Data-driven Subclassification of Treated Hypertensive Adults: Implications for Personalized Management

Introduction. Hypertension control remains a public health challenge worldwide, with variability in management outcomes. This study aimed to identify blood pressure control phenotypes among Mexican adults with previously diagnosed hypertension. Methods. We analyzed individuals (n = 1,308) aged 20 years and older with physician-confirmed hypertension. Controlled hypertension was defined as systolic/diastolic blood pressure < 130/80 mmHg, assessed via triplicate measurements. We used Principal Component Analysis (PCA) to reduce dimensionality and Gaussian Mixture Model (GMM) clustering on PCA-transformed data identified phenotypes. Results. GMM clustering identified 8 phenotypes, with cluster sizes ranging from 22 (1.7%) to 325 (24.8%) individuals. Cluster 3, integrated mainly by older women (mean age 66.5 years), long-term hypertension (> 10 years), and high socioeconomic status, was the largest cluster and showed better control. In contrast, Cluster 7 (n = 142) was mainly constituted by patients with low socioeconomic status and uncontrolled blood pressure. PCA variable contributions highlighted adherence to physical activity (9.07%), dietary modifications (7.24%), sex (7.14%), and type 2 diabetes (6.64%) as dominant factors in principal component 1 (PC1), whereas age explained 89.2% of PC2 variance. Conclusions. The presented results suggest heterogeneous hypertension control patterns influenced by demographic, clinical, and behavioral factors. Targeted interventions for high-risk phenotypes, particularly younger patients and those with poor adherence, could enhance blood pressure management strategies in Mexico. The integration of PCA and GMM offers a robust framework for phenotyping complex health conditions in population-based studies.