Essential hypertension is a non-communicable global epidemic disease. When complicated by major adverse cardiovascular and cerebrovascular events (MACCE), essential hypertension in associated with high incidence and high disability. Traditional Chinese medicine (TCM) has significant advantages in the prevention and treatment of this disease. Constructing a risk prediction model that combines disease and syndrome and has the characteristics of TCM syndrome elements can provide a reliable strategy for the prevention and treatment of hypertension combined with MACCE.

To explore the risk factors for the occurrence of MACCE in patients with essential hypertension, and to construct a risk prediction model based on the fusion of multi-modal data,using a combined classifier as the basic algorithm and involving TCM syndrome elements.

A total of 985 outpatients and inpatients with essential hypertension who were treated in Henan Integrative Medicine Hospital (Henan Academy of Chinese Medicine) from 2020 to 2023 were selected. Clinical data such as general information (including demographics), laboratory indicators, and TCM syndromes were collected,and the occurrence of MACCE was followed up. Taking whether MACCE occurred or not as the outcome indicator, the patients were divided into the MACCE group (284 cases) and the non-MACCE group (701 cases). The Cox proportional hazards regression model was used to screen independent risk factors and determine the final modeling variables. The random forest algorithm and the AdaBoost algorithm were respectively used to establish models, and the effectiveness of the two models was compared through the confusion matrix.

Through the collinearity diagnosis of the Cox proportional hazards regression model, combined with clinical observations, a total of 17 variables were finally established, including disease course, gender, family history of premature cardiovascular disease, BMI, homocysteine (Hcy), mean ankle-brachial index (ABI), mean arterial pressure, flow-mediated dilation (FMD), headache, Qi deficiency with blood stasis syndrome, Yin deficiency with Yang hyperactivity syndrome, liver-kidney Yin deficiency syndrome, Yin-Yang deficiency syndrome, impaired glucose tolerance, dyslipidemia, carotid-femoral pulse wave velocity (cfPWV), dizziness. A random forest model and an AdaBoost model were constructed. The effectiveness evaluated by the confusion matrix showed that the overall accuracy of the AdaBoost model in predicting the risk of MACCE in patients with essential hypertension was 85.57%, the precision was 91.02%, the recall was 92.09%, and the F1 value was 91.55%, which was much higher than those of the random forest model (the overall accuracy was 71.81%, the precision was 75.47%, the recall was 91.32%, and the F1 value was 82.64%).

Based on the condition of multi-modal data fusion, the risk prediction model of the combined classifier constructed with variables containing the characteristics of TCM syndrome elements has better effectiveness. It can preliminarily determine the occurrence of MACCE and play a scientific guiding role in clinical prevention and treatment decisions.