with 18.4 % and 35.2 % in urban and rural areas, respectively. According to the MLR analysis: Education, Age, Annual income, Waistline, Hypertension, and BMI were common factors. Physical activity (OR = 0.702 for Moderate PA), Marital status (OR = 0.651, Married), and Sleep quality (OR = 0.622, Good) are protective factors specific to urban adults. Alcohol consumption (OR = 3.549 for drinking more than once a week) and smoking (OR = 2.653 for smoking more than once a week) were risk factors specific to T2DM in rural adults (all P < 0.05). And the T2DM prediction model constructed through the RF algorithm exhibits excellent performance (Accuracy: 0.9407, Precision: 0.9920, AUC:0.9360 for urban adults; Accuracy: 0.9626, Precision: 0.9745, AUC: 0.9612 for rural adults). CONCLUSIONS: This study indicates differences in the key factors contributing to the incidence of T2DM between urban and rural residents. Distinct prediction models for T2DM were separately constructed for urban and rural populations, facilitating targeted prevention and precise intervention for T2DM. This study was sponsored by the Fujian provincial natural science foundation, grant number 2020 J01087, and Fujian provincial health technology project, grant number 2020GGA023