A Machine Learning Approach to Predict Hypertension Using Cross-Sectional & Two Years Follow Up Data from A Health & Demographic Cohort

Q: What are the key findings of this study?

The XGBoost machine learning algorithm achieved the highest performance for predicting hypertension with 92.7% accuracy and an AUC of 0.971, outperforming logistic regression (84.2% accuracy, AUC 0.889).

Q: What is the clinical evidence from this study?

Study design: Cohort. Population: Hypertension (n=2000). Intervention: XGBoost machine learning model vs. Logistic Regression and other supervised machine learning models (Random Forest, Decision Tree, Support Vector Machine, Artificial Neural Network). Primary outcome: Predictive accuracy for hypertension (AUC 0.971).

Q: What does this research mean for the field?

The XGBoost machine learning algorithm predicts hypertension with high accuracy (92.7%) and an AUC of 0.971 using demographic, clinical, and follow-up cohort data, outperforming traditional logistic regression and other machine learning models. Novelty: ClaimNovelty.CONFIRMATORY. Consensus alignment: ConsensusAlignment.NEUTRAL.

doi:10.25258/ijddt.16.72s.150

October 17, 2026International Journal of Drug Delivery TechnologyOpen Access

A Machine Learning Approach to Predict Hypertension Using Cross-Sectional & Two Years Follow Up Data from A Health & Demographic Cohort

View Full Paper

Ask AI

Bookmark

Share

Key result

XGBoost predicts hypertension with ~93% accuracy, outperforming standard logistic regression.

AUC 0.971
n=2,000

Why the study?

Hypertension is a major modifiable cardiovascular risk factor, and machine learning techniques have potential to improve its prediction by analyzing complex interactions among demographic, clinical, and lifestyle variables. This study evaluated the performance of different machine learning algorithms for predicting hypertension using cross-sectional and two-year follow-up cohort data.

Do machine learning algorithms improve the prediction of hypertension using cross-sectional and two-year follow-up data in adults aged ≥18 years?

Key Points

To evaluate the performance of different machine learning algorithms for predicting hypertension using baseline cross-sectional and two-year follow-up cohort data.
Conducted a prospective observational cohort study of 2,000 adults aged ≥18 years using demographic, anthropometric, clinical, and lifestyle variables.
Trained and evaluated Logistic Regression, Decision Tree, Random Forest, Support Vector Machine, XGBoost, and Artificial Neural Network models on baseline and two-year follow-up data.
XGBoost achieved the highest performance with 92.7% accuracy, 91.8% sensitivity, 93.4% specificity, 91.1% precision, an F1-score of 0.914, and an AUC of 0.971.
Random Forest showed the second-highest performance with an accuracy of 91.8% and an AUC of 0.963, while Logistic Regression achieved 84.2% accuracy and an AUC of 0.889.
Feature importance analysis identified age (importance score 0.284), body mass index (0.221), systolic blood pressure (0.197), family history of hypertension (0.124), and diabetes mellitus (0.081) as the strongest predictors.

Study Design

Type

Cohort (n=2,000)

Multicenter

No

Structured PICO

Do machine learning algorithms improve the prediction of hypertension using cross-sectional and two-year follow-up data in adults aged ≥18 years?

P

Population

2,000 adults aged ≥18 years (49.0% female) from a single health and demographic cohort with complete baseline and two-year follow-up clinical and blood pressure records, excluding pregnant individuals and those with severe systemic illnesses.

E

Exposure

Machine learning algorithms (Logistic Regression, Decision Tree, Random Forest, Support Vector Machine, Extreme Gradient Boosting [XGBoost], and Artificial Neural Network) utilizing baseline cross-sectional and two-year follow-up demographic, anthropometric, clinical, and lifestyle variables.

C

Comparator

Comparison across different machine learning models (including conventional Logistic Regression) and actual clinical hypertension diagnosis.

O

Outcome

Prediction and classification of hypertension (defined by systolic blood pressure, diastolic blood pressure, physician diagnosis, or current use of antihypertensive medications) over a two-year follow-up period, assessed by accuracy, sensitivity, specificity, precision, F1-score, receiver operating characteristic (ROC) curve, and area under the curve (AUC).surrogate

Main Result

Effect estimate: AUC 0.971

Absolute Event Rate: 92.7% vs 84.2%

Limitations

Single health and demographic cohort limiting generalizability to other populations
Lack of external validation using independent multicenter datasets
Missing data on dietary sodium intake, psychosocial stress, medication adherence, genetic biomarkers, and environmental exposures
Data utilized from a single health and demographic cohort, which may limit generalizability to other populations
Information regarding dietary sodium intake, psychosocial stress, medication adherence, genetic biomarkers, and environmental exposures was not available

Implication

Ensemble machine learning models, particularly XGBoost, can accurately predict hypertension using combined cross-sectional and two-year longitudinal data (AUC 0.971), potentially aiding early risk identification and preventive care.

Abstract

Background: Hypertension is a major public health problem and an important modifiable risk factor for cardiovascular disease, stroke, chronic kidney disease, and premature mortality.Machine learning techniques have shown considerable potential for improving hypertension prediction by analyzing complex interactions among multiple demographic, clinical, and lifestyle variables.Aim: To evaluate the performance of different machine learning algorithms for predicting hypertension using crosssectional and two-year follow-up data from a health and demographic cohort.Methods: A prospective observational cohort study was conducted among 2,000 adults aged ≥18 years using baseline cross-sectional and two-year follow-up data.Demographic, anthropometric, clinical, and lifestyle variables were collected using standardized methods.Data preprocessing included missing value handling, normalization, categorical encoding, and feature selection.The dataset was divided into training and testing sets for model development.Logistic Regression, Decision Tree, Random Forest, Support Vector Machine, Extreme Gradient Boosting (XGBoost), and Artificial Neural Network models were developed and evaluated.Model performance was assessed using accuracy, sensitivity, specificity, precision, F1-score, receiver operating characteristic (ROC) curve, and area under the curve (AUC).Results: Among the 2,000 participants, 720 (36.0%) had hypertension and 1,280 (64.0%) were non-hypertensive.Hypertension was significantly associated with age ≥50 years, body mass index ≥25 kg/m², diabetes mellitus, family history of hypertension, smoking, and physical inactivity (p<0.05).XGBoost demonstrated the best predictive performance with an accuracy of 92.7%, sensitivity of 91.8%, specificity of 93.4%, precision of 91.1%, F1-score of 0.914, and AUC of 0.971.Random Forest showed the second-highest performance with an accuracy of 91.8% and AUC of 0.963.Feature importance analysis identified age, body mass index, systolic blood pressure, family history of hypertension, and diabetes mellitus as the strongest predictors of hypertension.Conclusion: Machine learning algorithms, particularly XGBoost, demonstrated excellent performance in predicting hypertension using combined cross-sectional and longitudinal cohort data.These models may facilitate early identification of high-risk individuals, support personalized preventive strategies, and improve clinical decisionmaking.Further external validation using multicenter population-based cohorts is recommended before routine clinical implementation.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

A 2026 study conducted a cohort in Hypertension (n=2,000). XGBoost machine learning model vs. Logistic Regression and other supervised machine learning models (Random Forest, Decision Tree, Support Vector Machine, Artificial Neural Network) was evaluated on Predictive accuracy for hypertension (AUC 0.971). The XGBoost machine learning algorithm achieved the highest performance for predicting hypertension with 92.7% accuracy and an AUC of 0.971, outperforming logistic regression (84.2% accuracy, AUC 0.889).

synapsesocial.com/papers/6ae28e75cbadf70daf6c4281 https://doi.org/10.25258/ijddt.16.72s.150

View Full Paper

Ask AI

Bookmark

Share

A Machine Learning Approach to Predict Hypertension Using Cross-Sectional & Two Years Follow Up Data from A Health & Demographic Cohort

Key Points

Study Design

Structured PICO

Main Result

Limitations

Implication

Abstract

Discussion

Cite This Study

A Machine Learning Approach to Predict Hypertension Using Cross-Sectional & Two Years Follow Up Data from A Health & Demographic Cohort

Key Points

Study Design

Structured PICO

Main Result

Limitations

Implication

Abstract

Discussion

Cite This Study

Keep reading

Keep reading