Hypertension control in resource-constrained settings: Bridging socioeconomic gaps with predictive insights

Background:

Hypertension continues to be a pivotal driver of global cardiovascular disease burden and adverse health outcomes, particularly in resource-constrained settings where disparities in socioeconomic status and clinical infrastructure hinder effective management. Despite medical advancements, achieving optimal blood pressure (BP) control remains a formidable challenge, necessitating a nuanced understanding of multifactorial risk determinants.

Methods:

A cross-sectional analysis was conducted on 1,000 hypertensive patients from a larger dataset comprising 100,000 population size. Three hundred patients were examined for personalised BP control predictors who met the inclusion criteria of being treated for at least one year at the Hypertension and Research Centre in Rangpur, Bangladesh, between January 2020 and January 2021. BP control was assessed using World Health Organisation (WHO) and National Institute for Clinical Excellence (NICE) guidelines, and a comprehensive analysis of the sociodemographic and clinical variables was performed using multivariate logistic regression. Machine learning models such as K-Nearest Neighbours (KNN) were utilised to predict BP control with good performance using cross-validation techniques compared to other models. Explainable AI tools like Shapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME) provide interpretations of key variables with predictive qualities.

Results:

The mean age of participants was 49.37 ± 12.81 years, with 54.7% aged 40–59 years and 57.7% male. The overall BP control rate among the study population was 28%. Among those with controlled hypertension, 42% were rural residents ($p=0.005$) and 37% were homemakers ($p$ $<$ 0.001), indicating better control in these subgroups. Key facilitators of BP control included higher education levels (e.g., post-graduate OR = 1.17, $p<0.001$), lower cholesterol levels (SHAP value = 0.097), and adherence to combination therapy (75% of controlled cases). Conversely, diabetes mellitus (SHAP value = 0.069) and ischemic heart disease (OR = 0.95, $p=0.004$) emerged as significant impediments to BP control. Advanced machine learning models, including KNN, achieved an unparallelled predictive accuracy of 99%, underscoring precision-based interventions’ transformative potential. SHAP analysis revealed dietary habits (SHAP value = 0.077) and physical activity (SHAP value = 0.079) as modifiable predictors, highlighting the efficacy of personalised lifestyle strategies. Simulation-based interventions grounded in machine learning insights reduced high-risk classifications by 15%, further reinforcing predictive analytics’ value in hypertension management. Sensitivity analysis highlighted the dominance of socioeconomic factors, with income level (sensitivity: 0.85) and healthcare accessibility (sensitivity: 0.78) emerging as critical predictors, reinforcing the importance of addressing health inequities in hypertension management.

Conclusion:

The study elucidates critical gaps in hypertension management, emphasising the urgent need to address modifiable risk factors, tailor therapeutic regimens, and integrate socioeconomic considerations into public health frameworks. The findings advocate for scalable, data-driven interventions to bridge the hypertension care gap, thereby mitigating cardiovascular disease risks and enhancing health equity in underserved regions.