Abstract
Objective
To explore the influencing factors of coronary heart disease in Tianjin and construct a suitable early screening model for coronary heart disease in the Tianjin community population that integrates traditional Chinese and Western medicine.
Methods
Utilizing annual physical examination data from the electronic health records of community residents participating in the national basic public health service project as the data source and modeling index. The propensity score matching method (PSM) was applied to match age and gender, establishing a standard database. Data were randomly divided into training (8:2 ratio) and test sets using R 4.1.3 software. The training set was used to build the model, and the test set was used to verify the model's discrimination and calibration. Univariate and multivariate Logistic analysis methods were employed to screen predictors and build the model. The model was evaluated through internal and external validation, including goodness of fit, ROC curve analysis, sensitivity, specificity, accuracy, likelihood ratios, and AUC. The "caret" package in R software was used for internal validation via the Bootstrap method, and clinical decision curve analysis (DCA) was applied to assess clinical efficacy.
Results
A total of 24,792 community individuals with complete archives were screened, including 3,487 cases of coronary heart disease. After propensity matching, 3,487 cases each for the coronary heart disease and non-coronary heart disease groups were matched. The training set consisted of 5,580 cases, and the test set of 1,394 cases, with no significant baseline data differences (P > 0.05). A joint early screening model of traditional Chinese and Western medicine was established, incorporating variables such as residential area, waist circumference, diastolic blood pressure, auscultation murmur, dietary salt preference, coexisting conditions like stroke and anemia, and TCM constitution types. The Hosmer-Lemeshow test indicated a good model fit (χ²=14.519, P = 0.065 > 0.05), with an AUC of 0.796. The model demonstrated predictive efficacy and clinical value, suggesting its potential for quantitatively evaluating the risk of coronary heart disease in community residents without reliance on biochemical indicators.
Conclusions
The early screening model for coronary heart disease, based on annual physical examination data of the community population, shows good clinical predictive efficiency. It can quantitatively evaluate the risk of coronary heart disease in community residents, assist community doctors in screening high-risk populations, and is valuable for early diagnosis.