Predicting Hypertension Risk Research Articles (Page 1)

Hypertension is a major risk factor for cardiovascular diseases, and early detection is crucial for effective management. This study compares the predictive performance of three modeling techniques—Logistic Regression (LR), Neural Network (NN), and Deep Learning (DL)—in estimating the risk of hypertension. The dataset, obtained from Kaggle, consists of demographic and clinical variables with binary labels indicating the presence or absence of hypertension. Each model was trained and evaluated using RapidMiner, with performance assessed through accuracy and Root Mean Squared Error (RMSE). The results indicate that the Neural Network outperformed both Deep Learning and Logistic Regression, achieving the highest accuracy (99.88%) and the lowest RMSE (0.124). These findings suggest that shallow neural networks can provide reliable and efficient predictions for hypertension risk, sometimes even surpassing more complex deep learning architectures.

ABSTRACTThis review evaluates classical machine learning‐based hypertension prediction models, emphasizing their role in addressing global health burdens, particularly in low‐ and middle‐income countries. Hypertension affects over 1.28 billion people globally and contributes to cardiovascular disease and mortality. The review compares machine‐learning techniques with traditional methods, focusing on key datasets, evaluation metrics, and model development to advance early detection and effective hypertension management. The review used the PRISMA framework, using databases such as Google Scholar, PubMed, and IEEE explorer to identify studies published between 2020 and 2024 on machine learning techniques, predictive models, and early detection of hypertension based on relevance, methodological rigor, and inclusion criteria. The study analyzed hypertension prediction models across various countries, including the US, England, Korea, Japan, China, Indonesia, Thailand, India, Bangladesh, Nepal, and several African countries. The models' performance varied with AUC statistic values ranging from 0.6 to 0.9, indicating a wide range of predictive accuracy. Machine learning techniques generally reported higher performance metrics than traditional statistical methods. Risk factor heterogeneity was evident, with models like random forest, logistic regression, and gradient‐boosted trees showing high predictive accuracy. Emerging techniques like SMOTE (Synthetic Minority Oversampling Technique) and ensemble methods improved unbalanced data set performance. The review explores the potential of machine learning‐based hypertension prediction models in healthcare, highlighting their ability to accurately predict hypertension risk, tailor interventions to specific populations, and optimize healthcare resources in low‐ and middle‐income countries. However, challenges include data quality, model explainability, and ethical considerations. Despite these, ML integration offers scalable and cost‐effective solutions, especially in resource‐limited settings. Future research should focus on diverse datasets, advanced feature integration, and longitudinal validations.

The present study attempts to associate anthropometric markers and blood pressure, thereby determining the most effective predictive marker for hypertension. A total sample of 350 adults aged 20 to 70 years was included in the study after getting consent. Anthropometric data and blood pressure were collected using standard procedures. Significant t-tests, chi-square tests, one-way ANOVA and Pearson correlation analyses were performed between the anthropometric body adiposity measures and blood pressure. The adjusted odds ratio (AOR) was calculated to assess the risk of hypertension. The receiver operating characteristic (ROC) curve analysis was performed to find the anthropometric markers predictive cut-off values. The prevalence of hypertension was 23.4% in males and 22.3% in females, respectively. The nutritional status of overweight and obese individuals was greater in females (38.9% and 13.1%) compared to males (37.1% and 8.6%). Anthropometric indicators, including BMI, WC, WHR, and WHtR, strongly correlate with elevated blood pressure in both sexes. The ROC analysis shows WHtR has the highest area under curve (AUC) value of 0.669 in males, whereas BMI has an AUC of 0.804 in females, indicating it is a more effective predictor for high blood pressure. A multivariate logistic regression analysis showed that age (6.20; 5.30), anxiety (8.13; 6.69), alcohol use (11.81; 6.64), physical activity (8.32; 9.17), smoking (10.05; 10.44), and salt intake (6.31; 5.40) were all independently associated risk factors of hypertension, irrespective of sex. The study concluded that anthropometric markers such as BMI, WC, WHR, and WHtR could predict hypertension risk among the Phalee Tangkhul. Lifestyle factors such as physical activity, smoking, alcohol consumption, salt intake, age, and anxiety are independent risk factors for high blood pressure.

Introduction:Genome-wide polygenic risk scores (PRS) are useful for stratifying individuals’ risk for polygenic diseases such as hypertension. However, a downside of genome-wide PRS is the lack of information about the distribution of risk burden across biologic pathways. We used pathway-specific PRS to investigate these effects within common anti-hypertensive therapy-target pathways on disease risk in a cohort of West Africans.Methods:A total of 11 pathways comprising 1,149 unique genes were selected based on the targets of common anti-hypertensive agents. Pathway-specific PRS for hypertension (individuals with systolic blood pressure (SBP) ≥140 mmHg, diastolic blood pressure (DBP) ≥90 mmHg, or taking anti-hypertensive medications) were computed in a cohort of 2,395 individuals. The model was then validated and tested in 1,614 and 966 separate individuals, respectively. All participants were recruited from the International Collaborative Study on Hypertension in Blacks.Results:In combined pathways analysis, PRS predicted risk better than base models fitted with only sex, age, and principal components. Compared to base models without the PRS, the incremental increases in R2 attributable to inclusion of PRS in predictive models were 2.6% for SBP (p = 0.009); 1.4% for DBP (p = 0.012); and 1.1% for mean arterial pressure (MAP) (p = 0.044.) PRS from certain pathways (MAPK, cAMP, and adrenergic signaling in cardiomyocytes) could stratify individuals in the top and bottom deciles for DBP. Adrenergic signaling in cardiomyocytes was also predictive of MAP when comparing top and bottom deciles.Conclusions:Combined pathway polygenic risk scores constructed from genes in well-defined genetic pathways predict hypertension risk in individuals of African ancestry. However, pathway-specific PRS’s relatively low predictability supports the need to explore the broader influence of genetic, environmental, and epigenetic factors that cannot be captured by pathway-specific PRS alone.

Hypertension, known as the "silent killer," is one of the leading causes of global mortality, with a steadily increasing prevalence. Worldwide, the prevalence of hypertension reaches approximately 30%, with only 50% of cases being diagnosed and a low level of treatment adherence. Hypertension symptoms are often invisible, making early detection crucial to preventing serious complications. This paper aims to develop a hypertension prediction system using the Decision Tree and Random Forest algorithms, which are machine learning techniques used to solve classification and regression problems. These algorithms can predict hypertension risk based on clinical data, such as age, medical history, and lifestyle. The findings of this paper indicate that the Decision Tree and Random Forest algorithms are effective in predicting hypertension risk, achieving accuracies of 99.6% and 99.5%, respectively. This prediction system can provide fast and accurate information, assisting healthcare professionals in designing appropriate intervention strategies while also supporting better medical decision-making.

One-fourth of Indians are hypertensive, and the majority relies on groundwater for drinking. But the role of groundwater physicochemical properties and contamination in hypertension remains understudied. The study investigates the association between physicochemicalgroundwatercharacteristics andcontaminants and hypertension risk in India. This study useddata from the fifth round of the National Family Health Survey (NFHS-5 collected 2019-2021), including health, socio-demographics, and food and dietary information (n = 712,666 individuals). The physicochemical characteristics of groundwater data were derived from the Central Groundwater Board (CGWB, 2019-2021). This groundwater data from raster maps was linked to NFHS-5 records using cluster shapefiles and merging them with individual records via cluster IDs. Bivariate and multivariable regressions were used to identify factors associated with hypertension at the individual level. Moran's I statistics, Local Indicator of Spatial Association (LISA) cluster maps, and the Spatial Error Model(SEM) were used at district levels to investigate the spatial association. Machine learning models, including Artificial Neural Networks(ANN), Random Forest and Extreme Gradient Boosting(XGBoost), were used to predict hypertension risk zones. Physicochemical drinking water composition is a key factor in hypertension risk. Elevated groundwater pH (>8.5, AdjustedOdds Ratio(AOR): 2.12), electrical conductivity (>300 μS/cm, AOR: 1.06), sulphate (>200 mg/L, AOR: 1.16), arsenic (>0.01 mg/L, AOR: 1.09), nitrate (>45 mg/L, AOR: 1.07), and magnesium (>30 mg/L, AOR: 1.03) are associated to higher odds of hypertension. The Random Forest model demonstrated the highest predictive performance, with a coefficient of determination (R²) of 0.9970, mean absolute error (MAE) of 0.0012, and mean squared error (MSE) of 0.0077. It effectively identified high-risk zones in the northwestern (Delhi, Punjab, Haryana, and Rajasthan) and eastern (West Bengal and Bihar) regions of India. This study highlights how important groundwater quality is in determining the incidence of hypertension, pointing to groundwater physicochemical properties and contaminants such as electrical conductivity, sulphate, arsenic, nitrate, and magnesium as essential factors. Our research is the first of its kind to comprehensively map hypertension risk zones using machine learning models and geospatial analysis. The findings highlight that water quality is a modifiable risk factor, reinforcing the need for improved drinking water supply systems, regular water quality testing, and targeted interventions in high-risk regions. This study emphasizes the importance of intersectoral collaborations to enhance public health outcomes.

Hypertension is a major cause of cardiovascular disease, making early risk prediction essential. According to WHO, hypertension cases are estimated to reach 1.28 billion by 2023. This study aims to optimize the K-Nearest Neighbor (KNN) algorithm for predicting hypertension risk through hyperparameter tuning. Three methods Grid SearchCV, Bayes SearchCV, and Random SearchCV are compared to determine the best parameter configuration. The dataset, obtained from Kaggle, consists of 520 balanced samples (260 positive and 260 negative) with 18 health-related features such as age, gender, blood pressure, cholesterol, glucose, and others. After preprocessing, the KNN model is tuned using each method by testing combinations of neighbors (k), weight types, and distance metrics. Results show Bayes SearchCV achieved the highest accuracy of 92%, outperforming the baseline KNN model, which had 85% accuracy. The ROC AUC score of 0.96191 also indicates excellent classification performance. In conclusion, Bayes SearchCV significantly improves KNN's predictive ability in hypertension risk classification.

Hypertension remains a major global health challenge, contributing to significant morbidity and mortality. Advances in artificial intelligence (AI) and machine learning (ML) are transforming hypertension care by enhancing blood pressure (BP) measurement, risk assessment, and personalized treatment. AI-powered technologies have the potential to enable accurate non-invasive BP monitoring and facilitate tailored lifestyle modifications, enhancing adherence and outcomes. ML models can also predict hypertension risk based on demographic, lifestyle, and clinical data, enabling earlier intervention and prevention strategies. However, challenges such as the lack of standardized validation protocols and potential biases in AI systems may widen health disparities. Future research must prioritize rigorous validation across diverse populations and ensure algorithm transparency. By leveraging AI responsibly, we can revolutionize hypertension management, enhance health equity, and improve cardiovascular outcomes.

Insulin resistance (IR) is a complex abnormality and associated with hypertension. We aimed to assess the associations of six alternate IR measures and risk of hypertension, and to compare the predictive values for hypertension. We assessed 11,223 non-hypertensive Chinese adults enrolled in The Rural Chinese Cohort Study during 2007-2008. Six surrogate IR indexes were new visceral adiposity index (NVAI), Chinese visceral adiposity index (CVAI), weight-adjusted waist index (WWI), lipid accumulation product (LAP), triglyceride glucose (TyG) index, and visceral adiposity index (VAI). The relative risks (RR) and 95% confidence intervals (95% CI) of the six IR indicators and hypertension were estimated by using modified Poisson regression models with three adjusted models. During a median follow-up of 11.1 years, 3373 (30.05%) study participants developed hypertension. The cumulative incidence of hypertension showed an increasing trend with higher levels of all six IR surrogates. Significant associations of all the IR measures with incident hypertension were found in fully adjusted model, and the highest quartile group RRs (95% CIs) for hypertension were, 2.19 (1.88-2.55), 1.60 (1.42-1.81), 1.38 (1.25-1.53), 1.47 (1.31-1.65), 1.18 (1.04-1.34) and 1.25 (1.08-1.44) for NVAI, CVAI, WWI, LAP, TyG index and VAI, respectively, compared with lowest quartile group. Further, NVAI had the maximum predictive power for hypertension among six IR measures with the largest AUC of 0.706 (0.697-0.714). NVAI, CVAI, WWI, LAP, and TyG index were all independently associated with greater risk of incident hypertension, among which NVAI is the most powerful predictor for hypertension in rural Chinese adults. Association of the surrogate insulin resistance indexes with the risk of hypertension.

Hypertension, commonly known as high blood pressure, is a prevalent and serious condition affecting a significant portion of the adult population globally. It is a chronic medical issue that, if left unaddressed, can lead to severe health complications, including kidney problems, heart disease, and stroke. This study aims to develop a feature selection model using the XGBoost algorithm to identify specific single nucleotide polymorphisms (SNPs) as biomarkers for detecting hypertension risk. We propose using the high dimensionality of genetic variations (i.e., SNPs) to build a classifier model for prediction. In this study, SNPs were used as markers for hypertension in patients. We utilized the OpenSNP dataset, which includes 19,697 SNPs from 2,052 samples. Extreme gradient boosting (XGBoost) is an ensemble machine learning method employed here for feature selection, which incrementally adjusts weights in a series of steps. The experimental results identified 292 SNPs that exhibited high performance, with an F1-score of 98.55%, precision of 98.73%, recall of 98.38%, and overall accuracy of 98%. This study provides compelling evidence that the XGBoost feature selection method outperforms other representative feature selection methods, such as genetic algorithms, analysis of variance, chi-square, and principal component analysis, in predicting hypertension risk, demonstrating its effectiveness. We developed a model for predicting hypertension using the SNPs dataset. The high dimensionality of SNP data was effectively managed to identify significant features as biomarkers using the XGBoost feature selection method. The results indicate high performance in predicting the risk of hypertension.

Background Diabetes is closely related to hypertension, and insulin resistance-related indices are novel metrics used to evaluate the risk of diabetes and cardiovascular diseases. This study aims to explore the relationships between the TyG index, METS-IR, TG/HDL-C, and HOMA-IR with hypertension. Methods Data from the NHANES spanning ten consecutive survey cycles from 1998 to 2018 were utilized, focusing on adults with complete blood pressure data and comprehensive information for calculating the TyG index, METS-IR, TG/HDL-C, and HOMA-IR. A multivariable logistic regression model was employed to examine the relationship between insulin resistance indices and hypertension as well as blood pressure levels, while subgroup analyses were conducted to explore potential influencing factors. RCS curves were used to describe both linear and non-linear relationships. Results This NHANES-based study included 16,062 adults. Regardless of the adjustment for covariates, significant associations were found between the TyG index, METS-IR, TG/HDL-C, HOMA-IR and hypertension risk. The ROC curve demonstrated the stability of the TyG index, METS-IR, TG/HDL-C, and HOMA-IR in predicting hypertension risk. The RCS curves uncovered a linear relationship between the TyG index, METS-IR, and hypertension, whereas TG/HDL-C and HOMA-IR exhibited a non-linear association with hypertension. Subgroup analyses indicated that smoking and diabetes may influence the relationship between insulin resistance-related indices and hypertension. Conclusion Elevated levels of the insulin resistance indices TyG index, METS-IR, TG/HDL-C, and HOMA-IR are closely associated with hypertension risk. These indices can serve as effective markers for monitoring hypertension risk in clinical practice. However, larger-scale prospective cohort studies are needed to validate these findings and further explore the clinical application potential of the TyG index, METS-IR, TG/HDL-C, and HOMA-IR as tools for cardiovascular risk assessment. Such studies will help elucidate the specific causal relationships between these insulin resistance-related indices and hypertension and advance their practical application in clinical settings.

Hypertension remains a critical health issue, and complications such as cardiovascular disease, stroke, and renal failure similarly remain a global health concern. This study compared six supervised machine learning models – Support Vector Machines, k-nearest Neighbors, Random Forest Classifier, Naïve Bayes Classifier, Tree Bagging, and Extreme Gradient Boosting, based on the data from 2322 participants. The primary elements were SBP measured as equal to or more than 120 mmHg, BMI, Age, and the number of haemoglobin grams per litre, as well as demographic data. The research found that Random Forest yielded the highest evaluation metrics in Oversampling, with an accuracy of 100 %, balanced Accuracy of 100 %, Sensitivity of 100 %, specificity of 100 %, and AUC of 100 %; hence proved to be the best model to address the hypertension risk among patients. The feature importance of the SBP turned out to be higher according to the SHAP analysis, considering the "No" class where the SHAP value equalled 0.24, followed by BMI (0.05) and Gender (0.06). Variables such as advanced HIV status and log-centered creatinine showed negligible impact (SHAP value = 0.00). The random forest model was accurate and steady across all performance criteria, outperforming all other models with the No Information Rate (0.978) while illustrating the significance of physiological aspects of hypertension risk assessment. These results demonstrate the capability of Random Forest in predicting hypertension risk and give important suggestions for enhancing screening methods and specific public health initiatives.

Hypertension remains a critical health issue, and complications such as cardiovascular disease, stroke, and renal failure similarly remain a global health concern. This study compared six supervised machine learning models – Support Vector Machines, k-nearest Neighbors, Random Forest Classifier, Naïve Bayes Classifier, Tree Bagging, and Extreme Gradient Boosting, based on the data from 2322 participants. The primary elements were SBP measured as equal to or more than 120 mmHg, BMI, Age, and the number of haemoglobin grams per litre, as well as demographic data. The research found that Random Forest yielded the highest evaluation metrics in Oversampling, with an accuracy of 100 %, balanced Accuracy of 100 %, Sensitivity of 100 %, specificity of 100 %, and AUC of 100 %; hence proved to be the best model to address the hypertension risk among patients. The feature importance of the SBP turned out to be higher according to the SHAP analysis, considering the "No" class where the SHAP value equalled 0.24, followed by BMI (0.05) and Gender (0.06). Variables such as advanced HIV status and log-centered creatinine showed negligible impact (SHAP value = 0.00). The random forest model was accurate and steady across all performance criteria, outperforming all other models with the No Information Rate (0.978) while illustrating the significance of physiological aspects of hypertension risk assessment. These results demonstrate the capability of Random Forest in predicting hypertension risk and give important suggestions for enhancing screening methods and specific public health initiatives.

This study aims to enhance individual hypertension risk prediction in Indonesia using machine learning (ML) models. The research investigates the predictive accuracy of models with and without incorporating personal hypertension history, seeking to understand how data limitations impact model performance in a low-resource setting. Data from the SATUSEHAT IndonesiaKu (ASIK) system were preprocessed and filtered to create a dataset of 9.58 million adult health records. Two primary model variations were compared: Model A (incorporating patient history) and Model B (excluding patient history). We evaluated the model using five algorithms: XGBoost, LightGBM, CatBoost, Logistic Regression, and Random Forest. Model performance was assessed using the Area Under the Curve (AUC), sensitivity, and specificity metrics. Model A achieved superior predictive accuracy (AUC = 0.85) compared to Model B (AUC = 0.78). To mitigate potential bias, Model B was selected for further in-depth development. Evaluation of model B reveals that XGBoost and LightGBM algorithm achieved the highest performance (AUC 0.78) and LightGBM emerged as the best algorithm based on its performance. SHAP analysis was conducted and identified key predictors such as age, family history of hypertension, body weight, and waist circumference. This study finds that while a patient’s personal history of hypertension significantly enhances predictive accuracy, robust ML models can effectively predict hypertension risk using other accessible demographic, clinical, and lifestyle features. Model B offers a valuable and generalizable approach for broader risk screening, particularly where patient history may be unavailable or unreliable, while also providing insights into key modifiable and non-modifiable determinants of hypertension.Supplementary InformationThe online version contains supplementary material available at 10.1007/s10916-025-02253-5.

Determinants and prediction of hypertension among Chinese middle-aged and elderly adults with diabetes: A machine learning approach

Background and Objective: The integration of artificial intelligence (AI) in healthcare has revolutionized the management of various diseases, including hypertension. This review explores the potential of AI applications in improving hypertension diagnosis, treatment, and patient outcomes. Methods: A comprehensive literature review was conducted using databases such as PubMed, Scopus, and IEEE Xplore. Studies published between 2010 and 2023 were included, focusing on AI models used for hypertension management, such as machine learning algorithms and predictive analytics. Results: AI has demonstrated significant accuracy in predicting hypertension risk factors, optimizing treatment plans, and monitoring patient adherence. Machine learning models, such as neural networks and decision trees, have shown up to 95% accuracy in early hypertension detection. Furthermore, AI-driven personalized treatment plans have resulted in improved patient outcomes and reduced healthcare costs. Conclusions: AI holds tremendous potential in transforming hypertension management by enhancing diagnostic precision and personalizing treatment strategies. Future research should focus on integrating AI with electronic health records and developing real-time monitoring systems to further improve patient care.

Background: Most guidelines have used traditional methods (TM), such as regression, to analyze hypertension risk factors and predict high-risk populations. Many studies have developed hypertension prediction models using TM and machine learning (ML), but the risk factors and prediction accuracy vary. Objective: This work reviews articles that predict hypertension using ML. It aims to find the risk factors that contribute the most and evaluate ML performance utilizing the area under the receiver operating characteristic curve (AUC) indicator. Method: We conducted a systematic review using Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) for articles published from 2010 – 2023 through PubMed, Google Scholar, Science Direct, and Taylor & Francis. The keywords used were “hypertension,” “risk factor,” “prediction,” and “machine learning.” We used “AUC” as the inclusion criteria and filtered the articles with easy-to-collect risk factors, with 500 minimum subjects. Results: After applying the inclusion criteria, we obtained 38 articles. Only eight articles use easy-to-collect risk factors, including age, gender, family history of hypertension, alcohol consumption, fruit and vegetable consumption, physical activity, smoking, and body mass index. Compared to TM, ML algorithms such as Random Forest (RF), Gradient Boosting, Extreme Gradient Boosting, and Support Vector Machine have more significant values in predicting hypertension risk. For a large general population, the highest AUC for ML was 0.92 using RF, and for TM, it was 0.829 using logistic regression. Conclusion: ML can become an alternative method for predicting hypertension, providing better performance than TM.

Machine learning methods are widely used within the medical field to enhance prediction. However, little is known about the reliability and efficacy of these models to predict long-term medical outcomes such as blood pressure using lifestyle factors, such as diet. The authors assessed whether machine-learning techniques could accurately predict hypertension risk using nutritional information. A cross-sectional study using data from the National Health and Nutrition Examination Survey (NHANES) between January 2017 and March 2020. XGBoost was used as the machine-learning model of choice in this study due to its increased performance relative to other common methods within medical studies. Model prediction metrics (e.g., AUROC, Balanced Accuracy) were used to measure overall model efficacy, covariate Gain statistics (percentage each covariate contributes to the overall prediction) and SHapely Additive exPlanations (SHAP, method to visualize each covariate) were used to provide explanations to machine-learning output and increase the transparency of this otherwise cryptic method. Of a total of 9650 eligible patients, the mean age was 41.02 (SD=22.16), 4792 (50%) males, 4858 (50%) female, 3407 (35%) White patients, 2567 (27%) Black patients, 2108 (22%) Hispanic patients, and 981 (10%) Asian patients. From evaluation of model gain statistics, age was found to be the single strongest predictor of hypertension, with a gain of 53.1%. Additionally, demographic factors such as poverty and Black race were also strong predictors of hypertension, with gain of 4.33% and 4.18%, respectively. Nutritional Covariates contributed 37% to the overall prediction: Sodium, Caffeine, Potassium, and Alcohol intake being significantly represented within the model. Machine Learning can be used to predict hypertension.

Hypertension and abnormal liver enzymes are common health issues that frequently coexist, and recent research has suggested a possible association between them, warranting further investigation. Therefore, the aim of this study is to explore the relationship between liver enzymes and hypertension. This prospective cohort study utilized data from the Kharameh cohort study, which is a branch of the Prospective Epidemiological Studies in Iran (PERSIAN) database. The study included 7710 participants aged between 40 and 70 years. Hypertension was defined in accordance with the European guidelines for hypertension management, and the association between liver enzymes and hypertension was modeled and predicted using Firth logistic regression. This study investigated the association between liver enzymes and hypertension risk in a cohort of 7710 individuals aged 40-70 years. The results showed that higher levels of alanine aminotransferase (ALT), gamma glutamyl transferase (GGT), and alkaline phosphatase (ALP) were associated with an increased risk of hypertension, and this relationship remained significant even after adjusting for potential confounding factors. Additionally, separate analyses within age subgroups revealed a significant association between ALP concentration and high blood pressure in certain age ranges. The study demonstrated a significant association between high levels of ALT, GGT, and ALP and an increased risk of hypertension, regardless of other risk factors. These results suggest that monitoring liver enzymes, specifically ALT, GGT, and ALP, could serve as a useful tool to predict hypertension risk and identify individuals who could benefit from early intervention. Overall, these findings underscore the importance of monitoring liver function in preventing and managing hypertension.

BackgroundHypertension is a major public health problem, and its resulting other cardiovascular diseases are the leading cause of death worldwide. In this study, we constructed a convenient and high-performance hypertension risk prediction model to assist in clinical diagnosis and explore other important influencing factors.MethodsWe included 8,073 people from NHANES (2017—March 2020), using their 120 features to form the original dataset. After data pre-processing, we removed several redundant features through LASSO regression and correlation analysis. Thirteen commonly used machine learning methods were used to construct prediction models, and then, the methods with better performance were coupled with recursive feature elimination to determine the optimal feature subset. After data balancing through SMOTE, we integrated these better-performing learners to construct a fusion model based for predicting hypertension risk on stacking strategy. In addition, to explore the relationship between serum ferritin and the risk of hypertension, we performed a univariate analysis and divided it into four level groups (Q1 to Q4) by quartiles, with the lowest level group (Q1) as the reference, and performed multiple logistic regression analysis and trend analysis.ResultsThe optimal feature subsets were: age, BMI, waist, SBP, DBP, Cre, UACR, serum ferritin, HbA1C, and doctors recommend reducing salt intake. Compared to other machine learning models, the constructed fusion model showed better predictive performance with precision, accuracy, recall, F1 value and AUC of 0.871, 0.873, 0.871, 0.869 and 0.966, respectively. For the analysis of the relationship between serum ferritin and hypertension, after controlling for all co-variates, OR and 95% CI from Q2 to Q4, compared to Q1, were 1.396 (1.176–1.658), 1.499 (1.254–1.791), and 1.645 (1.360–1.989), respectively, with P < 0.01 and P for trend <0.001.ConclusionThe hypertension risk prediction model developed in this study is efficient in predicting hypertension with only 10 low-cost and easily accessible features, which is cost-effective in assisting clinical diagnosis. We also found a trend correlation between serum ferritin levels and the risk of hypertension.