A comparative study of antihypertensive drugs prediction models for the elderly based on machine learning algorithms

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Global hypertension management strategies show limited effectiveness, with low blood pressure control rates in treated patients.
• Inadequate control of high blood pressure poses significant health risks worldwide.

Purpose of the Study:

• To develop a predictive model for antihypertensive medication selection using patient attributes.
• To assist physicians in making faster and more rational medication choices for hypertension treatment.

Main Methods:

• Collected clinical data from elderly hypertensive patients.
• Employed statistical methods and machine learning (ML) algorithms for feature selection.
• Constructed and evaluated five ML models (RF, SVM, LightGBM, ANN, NB) using 5-fold cross-validation and micro-F1 scores.

Main Results:

• Identified key predictive features including Age, SBP, DBP, and various blood/biochemical markers (Lymph, RBC, HCT, MCHC, PLT, AST, TBIL, Cr, UA, Urea, K, Na, Ga, TP, GLU, TC, TG, γ-GT), along with Gender, HTN CAD, and RI.
• The LightGBM model demonstrated superior performance, achieving a micro-F1 score of 78.45%, outperforming the other four models.

Conclusions:

• The LightGBM model effectively predicts antihypertensive medication regimens.
• This predictive model holds potential for enhancing personalized hypertension treatment strategies.