Hypertension prevalence in Bangladesh was 18.04% and, among machine learning and deep learning models, weighted logistic regression achieved the highest accuracy and specificity while random forest achieved the highest recall and F1-score, with age, BMI, sex, family size, and educational level identified as the most important predictors.
The overall prevalence of hypertension among adults in Bangladesh was 18.04%, with higher prevalence among women than men.
Hypertension was significantly associated with age, BMI, diabetes, wealth index, education, household size, and region.
Weighted logistic regression (WLR) achieved the highest accuracy, precision, specificity, AUC-ROC, and AUC-PR among all models tested.
The random forest (RF) model achieved the highest recall and F1-score on the test data, indicating greater sensitivity in identifying individuals with hypertension.
Age, BMI, sex, family size, and educational level were identified as the most important predictors of hypertension among the variables included in the study.
Model performance was evaluated using multiple metrics including accuracy, precision, recall, specificity, F1 score, AUC-ROC, and AUC-PR to account for the imbalanced nature of the dataset.
This research suggests that nearly 1 in 5 adults in Bangladesh (about 18%) has hypertension, with women slightly more affected than men. Using data from over 14,000 adults surveyed in 2022, the researchers found that factors such as older age, higher BMI, having diabetes, lower education, household size, wealth, and geographic region were all meaningfully linked to hypertension risk. Among these, age, BMI, sex, family size, and education level stood out as the strongest predictors when fed into computer-based prediction models. The study tested six different artificial intelligence models — four machine learning and two deep learning approaches — to see which could best identify people with hypertension. No single model excelled at everything. The weighted logistic regression model was the most accurate overall and rarely gave false alarms, but it missed most actual hypertension cases (catching only about 7%). The random forest model, by contrast, correctly identified about 69% of people who actually had hypertension, making it potentially more useful for public health screening where missing cases is costly. The authors caution that before any model is used in practice, it would need to be tested on new, independent datasets to confirm its reliability. This research matters because hypertension is a major driver of heart disease and stroke, and Bangladesh, like many low- and middle-income countries, faces growing burdens from these conditions. The findings highlight that socioeconomic and demographic factors — not just biology — play a central role in hypertension risk, pointing to the importance of targeted public health programs. The comparison of prediction models also offers practical guidance for researchers and health planners on choosing the right tool depending on whether minimizing missed cases or minimizing false positives is the priority.
Chandra S, Molla M, Islam S, Rahaman M, Ali M, Ali M. (2026). Machine learning and deep learning-based prediction of hypertension and analysis of its major risk factors in Bangladesh.. PloS one. https://doi.org/10.1371/journal.pone.0358471