A Retrospective Analysis of Real-World Data: Assessing and Predicting Cardiovascular Risk Associated with Calcium Channel Blockers in Patients with Hypertension
Author(s)
Linke Zou, Master1, Xingwei Wu, PhD2, Ming Hu, PhD1;
1Sichuan University, Chengdu, China, 2Sichuan Provincial People's Hospital, Chengdu, China
1Sichuan University, Chengdu, China, 2Sichuan Provincial People's Hospital, Chengdu, China
OBJECTIVES: To evaluate and predict the preventable cardiovascular risk of calcium channel blockers(CCB) in the treatment of hypertension by analyzing multi-center healthcare data in China.
METHODS: Data on inpatients patients with ≥2 hospitalizations treated with CCB from 2019 to 2024 in three medical institutions were analyzed. Utilizing patient IDs and time series information, match the data fields and employ logistic regression analysis to evaluate the effectiveness of various CCB drugs. Data preprocessing involves impute missing values, data oversampling, and feature selection, etc. Nine classification algorithms were used to construct prediction models. The endpoints included major adverse cardiovascular events (MACE), which was assessed at 1 year, 2 years, 3 years. The performance was evaluated using metrics(such as AUC) to select the optimal prediction model.
RESULTS: A total of 19374 inpatient data were collected(sample siz: 3854: covariate size: 1843). The proportion of males(55.49%) and patients(59.82%) aged ≤ 65 is slightly higher. Coronary heart disease(53%), stroke(31%), and heart failure(11%) were the main outcome events. Logistic regression revealed that observation time, age, and the use of atorvastatin, furosemide, and dexamethasone injections significantly affected cardiovascular risk (P<0.05). The top three most effective CCB among 1165 subgroups are nifedipine sustained-release, controlled-release, and standard tablets, with effectiveness rates of 71.79%, 70.64%, and 65.22%. The CatBoost model excelled in 1 and 2-year CCB cardiovascular risk predictions (AUCs: 0.7978, 0.8289), while the Random Forest model performed the best for 3-year. (AUC=0.8329).
CONCLUSIONS: The data analysis results indicate that the future risk of CVD in hypertensive patients should be taken seriously. The subgroup analysis provide clinical practice references for individualized drug treatment of hypertension patients in China,while machine learning provides assistance in predicting the risk of cardiovascular disease. Risk prediction can mitigate the economic burden associated with cardiovascular complications, which may enhance cost-effectiveness in clinical practice.
METHODS: Data on inpatients patients with ≥2 hospitalizations treated with CCB from 2019 to 2024 in three medical institutions were analyzed. Utilizing patient IDs and time series information, match the data fields and employ logistic regression analysis to evaluate the effectiveness of various CCB drugs. Data preprocessing involves impute missing values, data oversampling, and feature selection, etc. Nine classification algorithms were used to construct prediction models. The endpoints included major adverse cardiovascular events (MACE), which was assessed at 1 year, 2 years, 3 years. The performance was evaluated using metrics(such as AUC) to select the optimal prediction model.
RESULTS: A total of 19374 inpatient data were collected(sample siz: 3854: covariate size: 1843). The proportion of males(55.49%) and patients(59.82%) aged ≤ 65 is slightly higher. Coronary heart disease(53%), stroke(31%), and heart failure(11%) were the main outcome events. Logistic regression revealed that observation time, age, and the use of atorvastatin, furosemide, and dexamethasone injections significantly affected cardiovascular risk (P<0.05). The top three most effective CCB among 1165 subgroups are nifedipine sustained-release, controlled-release, and standard tablets, with effectiveness rates of 71.79%, 70.64%, and 65.22%. The CatBoost model excelled in 1 and 2-year CCB cardiovascular risk predictions (AUCs: 0.7978, 0.8289), while the Random Forest model performed the best for 3-year. (AUC=0.8329).
CONCLUSIONS: The data analysis results indicate that the future risk of CVD in hypertensive patients should be taken seriously. The subgroup analysis provide clinical practice references for individualized drug treatment of hypertension patients in China,while machine learning provides assistance in predicting the risk of cardiovascular disease. Risk prediction can mitigate the economic burden associated with cardiovascular complications, which may enhance cost-effectiveness in clinical practice.
Conference/Value in Health Info
2025-05, ISPOR 2025, Montréal, Quebec, CA
Value in Health, Volume 28, Issue S1
Code
RWD91
Topic
Real World Data & Information Systems
Topic Subcategory
Health & Insurance Records Systems
Disease
SDC: Cardiovascular Disorders (including MI, Stroke, Circulatory)