Evaluating the Predictive Capacity of Obesity Indices for High Blood Pressure Among Zambians Aged 18–69 Years Using Machine Learning: Evidence From WHO STEPS Survey

Evaluating the Predictive Capacity of Obesity Indices for High Blood Pressure Among Zambians Aged 18–69 Years Using Machine Learning: Evidence From WHO STEPS Survey | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Evaluating the Predictive Capacity of Obesity Indices for High Blood Pressure Among Zambians Aged 18–69 Years Using Machine Learning: Evidence From WHO STEPS Survey Samuel Mutasha, Given Moonga, Wilbroad Mutale, Clyde Mulenga, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9585939/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background: Hypertension is a leading contributor to cardiovascular morbidity in sub-Saharan Africa. BMI, waist circumference (WC), waist-to-height ratio (WHtR), and waist-to-hip ratio (WHpR) are widely used for screening, yet their optimal thresholds and predictive utility vary by population. This study evaluated these indices for hypertension prediction in Zambian adults using traditional and machine learning approaches. Methods: This cross-sectional study assessed hypertension in 4,302 adults aged 18–69 years from the 2017 Zambia WHO STEPS survey. ROC analysis with Youden's Index identified optimal sex-stratified cut-offs for BMI, WC, WHtR, and WHpR. Five machine learning classifiers — logistic regression, linear SVM, random forest, gradient boosting, and XGBoost — were then trained on anthropometric, sociodemographic, and behavioural covariates using stratified 5-fold cross-validation, with recall prioritised as the primary metric given the imbalanced outcome. Results: Hypertension prevalence was 23.7% (n=1,021), higher in men (25.4% vs. 22.7%; p=0.046) and urban than rural residents (27.0% vs. 21.7%; p<0.001). WC and WHtR showed the highest discrimination (men: AUC 0.63/0.62; women: 0.66/0.66); locally derived cut-offs were below WHO thresholds (WC: 80.35/81.85 cm; WHtR: 0.48/0.53; BMI: 21.97/24.12 kg/m²). Logistic regression and linear SVM achieved the best ROC-AUC (0.74; recall 0.69–0.70; F1=0.51); ensemble models showed higher accuracy (0.78–0.79) but lower recall (0.22–0.26). Conclusion: Central adiposity measures (WC and WHtR) outperformed BMI for hypertension identification in Zambian adults, with locally derived cut-offs consistently below WHO thresholds. Logistic regression and linear SVM demonstrated the best predictive performance, supporting their utility in risk stratification using routinely collected data. Artificial Intelligence and Machine Learning hypertension blood pressure obesity indices anthropometric indices machine learning prediction Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction Hypertension and obesity are now among the most pressing global health challenges, with their prevalence rising sharply in both high- and low-income countries. Worldwide, over 1.28 billion adults are estimated to have hypertension, and more than 1 billion are living with obesity, a figure that has more than doubled since 1990, with the fastest increases seen in low- and middle-income regions, including sub-Saharan Africa 1–3 . In sub-Saharan Africa, hypertension prevalence ranges from 20% to over 40% in adults, with urbanization, dietary shifts, and reduced physical activity increasing a parallel surge in obesity 4–6 . Zambia mirrors these trends with recent national surveys indicating that about one in three adults has raised blood pressure, and overweight rates are rising, especially among women and urban residents 7,8 . Despite this, many cases remain undiagnosed or poorly managed, and the health system faces challenges in early detection and prevention. In Africa, the increasing prevalence of hypertension and obesity is driven by poor dietary habits, physical inactivity, smoking, alcohol consumption, and environmental factors such as urbanization and pollution 9 . Zambia demonstrates this trend, with an 18.9% national hypertension prevalence, rising to 34.8% in urban areas like Lusaka 10 . Obesity affects 24.4% of Zambian adults, with a higher prevalence among women 10 . Given the link between obesity and hypertension, assessing anthropometric indices is key for improving early detection and prevention strategies. However, the optimal anthropometric indicators and cut-off values for predicting high blood pressure may vary across populations, underscoring the need for population-specific evaluation. Anthropometric indices such as body mass index (BMI), waist circumference (WC), waist-to-hip ratio (WHR), and waist-to-height ratio (WHtR) are widely used to assess obesity and predict hypertension risk 11–13 . The World Health Organization (WHO) recommends BMI cut-offs of 25 kg/m² for overweight and 30 kg/m² for obesity, and WC thresholds of ≥94 cm for men and ≥80 cm for women (Europid values) 12,13,13 . However, evidence suggests these cut-offs may not optimally capture risk in African populations, where body composition and fat distribution differ from those in Western populations 14–16 . For example, at the same BMI, Africans may have higher cardiovascular risk, and indices like WC or WHtR may outperform BMI in predicting hypertension 11–13 . This raises a key question: should Zambia adopt cut-off values that are validated within its own population? Recent advances in machine learning (ML) provide a powerful approach to handle complex biomedical and healthcare prediction problems, enabling the extraction of patterns from large, high‑dimensional data and often improving classification and prediction beyond traditional statistical methods 17,18 . Machine learning models, including random forests, XGBoost, and neural networks, can analyze large, complex datasets to identify the most predictive anthropometric indices and optimal cut-off points for hypertension risk, tailored to specific populations 19–21 . Studies from Africa and other low-resource settings have shown that ML-based models can outperform traditional regression in predicting hypertension, especially when using easy-to-collect, non-invasive measures like age, BMI, WC, and WHtR 22,23 . For instance, Islam and colleagues in Albania demonstrated that ML models achieved high accuracy in hypertension prediction, and identified age, BMI, and WC as top predictors 22,24 . Similarly, multi-country analyses using WHO STEPS data have highlighted the need for country-specific models and thresholds to maximize predictive performance 25 . Leveraging on ML with WHO STEPS survey data presents an opportunity to compare the predictive value of different anthropometric indices, explore whether local cut-offs improve risk stratification, and ultimately inform more effective, context-specific screening and prevention strategies in Zambia. This approach could help close the gap in early detection, reduce the burden of undiagnosed hypertension, and support the development of tailored interventions for Zambia’s diverse adult population. Methods Study Design and Data Source This cross-sectional study used secondary data from the 2017 Zambia WHO STEPwise Approach to NCD Risk Factor Surveillance (STEPS) survey. The survey employed a nationally representative sampling frame covering urban and rural areas across all ten provinces, targeting adults aged 18–69 years. Data were collected in three steps: self-reported behavioural and sociodemographic information (Step 1), standardised physical and anthropometric measurements (Step 2), and biochemical measurements from fasting blood samples (Step 3). This study drew on data from all three steps. Sampling and Participants The survey used a multistage cluster sampling design. Standard Enumeration Areas were selected using probability proportional to size, households were selected systematically within each area, and one eligible adult per household was randomly chosen. Non-respondents were not replaced. Adults aged 18–69 years with complete blood pressure and anthropometric measurements were eligible. Pregnant women were excluded at enrollment because physiological changes during pregnancy affect blood pressure and body composition. The stated eligibility were largely applied at enrollment, this yielded a final analytical sample of 4,302 as shown in figure 1. Sample Size justification All eligible participants were retained to maximize the stability of machine learning model estimates. The minimum required sample size was calculated using the recommended diagnostic test accuracy framework 26 . Assumptions included 95% confidence level (Z = 1.96), expected sensitivity of 80%, expected specificity of 75%, hypertension prevalence of 32%, and a maximum confidence interval width of 10%, the minimum required sample was approximately 1,300. However, the study employed a full enumeration with the available 4,302 participants. Study Variables Outcome. Hypertension was defined as systolic blood pressure ≥140 mmHg, diastolic blood pressure ≥90 mmHg, or current use of antihypertensive medication 27 , coded as a binary variable (1 = hypertensive, 0 = normotensive). Anthropometric obesity indices. Four indices were evaluated: Body Mass Index (weight in kg/height in m²), Waist Circumference (measured at the midpoint between the lowest rib and iliac crest), Waist-to-Hip Ratio, and Waist-to-Height Ratio. All measurements followed WHO standardised procedures. Covariates. Sociodemographic variables included age, sex, residence (urban/rural), province, education level, and marital status. Behavioural variables included smoking status, alcohol use, physical activity, and salt intake frequency. Total cholesterol was included as a biochemical predictor. Handling Class Imbalance Approximately 24% of participants were hypertensive. To prevent classifiers from favouring the majority class, stratified sampling preserved the original class distribution in each cross-validation fold. Recall was prioritised as the primary metric to maximise detection of true hypertension cases 28 . Precision-Recall curves supplemented standard ROC curves. No synthetic oversampling was applied to preserve the original population distribution. Data Analysis Analyses were conducted using STATA version 17 29 for survey-weighted descriptive statistics. and Python 30 for machine learning using the popular packages scikit-learn 31 and XGBoost 32 . Exploratory analysis. Distributions of anthropometric indices were examined, a feature correlation matrix was computed, and mutual information scores were calculated to assess the standalone predictive information contributed by each variable toward the hypertension outcome. ROC and cutoff analysis. Receiver Operating Characteristic curves assessed each obesity index's discriminatory ability for hypertension. The Area Under the Curve (AUC) with 95% confidence intervals was computed overall and stratified by sex. Optimal cutoff values were identified using Youden's Index (sensitivity + specificity − 1). Sensitivity-oriented thresholds were also explored for screening applications. Classification models. Five models were fitted: Logistic Regression, Linear Support Vector Machine, Random Forest, Gradient Boosting Machine, and Extreme Gradient Boosting (XGBoost). All models used default hyperparameters to maintain comparability and avoid overfitting in the absence of an external validation dataset. Validation. Internal validation used stratified 5-fold cross-validation, preserving the outcome class distribution in each fold. Performance was averaged across folds. Performance metrics. Models were evaluated using ROC AUC, Precision-Recall AUC, accuracy, precision, recall, and F1-score. Receiver Operators Characteristics (ROC) AUC ≥0.70 was considered acceptable. Precision Re-call (PR) AUC was interpreted relative to the baseline prevalence. Accuracy was interpreted cautiously given the class imbalance. Feature importance. Permutation-based feature importance quantified each predictor's contribution by measuring the drop in model performance when that variable was randomly shuffled. Ethical Considerations Permission to use the 2017 Zambia STEPS dataset was formally granted by the World Health Organization and the Zambia Ministry of Health. Ethical clearance for this secondary analysis was obtained from the University of Zambia Biomedical Research Ethics Committee (UNZABREC Approval No. REF. 6-492-2025). The study was additionally registered with and approved by the National Health Research Authority of Zambia (NHRA-2135/15/04/2025) in accordance with national research governance requirements. The dataset was fully anonymised prior to analysis and no personal identifiers were accessed at any stage. All analyses were conducted in accordance with the principles of the Declaration of Helsinki regarding confidentiality, data protection, and responsible use of human subject data. Results Participant Characteristics The final sample included 4,302 adults aged 18–69 years, with women comprising 62.5% and men 37.5%. High blood pressure was present in 23.7% of participants (n = 1,021). Those with high blood pressure were considerably older (median 42 years, IQR 31–57) compared with normotensive participants (median 32 years, IQR 24–42; p < 0.001). Men had a higher prevalence than women (25.4% vs 22.7%; p = 0.046), urban residents higher than rural (27.0% vs 21.7%; p < 0.001), and alcohol users higher than non-users (26.6% vs 22.3%; p = 0.002). Neither smoking nor salt intake showed a significant association with high blood pressure. Median total cholesterol, waist circumference, and BMI were all significantly elevated among hypertensive compared with normotensive participants (all p < 0.001). Table 1 presents the full sample characteristics. Table 1 . Population Characteristics of the Study Population Stratified by Blood Pressure Status (N = 4,302). Variables Overall N = 4,302 No High BP n = 3,281 (76.3%) High BP n = 1,021 (23.7%) p-value Cholesterol (mmol/L), median (IQR) 3.35 (2.67–4.20) 3.10 (2.59–3.9) 3.59 (2.75–4.50) < 0.001ᶜ Waist circumference (cm), median (IQR) 80 (74–87) 77 (72–84) 83 (76–93) < 0.001ᶜ Body mass index (kg/m²), median (IQR) 22.9 (20.5–26.3) 22.0 (20.3–24.4) 23.8 (20.9–28.1) < 0.001ᶜ Waist-to-height ratio, mean (SD) 0.50 (0.08) 0.49 (0.07) 0.53 (0.10) < 0.001ᵇ Waist-to-hip ratio, mean (SD) 0.85 (0.08) 0.84 (0.08) 0.86 (0.10) < 0.001ᵇ Sex, n (%) Male 1,614 (37.5) 1,204 (74.6) 410 (25.4) Female 2,688 (62.5) 2,077 (77.3) 611 (22.7) 0.046ᵃ Age, median (IQR) 32 (24–42) 32 (24–42) 42 (31–57) < 0.001ᶜ Residence, n (%) Rural 2,658 (61.8) 2,081 (78.3) 577 (21.7) Urban 1,644 (38.2) 1,200 (73.0) 444 (27.0) < 0.001ᵃ Education level, n (%) No primary 1,546 (36.0) 1,159 (75.0) 387 (25.0) Primary 1,036 (24.1) 791 (76.4) 245 (23.7) Lower secondary 826 (19.2) 667 (80.8) 159 (19.3) Higher secondary 564 (13.1) 438 (77.7) 126 (22.3) College 327 (7.6) 226 (69.1) 101 (30.9) < 0.001ᵃ Marital status, n (%) Single 934 (21.8) 767 (82.1) 167 (17.9) Married/cohabiting 2,627 (61.2) 2,020 (76.9) 607 (23.1) Other 731 (17.0) 491 (67.2) 240 (32.8) < 0.001ᵃ Current smoker, n (%) Yes 475 (11.0) 354 (74.5) 121 (25.5) No 3,826 (89.0) 2,927 (76.5) 899 (23.5) 0.339ᵃ Salt intake, n (%) Rarely/never 1,352 (31.5) 1,024 (75.7) 328 (24.3) Sometimes 1,319 (30.8) 1,036 (78.5) 283 (21.5) Often/always 1,618 (37.7) 1,211 (74.9) 407 (25.1) 0.055ᵃ Alcohol use, n (%) Yes 1,398 (32.5) 1,026 (73.4) 372 (26.6) No 2,903 (67.5) 2,255 (77.7) 648 (22.3) 0.002ᵃ ᵃChi-square test; ᵇIndependent samples t-test; ᶜWilcoxon rank-sum test. BP – blood pressure; IQR – interquartile range; SD – standard deviation. p-values < 0.05 considered statistically significant. Sex-Stratified Anthropometric Differences Among participants with high blood pressure, women had higher mean BMI (26.28 ± 6.28 vs 23.87 ± 5.14 kg/m²), waist circumference (87.29 ± 16.65 vs 82.95 ± 13.0 cm), and waist-to-height ratio (0.55 ± 0.10 vs 0.50 ± 0.08) than men. Waist-to-hip ratio was similar between sexes. The same pattern held among normotensive participants. All differences between hypertensive and normotensive groups were significant within each sex (all p < 0.001; Table 2). Table 2 . Anthropometric Index Values According to Blood Pressure Status by Sex Among Adult Residents of Zambia. Anthropometric Index High BP Yes High BP No p-value Male n 353 1,212 BMI (kg/m²) 23.87 (5.14) 22.00 (3.18) < 0.001 WstC (cm) 82.95 (13.00) 77.77 (8.24) < 0.001 WHpR 0.86 (0.09) 0.85 (0.06) < 0.001 WHtR 0.50 (0.08) 0.47 (0.05) < 0.001 Female n 480 1,960 BMI (kg/m²) 26.28 (6.28) 23.46 (4.91) < 0.001 WstC (cm) 87.29 (16.65) 79.40 (11.10) < 0.001 WHpR 0.86 (0.10) 0.83 (0.08) < 0.001 WHtR 0.55 (0.10) 0.50 (0.07) < 0.001 Both sexes combined n 833 3,172 BMI (kg/m²) 25.26 (5.94) 22.90 (4.39) < 0.001 WstC (cm) 85.45 (15.36) 78.80 (10.13) < 0.001 WHpR 0.86 (0.10) 0.84 (0.08) < 0.001 WHtR 0.53 (0.10) 0.49 (0.07) < 0.001 Data presented as mean (SD). p-values derived from independent samples t-test. BMI – body mass index; WstC – waist circumference; WHpR – waist-to-hip ratio; WHtR – waist-to-height ratio; BP – blood pressure; SD – standard deviation. Discriminatory Ability of Anthropometric Indices The ROC analysis showed that all four indices had fair discriminatory ability for high blood pressure (AUC range: 0.59–0.66). Among men, waist circumference had the highest AUC (0.63, 95% CI 0.60–0.67), followed by waist-to-height ratio (0.62), BMI (0.60), and waist-to-hip ratio (0.59). Among women, waist-to-height ratio and waist circumference performed best (both AUC 0.66), followed by BMI (0.64) and waist-to-hip ratio (0.60). Central adiposity measures consistently matched or outperformed BMI in both sexes (Figure 2). Optimal cutoff values derived from Youden's Index are shown in Table 3. The BMI cutoff was 21.97 kg/m² for men (sensitivity 58%, specificity 58%) and 24.12 kg/m² for women (sensitivity 58%, specificity 67%). For waist circumference, cutoffs were 80.35 cm (men) and 81.85 cm (women). For waist-to-height ratio, the cutoff was 0.48 for men and 0.53 for women, with the latter yielding the best combination of sensitivity (55%) and specificity (71%) among all indices in women. Table 3. Area Under the Curve (AUC), Optimal Cutoff Values, Sensitivity, and Specificity of Selected Anthropometric Indices for Predicting High Blood Pressure by Sex Among Adult Residents of Zambia. Indices Sex AUC (95% CI) Cut off value Sensitivity Specificity Younden Index High blood pressure BMI Male 0.60 [0.57 - 0.64] 21.97 0.58 0.58 0.16 Female 0.64 [0.61 - 0.67] 24.12 0.58 0.67 0.25 wstc Male 0.63 [0.60 - 0.67] 80.35 0.52 0.69 0.21 Female 0.66 [0.63 - 0.69] 81.85 0.61 0.64 0.25 whpr Male 0.58 [ 0.55 - 0.62] 0.86 0.52 0.61 0.13 Female 0.60 [0.57 - 0.63] 0.84 0.59 0.56 0.15 whtr Male 0.62 [0.59 - 0.66] 0.48 0.52 0.68 0.2 Female 0.66 [0.63 - 0.69] 0.53 0.55 0.71 0.26 Systolic bp > 140 mmHg BMI Male 0.61 [0.57 - 0.65] 22.21 0.57 0.6 0.17 Female 0.65 [0.62 - 0.69] 24.12 0.61 0.66 0.27 wstc Male 0.64 [0.60 - 0.68] 80.35 0.54 0.68 0.22 Female 0.68 [0.65 - 0.71] 82.45 0.61 0.66 0.27 whpr Male 0.59 [0.55 - 0.63] 0.86 0.55 0.6 0.15 Female 0.61 [0.58 - 0.65] 0.84 0.62 0.55 0.17 whtr Male 0.62 [0.58 - 0.66] 0.48 0.54 0.67 0.21 Female 0.68 [0.65 - 0.71] 0.53 0.59 0.7 0.29 Diastolic bp > 90 mmHg BMI Male 0.63 [0.59 - 0.67] 21.78 0.67 0.54 0.21 Female 0.63 [0.60 - 0.67] 24.12 0.59 0.66 0.25 wstc Male 0.65 [0.60 - 0.69] 81.45 0.53 0.73 0.26 Female 0.65 [0.61 - 0.68] 81.85 0.61 0.62 0.23 whpr Male 0.58 [ 0.53 - 0.62] 0.87 0.5 0.65 0.15 Female 0.58 [0.54 - 0.61] 0.84 0.59 0.55 0.14 whtr Male 0.64 [ 0.59 - 0.68] 0.49 0.51 0.75 0.26 Female 0.64 [0.61 - 0.67] 0.54 0.52 0.72 0.24 AUC values presented with 95% confidence intervals (CI) in parentheses. Optimal cutoff values determined by the Youden Index.BMI – body mass index; WstC – waist circumference; WHpR – waist-to-hip ratio; WHtR – waist-to-height ratio; AUC – area under the receiver operating characteristic curve; BP – blood pressure. Exploratory Feature Analysis and Mutual Information (MI) The feature correlation matrix (Figure 3) showed moderate correlations among anthropometric indices (e.g., BMI with waist circumference, r = 0.70), reflecting shared adiposity constructs. Pairwise correlations between individual predictors and high blood pressure were modest (r = 0.20–0.23), confirming that no single variable dominates hypertension risk. Mutual information scores shown in Figure 4 indicated that age contributed the most standalone information toward hypertension (MI = 0.031), followed by waist-to-hip ratio (MI = 0.031), waist-to-height ratio (MI = 0.026), waist circumference (MI = 0.025), and BMI (MI = 0.022). Behavioural variables such as smoking (MI = 0.015), salt intake (MI = 0.011), and cholesterol (MI = 0.008) contributed modestly, while physical activity, sex, education, and residence contributed minimal information (MI < 0.002). Machine Learning Model Performance Table 4 summarizes model performance across all five classifiers. Logistic Regression and Linear SVM achieved the highest ROC AUC (both 0.74) and the highest recall (0.70 and 0.69, respectively), indicating superior identification of hypertensive individuals. Their F1 scores (both 0.51) reflected a reasonable balance between sensitivity and precision. Tree-based models (Random Forest, Gradient Boosting, and XGBoost) achieved higher overall accuracy (0.78–0.79) but substantially lower recall (0.22–0.26), indicating that these models favoured the majority normotensive class. Table 4 . Machine Learning Model Performance for Predicting High Blood Pressure Using the Zambia WHO STEPS Dataset Model ROC AUC PR AUC Accuracy Precision Recall F1 Score Linear SVM 0.74 0.50 0.69 0.41 0.69 0.51 Logistic Regression 0.74 0.49 0.69 0.40 0.70 0.51 Random Forest 0.73 0.49 0.79 0.64 0.22 0.32 Gradient Boosting 0.70 0.47 0.78 0.60 0.23 0.33 XGBoost 0.69 0.46 0.78 0.57 0.26 0.36 Performance metrics evaluated on a held-out test set. Bold values indicate the best-performing model for each metric. ROC AUC – area under the receiver operating characteristic curve; PR AUC – area under the precision-recall curve; F1 – harmonic mean of precision and recall; SVM – support vector machine. The ROC curves in Figure 5 show moderate and closely aligned discrimination, with Logistic Regression and Linear SVM achieving the highest ROC AUC values (about 0.74), followed by Random Forest (about 0.73). These patterns indicate that traditional and machine-learning models achieved similar discriminatory performance, with no substantial advantage from increased model complexity in this exploratory analysis. Feature Importance The SHAP analysis in Figure 6 from the best performing logistic regression model identifies age as the strongest predictor of high blood pressure among Zambian adults aged 18–69 years in the WHO STEPS Survey, with risk increasing steadily with advancing age. Measures of adiposity, particularly waist circumference and body mass index (BMI), also show strong positive associations, highlighting the importance of central and overall obesity in hypertension risk. Waist-to-height ratio demonstrates a more complex, potentially non-linear relationship, while civil or marital status and sex contribute moderate effects. Fasting blood glucose shows strong positive effects among individuals with very high levels, consistent with the link between dysglycaemia and hypertension, whereas total cholesterol shows some inverse associations in specific subgroups. Other factors, including salt intake, physical activity, diet, education, and smoking, contribute minimally after accounting for age and obesity measures. Overall, age and key adiposity, particularly waist circumference, indicators emerge as the strongest predictors, supporting their use in screening for high blood pressure in this population. Discussion Central obesity markers identified high blood pressure more accurately than BMI in our study. In women, waist-to-height ratio and waist circumference outperformed BMI, while waist-to-hip ratio performed poorly in men. The optimal cut-offs for all measures fell below WHO thresholds. BMI showed moderate sensitivity and specificity, while waist-to-height ratio and waist circumference performed similarly, and waist-to-hip ratio remained less consistent. The models showed moderate discrimination, with random forest achieving the highest accuracy but lower recall, while logistic regression and linear SVM achieved better recall and overall balance, and age ranked as the strongest predictor followed by BMI, waist circumference, waist-to-height ratio, and cholesterol. Anthropometric Differences and the Role of Central Obesity in High Blood Pressure Individuals with high blood pressure in this study demonstrated significantly higher body mass index (BMI), waist circumference (WC), waist-to-height ratio (WHtR), and waist-to-hip ratio (WHpR) across both sexes and all blood pressure categories. This pattern aligns with findings reported in various global regions. Central obesity indicators, particularly WC and WHtR, exhibited greater discriminatory power for high blood pressure status compared to BMI. This distinction was especially evident among women, as the area under the curve (AUC) values for WHtR and WC approached 0.66, whereas BMI reached 0.64. These findings are consistent with studies conducted in rural India, Nigeria, and Ethiopia, which also identified strong associations between both BMI and central adiposity measures and elevated blood pressure. 33–36 . Similarly, research from China and Korea demonstrates that central obesity indices (WC, WHR, WHtR) are superior to BMI in predicting hypertension 33–36 . The biological credibility of these associations is well established, with mechanisms including insulin resistance, sodium retention, sympathetic nervous system activation, and vascular dysfunction contributing to the link between adiposity and hypertension 33,37,38 . Discriminatory Performance and Optimal Cutoff Values of Obesity Indices Our results reinforce that BMI, waist circumference (WC), waist-to-hip ratio (WHpR), and waist-to-height ratio (WHtR) can all help identify high blood pressure in Zambian adults. However, our ROC analysis shows that WHtR and WC were more accurate than BMI, while WHpR was less reliable, especially for men (AUC = 0.59). This trend, where central obesity measures outperform BMI,has also been seen in studies from Malaysia, China, and Vietnam, where WHtR and WC offered higher AUC values and were better at predicting hypertension, particularly for women. 34,39–41 . Notably, the optimal cut-off values for these indices in our Zambian sample were generally lower than the WHO-recommended thresholds, a trend also observed in Ethiopia, several Asian populations, and among Albanians, where BMI and WC cut-offs for hypertension risk were consistently below international standards 33,42–44 . For example, our study identified BMI cut-offs of 21.97 kg/m² for men and 24.12 kg/m² for women, compared to the WHO’s 25 kg/m², and WC cut-offs of 80.35 cm for men and 81.85 cm for women, both lower than the WHO’s 94 cm and 80 cm, respectively. Similar lower thresholds have been reported in Ethiopia, Hong Kong, Shandong (China), and Taiwan, particularly among men, while studies in Albania and Korea also found optimal BMI and WC cut-offs below WHO recommendations 12,33,35,42,43,45,46 . Sensitivity and Specificity of Index Cutoffs and Implications for Population Screening The performance of WHpR in our study, with cut-offs of 0.86 for men and 0.84 for women, was also lower than the WHO definition of abdominal obesity (≥0.90 for men and ≥0.85 for women), mirroring findings from Ethiopia and Hong Kong, though sensitivity estimates were generally lower than those observed in Chinese populations 33,44,45 . In contrast, WHtR demonstrated the most consistent performance, with cut-off values of 0.48 for men and 0.53 for women, closely aligning with findings from Hong Kong, Taiwan, Shandong (China), Western Ethiopia, and Vietnam, where optimal WHtR thresholds ranged between 0.47 and 0.55 33,39,45 . While specificity estimates for WHtR were comparable across studies, sensitivity remained modest in both sexes, underscoring the need for population-specific thresholds. Population and Sex Differences in Optimal Anthropometric Thresholds Comparative studies across regions highlight the variability in optimal anthropometric cut-offs, often influenced by population structure, urbanization, and ethnic differences in body composition. For instance, urban Ethiopian cohorts reported higher BMI cut-offs than rural populations, and studies in Nigeria and Iran found higher WC cut-offs for women than men, diverging from WHO standards 12,33,47 . In South Africa and Cambodia, WC and WHtR were better predictors of hypertension in women, while BMI and WC were more appropriate for men, emphasizing the importance of sex-specific and regionally tailored screening tools 37,46 . Furthermore, meta-analyses and large-scale studies in China, Korea, and the US confirm that central obesity measures (WC, WHtR, WHR) are generally superior to BMI for hypertension prediction, though the magnitude of association and optimal thresholds vary by ethnicity, age, and sex 34,48–51 . Machine Learning–Based Prediction of Hypertension Using Obesity Indices Our study also evaluated the predictive performance of several machine learning models, including Logistic Regression, Linear SVM, Random Forest, and Gradient Boosting, for hypertension classification using anthropometric indices and additional demographic and biochemical factors. Consistent with prior research, our results showed that Logistic Regression and Linear SVM achieved the highest discriminatory performance (ROC AUC = 0.74), closely followed by Random Forest (AUC = 0.73) and Gradient Boosting (AUC = 0.72). Ensemble models such as Random Forest demonstrated higher overall accuracy (up to 79.0%) but lower recall (0.22–0.26), reflecting challenges with class imbalance, while Logistic Regression and Linear SVM provided more balanced recall (0.68–0.69) and F1 scores (0.51), indicating better sensitivity to hypertensive cases. Feature importance analysis confirmed that age was the dominant predictor, followed by BMI, waist circumference, waist-to-height ratio, and cholesterol, supporting the continued relevance of these indices in multivariable models 19,52 . Comparative Performance of Traditional and Machine Learning Models for Prediction Our findings align with previous studies that have compared traditional and machine learning approaches for hypertension prediction. For example, ensemble models such as Random Forest and XGBoost have been shown to achieve high accuracy and AUC in various populations, with feature importance analyses consistently highlighting age, BMI, waist circumference, and related anthropometric measures as key predictors 53,54 . However, as in our study, no single obesity index or model consistently outperformed others across all metrics or populations. The modest improvements observed when integrating multiple indices and demographic variables into predictive models underscore the value of combining simple, non-invasive measures for population-level risk stratification, as also demonstrated in large-scale studies using WHO STEPS and similar datasets 12,19 . Clinical Implications A key clinical implication for hypertension screening in Zambia is the need to use locally derived anthropometric cut-off values, rather than relying solely on WHO thresholds, for measures such as waist circumference (WC), waist-to-height ratio (WHtR), and body mass index (BMI). Evidence from Zambia and neighboring countries shows that standard WHO cut-offs may not accurately identify all individuals at risk, potentially leading to underdiagnosis and missed opportunities for early intervention. Adopting population-specific thresholds, informed by local data, can improve the sensitivity and specificity of screening programs, ensuring that more at-risk individuals are identified and managed appropriate 55,56 . Secondly, integrating simple, validated predictive models that combine age and anthropometric indicators into primary health care is especially important in resource-limited settings. These models, when embedded within existing primary care infrastructure and supported by community health workers, can expand the reach of hypertension screening, improve case detection, and facilitate timely linkage to care. Community-based screening, particularly when integrated with other chronic disease services, has been shown to be both effective and cost-efficient, leading to significant reductions in cardiovascular morbidity and mortality across Africa 57,58 . Sex differences in hypertension risk prediction also permit attention. Central obesity measures like WC and WHtR are particularly strong predictors of hypertension in women, suggesting that sex-specific cut-off values or screening protocols may further enhance detection rates 56,59 . These recommendations are urgent given Zambia’s rising burden of non-communicable diseases, driven by urbanization and lifestyle changes, and the persistently low rates of hypertension awareness, treatment, and control, especially in rural and underserved communities. Addressing these gaps requires expanding universal screening, adopting locally relevant anthropometric thresholds, integrating risk prediction tools into primary care, and ensuring equitable access to follow-up and treatment 56,60 . Public Health Relevance and Future Directions for Hypertension Risk Prediction Importantly, our models were not tuned, as this was an exploratory analysis, yet their performance was comparable to or slightly better than those reported in earlier studies using similar datasets and methods 19,53,54 . This suggests that even untuned, interpretable models can provide valuable insights for public health screening and risk stratification, particularly in resource-limited settings. Future research should focus on model tuning, external validation, and the development of country- or region-specific models to further enhance predictive accuracy and generalizability. Additionally, the integration of machine learning-based prediction tools into health systems could support targeted interventions and more efficient allocation of resources for hypertension prevention and control. Strengths and Limitations Our study has several limitations. Its cross-sectional design precludes causal inference between anthropometric indicators and hypertension. Blood pressure and behavioural variables were measured at a single point in time, which may not fully capture long-term exposure or variability. Although the WHO STEPS methodology ensures standardized measurements, some variables relied on self-reported information and may be subject to recall or social desirability bias. In addition, the machine-learning models were not extensively tuned or externally validated, and their predictive performance should therefore be interpreted as exploratory. Despite these limitations, the large, nationally representative sample and standardized measurements strengthen the reliability and public-health relevance of the findings. Conclusion Central obesity indices, particularly WHtR and WC, outperformed BMI in discriminating high blood pressure among Zambian adults, with optimal cutoff values consistently lower than WHO recommended thresholds. Simple, interpretable models combining age and anthropometric indices achieved discriminatory performance comparable to more complex machine learning algorithms. These findings support the adoption of locally derived, sex specific anthropometric cutoffs for hypertension screening in Zambia and the integration of simple prediction tools within primary care and community health programmes to strengthen early detection of undiagnosed hypertension. Abbreviations AUC Area Under the Curve BMI Body Mass Index CI Confidence Interval LMICs Low- and Middle-Income Countries NCDs Non-Communicable Diseases OR Odds Ratio ROC Receiver Operating Characteristic SSA Sub-Saharan Africa STEPS STEPwise Approach to NCD Risk Factor Surveillance WC Waist Circumference WHpR Waist-to-Hip Ratio WHtR Waist-to-Height Ratio WHO World Health Organization Declarations Acknowledgements The authors thank the World Health Organization and the Ministry of Health, Zambia for providing access to the 2017 Zambia WHO STEPwise Approach to Surveillance (STEPS) dataset for research purposes. The authors also acknowledge the survey teams and all study participants for their valuable contributions to the collection of this nationally representative dataset. We sincerely acknowledge the US–Zambia NCD Risk Project for its support and collaboration. We thank mentors at The George Washington University Milken Institute School of Public Health, including Dr. Heather Rosen and Dr. Nino Paichadze, as well as colleagues at The University of Zambia, including Dr. Cosmas Zyambo, Prof. Choolwe Jacobs, and Dr. Adam Silumbwe, for their guidance and mentorship throughout this study. The authors confirm that these institutions had no role in the design, analysis, interpretation, or writing of this manuscript. Any errors or omissions remain the responsibility of the authors Author Contributions SM led the conceptualisation of the study, conducted the statistical analysis, and drafted the manuscript. CM, GM, WM and PM contributed to study conceptualisation, interpretation of the results, and critically reviewed the analysis and manuscript. All authors reviewed and approved the final version of the manuscript. Funding This study received no specific funding. Data Availability The data analyzed in our study are derived from the 2017 Zambia WHO STEPwise Approach to NCD Risk Factor Surveillance (STEPS) survey and are publicly available through the World Health Organization NCD Microdata Repository. Access to the data is subject to World Health Organization data access approval procedures. Ethics Approval and Consent to Participate The 2017 Zambia WHO STEPwise Approach to Surveillance (STEPS) survey was conducted in accordance with the ethical principles outlined in the Declaration of Helsinki. Ethical approval for the original survey was obtained from the relevant national ethics authorities in Zambia, with technical oversight from the World Health Organization. This study was a secondary analysis of de-identified, publicly available STEPS data accessed through an approved World Health Organization data request. Ethical clearance for the secondary analysis was obtained from the University of Zambia Biomedical Research Ethics Committee (UNZABREC) with approval No. REF. 6-492-2025 and registered with the National Health Research Authority (NHRA) with registration number NHRA-2135/15/04/2025. The analysis involved no direct contact with participants and utilized anonymized data. Consent for Publication Not applicable. This manuscript does not contain any individual-level identifiable data. Competing Interests The authors declare no competing interest s References Blüher, M. Obesity: global epidemiology and pathogenesis. Nat. Rev. Endocrinol. 15 , 288–298 (2019). Hruby, A. & Hu, F. B. The Epidemiology of Obesity: A Big Picture. PharmacoEconomics 33 , 673–689 (2015). Mills, K. T., Stefanescu, A. & He, J. The global epidemiology of hypertension. Nat. Rev. Nephrol. 16 , 223–237 (2020). Attia, A. et al. Abstract TH122: Prevalence and Associated Risk Factors of Hypertension in Urban and Rural Communities of Sub-Saharan Africa. Hypertension 82 , hyp.82.suppl_1.TH122 (2025). Guwatudde, D. et al. The burden of hypertension in sub-Saharan Africa: a four-country cross sectional study. BMC Public Health 15 , 1211 (2015). Nwankwo, M. et al. Hypertension in West Africa: a systematic review and meta-analysis of prevalence and associated risk factors. BMC Public Health 25 , 3269 (2025). Hendriks, M. E. et al. Hypertension in Sub-Saharan Africa: Cross-Sectional Surveys in Four Rural and Urban Communities. PLoS ONE 7 , e32638 (2012). Price, A. J. et al. Prevalence of obesity, hypertension, and diabetes, and cascade of care in sub-Saharan Africa: a cross-sectional, population-based study in rural and urban Malawi. Lancet Diabetes Endocrinol. 6 , 208–222 (2018). Schutte, A. E. Hypertension in Low- and Middle-Income Countries. HHS Pub Lic Access Physiol. Behav. 176 , 139–148 (2021). Pengpid, S. & Peltzer, K. Prevalence and correlates of multiple non-communicable disease risk factors among adults in zambia: Results of the first national steps survey in 2017. Pan Afr. Med. J. 37 , 1–12 (2020). Haregu, T. N., Oti, S., Egondi, T. & Kyobutungi, C. Measurement of overweight and obesity an urban slum setting in sub-Saharan Africa: a comparison of four anthropometric indices. BMC Obes. 3 , 46 (2016). Patel, S. A. et al. Obesity and its Relation With Diabetes and Hypertension. Glob. Heart 11 , 71-79.e4 (2016). Zhou, B. et al. Worldwide trends in hypertension prevalence and progress in treatment and control from 1990 to 2019: a pooled analysis of 1201 population-representative studies with 104 million participants. The Lancet 398 , 957–980 (2021). Akpa, O. M. et al. Regional Patterns and Association Between Obesity and Hypertension in Africa: Evidence From the H3Africa CHAIR Study. Hypertension 75 , 1167–1178 (2020). Haregu, T. N., Oti, S., Egondi, T. & Kyobutungi, C. Measurement of overweight and obesity an urban slum setting in sub-Saharan Africa: a comparison of four anthropometric indices. BMC Obes. 3 , 46 (2016). Patel, S. A. et al. Obesity and its Relation With Diabetes and Hypertension: A Cross-Sectional Study Across 4 Geographical Regions. Glob. Heart 11 , 71 (2016). Jovel, J. & Greiner, R. An Introduction to Machine Learning Approaches for Biomedical Research. Front. Med. 8 , 771607 (2021). Sidey-Gibbons, J. A. M. & Sidey-Gibbons, C. J. Machine learning in medicine: a practical introduction. BMC Med. Res. Methodol. 19 , 64 (2019). Latifa, A. A., Id, L. S. A., Al, M. F., Id, A. & Id, M. E. A. Predicting hypertension using machine learning : Findings from Qatar Biobank Study. 1–17 (2020) doi:10.1371/journal.pone.0240370. Lopez-Jimenez, F. et al. Obesity and cardiovascular disease: mechanistic insights and management strategies. A joint position paper by the World Heart Federation and World Obesity Federation. Eur. J. Prev. Cardiol. 29 , 2218–2237 (2022). Zhao, H. et al. Predicting the Risk of Hypertension Based on Several Easy-to-Collect Risk Factors: A Machine Learning Method. Front. Public Health 9 , 619429 (2021). Islam, Md. M. et al. Predicting the risk of hypertension using machine learning algorithms: A cross sectional study in Ethiopia. PLOS ONE 18 , e0289613 (2023). Septian, E. et al. Prediction of Personalised Hypertension Using Machine Learning in Indonesian Population. J. Med. Syst. 49 , 137 (2025). Sifat, I. K. & Kibria, Md. K. Optimizing hypertension prediction using ensemble learning approaches. PLOS ONE 19 , e0315865 (2024). Bisong, E. et al. Predicting high blood pressure using machine learning models in low- and middle-income countries. BMC Med. Inform. Decis. Mak. 24 , 234 (2024). Bujang, M. A. Requirements for Minimum Sample Size for Sensitivity and Specificity Analysis. J. Clin. Diagn. Res. https://doi.org/10.7860/JCDR/2016/18129.8744 (2016) doi:10.7860/JCDR/2016/18129.8744. Egan, B. M. & Zhao, Y. Different Definitions of Prevalent Hypertension Impact: The Clinical Epidemiology of Hypertension and Attainment of Healthy People Goals. J. Clin. Hypertens. 15 , 154–161 (2013). Haibo He & Garcia, E. A. Learning from Imbalanced Data. IEEE Trans. Knowl. Data Eng. 21 , 1263–1284 (2009). StataCorp LLC. Stata Statistical Software: Release 17 . (StataCorp LLC, College Station, TX, USA, 2021). Python Software Foundation. Python: A Programming Language. (2023). Pedregosa, F. et al. Scikit-learn: Machine learning in Python. J. Mach. Learn. Res. 12 , 2825–2830 (2011). Chen, T. & Guestrin, C. Xgboost: A scalable tree boosting system. in Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining 785–794 (2016). Gutema, B. T. et al. Predictive capacity of obesity indices for high blood pressure among southern Ethiopian adult population: A WHO STEPS survey. BMC Cardiovasc. Disord. 20 , 1–9 (2020). Lee, X., Gao, Y., Zhang, Y. & Feng, Y. Comparison of obesity-related indices for predicting hypertension based on ROC analysis in Chinese adults. Front. Public Health https://doi.org/doi:%2010.3389/fpubh.2022.1042236 (2022) doi:doi: 10.3389/fpubh.2022.1042236. Lee, J., Lim, N., Baek, T., Park, S. & Park, H. Anthropometric indices as predictors of hypertension among men and women aged 40 – 69 years in the Korean population : the Korean Genome and Epidemiology Study. 1–7 (2015) doi:10.1186/s12889-015-1471-5. Adegoke, O., Ozoh, O. B., Odeniyi, I. A., Bello, B. T. & Okubadejo, N. U. Prevalence of obesity and an interrogation of the correlation between anthropometric indices and blood pressures in urban. Sci. Rep. 1–12 (2021) doi:10.1038/s41598-021-83055-w. Modjadji, P., Salane, M. C., Mokwena, K. E., Mudau, T. S. & Mphekgwana, P. M. Utility of Obesity Indicators for Predicting Hypertension among Older Persons in Limpopo Province, South Africa. Appl. Sci. Switz. 12 , (2022). Chen, L. et al. Association of different obesity patterns with hypertension in US male adults : a cross ‑ sectional study. 1–10 (2023) doi:10.1038/s41598-023-37302-x. Zhang, Y. et al. Association between anthropometric indicators of obesity and cardiovascular risk factors among adults in Shanghai , China. 1–9 (2019). Quan, N. M. & Nguyen, M. H. Anthropometric Indexes for Predicting High Blood Pressure in Vietnamese Adults : A Cross-Sectional Study. Dovepress 181–186 (2020). Ying, J., Tee, H., Gan, W. Y. & Lim, P. Y. Comparisons of body mass index , waist a body shape index ( ABSI ) in predicting high blood pressure among Malaysian adolescents : a cross- ­ sectional study. 1–9 (2020) doi:10.1136/bmjopen-2019-032874. Gupta, R. Das & Hashan, M. R. Association Between Abdominal Obesity , Body Mass Index , and Hypertension in India : Evidence From a Large Nationally Representative Data. J. Clin. Hypertens. https://doi.org/10.1111/jch.70034 (2025) doi:10.1111/jch.70034. Hu, J., Zhong, Y., Ge, W., Lv, H. & Ding, Z. Comparisons of tri-ponderal mass index and body mass index in discriminating hypertension at three separate visits in adolescents : A retrospective cohort study. 1–11 (2022) doi:10.3389/fnut.2022.1028861. Kawamoto, R., Kikuchi, A., Akase, T., Ninomiya, D. & Kumagi, T. Usefulness of waist-to-height ratio in screening incident hypertension among Japanese community-dwelling middle-aged and elderly individuals. Clin. Hypertens. 26 , 1–9 (2020). Islam, M. R., Md, M. & Samaha, M. S. Relationship of Anthropometric Indicators of General and Abdominal Obesity with Hypertension and Their Predictive Performance among Albanians : A Nationwide. MDPI Nutr. (2021). Shimotake, Y. et al. Heliyon Comparative evaluation of anthropometric measurements and prevalence of hypertension : community based cross-sectional study in rural male and female Cambodians. Heliyon 6 , e04432 (2020). Adegoke, O. et al. Prevalence of obesity and an interrogation of the correlation between anthropometric indices and blood pressures in urban Lagos, Nigeria. Sci. Rep. 11 , 1–12 (2021). Gui, J. et al. Obesity ‑ and lipid ‑ related indices as a risk factor of hypertension in mid ‑ aged and elderly Chinese : a cross ‑ sectional study. BMC Geriatr. 1–13 (2024) doi:10.1186/s12877-023-04650-2. Jiang, X. et al. Gender differences in associations between obesity and hypertension , diabetes , dyslipidemia : evidence from electronic health records of 3 . 5 million Chinese senior population. (2025). Chao, Wanga. ;, Fub., W. & Caoa, S. Association of Adiposity Indicators with Hypertension among Chinese Adults Running title : Obesity indicators and hypertension Chao Wang. Sciencedirect (2021). Ge, Q. et al. Comparison of different obesity indices related with hypertension among different sex and age groups in China. Nutr. Metab. Cardiovasc. Dis. 313 793-801 31(3), 793 , (2021). Sifat, I. K. & Kaderi, K. Optimizing hypertension prediction using ensemble learning approaches. 1–17 (2024) doi:10.1371/journal.pone.0315865. Silva, G. F. S., Fagundes, T. P. & Teixeira, B. C. Machine Learning for Hypertension Prediction: a Systematic Review. https://link.springer.com/ (2022). Ahmed, F., Ali, S., Rahman, J. & Roy, D. C. Predicting the risk of hypertension using machine learning algorithms : A cross sectional study in Ethiopia. 1–20 (2023) doi:10.1371/journal.pone.0289613. Shah, P. et al. Abstract P1057: Hypertension Screening in Children and Adolescents: A Global Policy Analysis. Circulation 151 , cir.151.suppl_1.P1057 (2025). Tateyama, Y. et al. Hypertension, its correlates and differences in access to healthcare services by gender among rural Zambian residents: A cross-sectional study. BMJ Open 12 , (2022). Hickey, M. D. et al. Cost-effectiveness of leveraging existing HIV primary health systems and community health workers for hypertension screening and treatment in Africa: An individual-based modeling study. PLOS Med. 22 , e1004531 (2025). Mutelo, L. et al. Prevalence and correlates of microalbuminuria in high-risk persons with hypertension, diabetes, or HIV at a tertiary hospital in Zambia. PLOS One 20 , e0328529 (2025). Kim, J.-H. et al. Examining geospatial and temporal distribution of invasive non-typhoidal Salmonella disease occurrence in sub-Saharan Africa: a systematic review and modelling study. BMJ Open 14 , e080501 (2024). Lucinde, R. K. et al. Diagnostic Performance of Unattended Automated Office Blood Pressure Measurement for Hypertension Screening Among People With and Without HIV. J. Am. Heart Assoc. 14 , e043957 (2025). Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9585939","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":632953881,"identity":"9943883e-ff04-427e-a26e-10a9c21a68c9","order_by":0,"name":"Samuel Mutasha","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABBElEQVRIiWNgGAWjYNCCAgjFzMBgkwBmJRQQ0mIAxGxgLWkJIAZDggHxWg5DtDDg0SLffjpN4oMBQ7TB/ebHnwtqzufxy3cnfnhgwCDPL3YAu/lncrdJzjBgyN1wjM1Mesax28WSbbybJYAOM5w5OwGHk3K3SfOAtTCYMfOw3U7ccIx3A0hLgsFt7Frk+99uk/4D1sL++TPPv3MgLZt/4NPCcANoCwNYC4+BNG/bAZCWbXhtMbjxdrNlD1DLzGM5ZdK8fcmJM9tyt1kkGEjg9It8f+7GGz8qGHL7Dh/f/Jnnm11iP/PZzTd/VNjI80vjcBgE/McQkcCnfBSMglEwCkYBAQAA+TleeerS74QAAAAASUVORK5CYII=","orcid":"https://orcid.org/0009-0009-5480-5553","institution":"UNZA: University of Zambia","correspondingAuthor":true,"prefix":"","firstName":"Samuel","middleName":"","lastName":"Mutasha","suffix":""},{"id":632958115,"identity":"c74b8aa4-6b57-4e99-9f9c-280076e9bccd","order_by":1,"name":"Given Moonga","email":"","orcid":"https://orcid.org/0000-0003-0062-1475","institution":"UNZA: University of Zambia","correspondingAuthor":false,"prefix":"","firstName":"Given","middleName":"","lastName":"Moonga","suffix":""},{"id":632958116,"identity":"c7836c78-225e-4ace-89c8-59e6a5453f5e","order_by":2,"name":"Wilbroad Mutale","email":"","orcid":"https://orcid.org/0000-0002-4891-6750","institution":"UNZA: University of Zambia","correspondingAuthor":false,"prefix":"","firstName":"Wilbroad","middleName":"","lastName":"Mutale","suffix":""},{"id":632958117,"identity":"f43aa5e1-f0e9-4e59-801b-0bd63f10313b","order_by":3,"name":"Clyde Mulenga","email":"","orcid":"https://orcid.org/0000-0003-0037-1365","institution":"UNZA: University of Zambia","correspondingAuthor":false,"prefix":"","firstName":"Clyde","middleName":"","lastName":"Mulenga","suffix":""},{"id":632958118,"identity":"23d4d116-d581-4b21-82d1-187139dc8d66","order_by":4,"name":"Patrick Musonda","email":"","orcid":"https://orcid.org/0000-0002-1204-6300","institution":"UNZA: University of Zambia","correspondingAuthor":false,"prefix":"","firstName":"Patrick","middleName":"","lastName":"Musonda","suffix":""}],"badges":[],"createdAt":"2026-05-01 12:21:29","currentVersionCode":1,"declarations":{"humanSubjects":true,"vertebrateSubjects":true,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":true,"humanSubjectConsent":true,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":true},"doi":"10.21203/rs.3.rs-9585939/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9585939/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108417409,"identity":"5c3e6c0a-1d9b-476a-b1f4-beeccc77d869","added_by":"auto","created_at":"2026-05-04 11:45:12","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":64670,"visible":true,"origin":"","legend":"\u003cp\u003eParticipant Flow Diagram\u003c/p\u003e","description":"","filename":"Figure1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9585939/v1/0758fc3f1dfa289574c2e45c.jpg"},{"id":108417407,"identity":"9e4e130a-ec5c-43c2-9ebc-4eafd8ef585c","added_by":"auto","created_at":"2026-05-04 11:45:12","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":128198,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eROC curve for the assessment of the risk of high-blood pressure on anthropometric indices by sex\u003c/em\u003e\u003c/p\u003e","description":"","filename":"Figure2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9585939/v1/2741b56918bd9f84e2db3964.jpg"},{"id":108493408,"identity":"8eda3045-8f0a-49ab-9d09-d87f155ecdf6","added_by":"auto","created_at":"2026-05-05 10:00:18","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":32861,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eFeature correlation matrix\u003c/em\u003e\u003c/p\u003e","description":"","filename":"Figure3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9585939/v1/5dfe3e2945bd2440e0e09bf4.jpg"},{"id":108493879,"identity":"08971634-411c-48e9-bbed-0768845afcb6","added_by":"auto","created_at":"2026-05-05 10:02:00","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":97225,"visible":true,"origin":"","legend":"\u003cp\u003eMutual Information (MI) Score\u003c/p\u003e","description":"","filename":"Figure4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9585939/v1/db7bd972b7d282f7eef2a02b.jpg"},{"id":108492665,"identity":"74bec0cf-981c-4932-b08c-c3b8ccb1649d","added_by":"auto","created_at":"2026-05-05 09:58:16","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":51253,"visible":true,"origin":"","legend":"\u003cp\u003eReceiver Operating Characteristic (ROC) and Precision Re-call curves comparing machine-learning models for hypertension prediction\u003c/p\u003e","description":"","filename":"Figure5.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9585939/v1/2fa7e9fb883668aa5750865c.jpg"},{"id":108417412,"identity":"502a856e-2c26-4e96-91fc-12118ad73202","added_by":"auto","created_at":"2026-05-04 11:45:12","extension":"jpg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":111647,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP summary (Logistic Regression)\u003c/p\u003e","description":"","filename":"Figure6.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9585939/v1/10c8f295bdb6e161a8aa1145.jpg"},{"id":108804682,"identity":"5e71d85a-8c29-4fbf-9e3c-8aefb1aa6314","added_by":"auto","created_at":"2026-05-08 15:22:46","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1082630,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9585939/v1/9e88a538-8cea-4824-983e-a1eb4479fd06.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eEvaluating the Predictive Capacity of Obesity Indices for High Blood Pressure Among Zambians Aged 18–69 Years Using Machine Learning: Evidence From WHO STEPS Survey\u003c/p\u003e","fulltext":[{"header":"Introduction","content":"\u003cp\u003eHypertension and obesity are now among the most pressing global health challenges, with their prevalence rising sharply in both high- and low-income countries. Worldwide, over 1.28 billion adults are estimated to have hypertension, and more than 1 billion are living with obesity, a figure that has more than doubled since 1990, with the fastest increases seen in low- and middle-income regions, including sub-Saharan Africa\u0026nbsp;\u003csup\u003e1\u0026ndash;3\u003c/sup\u003e.\u0026nbsp;In sub-Saharan Africa, hypertension prevalence ranges from 20% to over 40% in adults, with urbanization, dietary shifts, and reduced physical activity increasing a parallel surge in obesity\u003csup\u003e4\u0026ndash;6\u003c/sup\u003e. Zambia mirrors these trends with recent national surveys indicating that about one in three adults has raised blood pressure, and overweight rates are rising, especially among women and urban residents\u0026nbsp;\u003csup\u003e7,8\u003c/sup\u003e.\u0026nbsp;Despite this, many cases remain undiagnosed or poorly managed, and the health system faces challenges in early detection and prevention.\u003c/p\u003e\n\u003cp\u003eIn Africa, the increasing prevalence of hypertension and obesity is driven by poor dietary habits, physical inactivity, smoking, alcohol consumption, and environmental factors such as urbanization and pollution\u0026nbsp;\u003csup\u003e9\u003c/sup\u003e. Zambia demonstrates this trend, with an 18.9% national hypertension prevalence, rising to 34.8% in urban areas like Lusaka\u0026nbsp;\u003csup\u003e10\u003c/sup\u003e. Obesity affects 24.4% of Zambian adults, with a higher prevalence among women\u003csup\u003e10\u003c/sup\u003e. Given the link between obesity and hypertension, assessing anthropometric indices is key for improving early detection and prevention strategies. However, the optimal anthropometric indicators and cut-off values for predicting high blood pressure may vary across populations, underscoring the need for population-specific evaluation.\u003c/p\u003e\n\u003cp\u003eAnthropometric indices such as body mass index (BMI), waist circumference (WC), waist-to-hip ratio (WHR), and waist-to-height ratio (WHtR) are widely used to assess obesity and predict hypertension risk\u003csup\u003e11\u0026ndash;13\u003c/sup\u003e . The World Health Organization (WHO) recommends BMI cut-offs of 25 kg/m\u0026sup2; for overweight and 30 kg/m\u0026sup2; for obesity, and WC thresholds of \u0026ge;94 cm for men and \u0026ge;80 cm for women (Europid values)\u0026nbsp;\u003csup\u003e12,13,13\u003c/sup\u003e.\u0026nbsp;However, evidence suggests these cut-offs may not optimally capture risk in African populations, where body composition and fat distribution differ from those in Western populations\u003csup\u003e14\u0026ndash;16\u003c/sup\u003e . \u0026nbsp;For example, at the same BMI, Africans may have higher cardiovascular risk, and indices like WC or WHtR may outperform BMI in predicting hypertension\u0026nbsp;\u003csup\u003e11\u0026ndash;13\u003c/sup\u003e.\u0026nbsp;This raises a key question: should Zambia adopt cut-off values that are validated within its own population?\u003c/p\u003e\n\u003cp\u003eRecent advances in machine learning (ML) provide a powerful approach to handle complex biomedical and healthcare prediction problems, enabling the extraction of patterns from large, high‑dimensional data and often improving classification and prediction beyond traditional statistical methods\u003csup\u003e17,18\u003c/sup\u003e. Machine learning models, including random forests, XGBoost, and neural networks, can analyze large, complex datasets to identify the most predictive anthropometric indices and optimal cut-off points for hypertension risk, tailored to specific populations\u003csup\u003e19\u0026ndash;21\u003c/sup\u003e. Studies from Africa and other low-resource settings have shown that ML-based models can outperform traditional regression in predicting hypertension, especially when using easy-to-collect, non-invasive measures like age, BMI, WC, and WHtR\u0026nbsp;\u003csup\u003e22,23\u003c/sup\u003e.\u0026nbsp;For instance, Islam and colleagues in Albania demonstrated that ML models achieved high accuracy in hypertension prediction, and identified age, BMI, and WC as top predictors\u003csup\u003e22,24\u003c/sup\u003e. Similarly, multi-country analyses using WHO STEPS data have highlighted the need for country-specific models and thresholds to maximize predictive performance\u0026nbsp;\u003csup\u003e25\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eLeveraging on ML with WHO STEPS survey data presents an opportunity to compare the predictive value of different anthropometric indices, explore whether local cut-offs improve risk stratification, and ultimately inform more effective, context-specific screening and prevention strategies in Zambia. This approach could help close the gap in early detection, reduce the burden of undiagnosed hypertension, and support the development of tailored interventions for Zambia\u0026rsquo;s diverse adult population.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003eStudy Design and Data Source\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis cross-sectional study used secondary data from the 2017 Zambia WHO STEPwise Approach to NCD Risk Factor Surveillance (STEPS) survey. The survey employed a nationally representative sampling frame covering urban and rural areas across all ten provinces, targeting adults aged 18\u0026ndash;69 years. Data were collected in three steps: self-reported behavioural and sociodemographic information (Step 1), standardised physical and anthropometric measurements (Step 2), and biochemical measurements from fasting blood samples (Step 3). This study drew on data from all three steps.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSampling and Participants\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe survey used a multistage cluster sampling design. Standard Enumeration Areas were selected using probability proportional to size, households were selected systematically within each area, and one eligible adult per household was randomly chosen. Non-respondents were not replaced. Adults aged 18\u0026ndash;69 years with complete blood pressure and anthropometric measurements were eligible. Pregnant women were excluded at enrollment because physiological changes during pregnancy affect blood pressure and body composition. The stated eligibility were largely applied at enrollment, this yielded a final analytical sample of 4,302 as shown in figure 1.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSample Size justification\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll eligible participants were retained to maximize the stability of machine learning model estimates. The minimum required sample size was calculated using the recommended diagnostic test accuracy framework\u003csup\u003e26\u003c/sup\u003e. Assumptions included \u0026nbsp;95% confidence level (Z = 1.96), expected sensitivity of 80%, expected specificity of 75%, hypertension prevalence of 32%, and a maximum confidence interval width of 10%, the minimum required sample was approximately 1,300. However, the study employed a full enumeration with the available 4,302 participants.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStudy Variables\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eOutcome.\u003c/em\u003e Hypertension was defined as systolic blood pressure \u0026ge;140 mmHg, diastolic blood pressure \u0026ge;90 mmHg, or current use of antihypertensive medication\u003csup\u003e27\u003c/sup\u003e, coded as a binary variable (1 = hypertensive, 0 = normotensive).\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eAnthropometric obesity indices.\u003c/em\u003e Four indices were evaluated: Body Mass Index (weight in kg/height in m\u0026sup2;), Waist Circumference (measured at the midpoint between the lowest rib and iliac crest), Waist-to-Hip Ratio, and Waist-to-Height Ratio. All measurements followed WHO standardised procedures.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eCovariates.\u003c/em\u003e Sociodemographic variables included age, sex, residence (urban/rural), province, education level, and marital status. Behavioural variables included smoking status, alcohol use, physical activity, and salt intake frequency. Total cholesterol was included as a biochemical predictor.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHandling Class Imbalance\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eApproximately 24% of participants were hypertensive. To prevent classifiers from favouring the majority class, stratified sampling preserved the original class distribution in each cross-validation fold. Recall was prioritised as the primary metric to maximise detection of true hypertension cases\u003csup\u003e28\u003c/sup\u003e. Precision-Recall curves supplemented standard ROC curves. No synthetic oversampling was applied to preserve the original population distribution.\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAnalyses were conducted using STATA version 17 \u003csup\u003e29\u003c/sup\u003e for survey-weighted descriptive statistics. and Python \u003csup\u003e30\u003c/sup\u003e for machine learning using the popular packages scikit-learn\u003csup\u003e31\u003c/sup\u003e and XGBoost\u003csup\u003e32\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eExploratory analysis.\u003c/em\u003e Distributions of anthropometric indices were examined, a feature correlation matrix was computed, and mutual information scores were calculated to assess the standalone predictive information contributed by each variable toward the hypertension outcome.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eROC and cutoff analysis.\u003c/em\u003e Receiver Operating Characteristic curves assessed each obesity index\u0026apos;s discriminatory ability for hypertension. The Area Under the Curve (AUC) with 95% confidence intervals was computed overall and stratified by sex. Optimal cutoff values were identified using Youden\u0026apos;s Index (sensitivity + specificity \u0026minus; 1). Sensitivity-oriented thresholds were also explored for screening applications.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eClassification models.\u003c/em\u003e Five models were fitted: Logistic Regression, Linear Support Vector Machine, Random Forest, Gradient Boosting Machine, and Extreme Gradient Boosting (XGBoost). All models used default hyperparameters to maintain comparability and avoid overfitting in the absence of an external validation dataset.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eValidation.\u003c/em\u003e Internal validation used stratified 5-fold cross-validation, preserving the outcome class distribution in each fold. Performance was averaged across folds.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003ePerformance metrics.\u003c/em\u003e Models were evaluated using ROC AUC, Precision-Recall AUC, accuracy, precision, recall, and F1-score. Receiver Operators Characteristics (ROC) AUC \u0026ge;0.70 was considered acceptable. Precision Re-call (PR) AUC was interpreted relative to the baseline prevalence. Accuracy was interpreted cautiously given the class imbalance.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eFeature importance.\u003c/em\u003e Permutation-based feature importance quantified each predictor\u0026apos;s contribution by measuring the drop in model performance when that variable was randomly shuffled.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthical Considerations\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePermission to use the 2017 Zambia STEPS dataset was formally granted by the World Health Organization and the Zambia Ministry of Health. Ethical clearance for this secondary analysis was obtained from the University of Zambia Biomedical Research Ethics Committee (UNZABREC Approval No. REF. 6-492-2025). The study was additionally registered with and approved by the National Health Research Authority of Zambia (NHRA-2135/15/04/2025) in accordance with national research governance requirements. The dataset was fully anonymised prior to analysis and no personal identifiers were accessed at any stage. All analyses were conducted in accordance with the principles of the Declaration of Helsinki regarding confidentiality, data protection, and responsible use of human subject data.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003eParticipant Characteristics\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe final sample included 4,302 adults aged 18\u0026ndash;69 years, with women comprising 62.5% and men 37.5%. High blood pressure was present in 23.7% of participants (n = 1,021). Those with high blood pressure were considerably older (median 42 years, IQR 31\u0026ndash;57) compared with normotensive participants (median 32 years, IQR 24\u0026ndash;42; p \u0026lt; 0.001). Men had a higher prevalence than women (25.4% vs 22.7%; p = 0.046), urban residents higher than rural (27.0% vs 21.7%; p \u0026lt; 0.001), and alcohol users higher than non-users (26.6% vs 22.3%; p = 0.002). Neither smoking nor salt intake showed a significant association with high blood pressure. Median total cholesterol, waist circumference, and BMI were all significantly elevated among hypertensive compared with normotensive participants (all p \u0026lt; 0.001). Table 1 presents the full sample characteristics.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e1\u003c/strong\u003e\u003cstrong\u003e.\u003c/strong\u003e Population Characteristics of the Study Population Stratified by Blood Pressure Status (N = 4,302).\u003c/p\u003e\n\u003cdiv align=\"\"\u003e\n \u003ctable style=\"width: 4.5e+2pt;border: none;\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eVariables\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eOverall\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003eN = 4,302\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eNo High BP\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003en = 3,281 (76.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eHigh BP\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003en = 1,021 (23.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003ep-value\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eCholesterol (mmol/L), median (IQR)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e3.35 (2.67\u0026ndash;4.20)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e3.10 (2.59\u0026ndash;3.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e3.59 (2.75\u0026ndash;4.50)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt; 0.001ᶜ\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eWaist circumference (cm), median (IQR)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e80 (74\u0026ndash;87)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e77 (72\u0026ndash;84)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e83 (76\u0026ndash;93)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt; 0.001ᶜ\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eBody mass index (kg/m\u0026sup2;), median (IQR)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e22.9 (20.5\u0026ndash;26.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e22.0 (20.3\u0026ndash;24.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e23.8 (20.9\u0026ndash;28.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt; 0.001ᶜ\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eWaist-to-height ratio, mean (SD)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.50 (0.08)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.49 (0.07)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.53 (0.10)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt; 0.001ᵇ\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eWaist-to-hip ratio, mean (SD)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.85 (0.08)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.84 (0.08)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.86 (0.10)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt; 0.001ᵇ\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"5\"\u003e\n \u003cp\u003e\u003cstrong\u003eSex, n (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; Male\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1,614 (37.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1,204 (74.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e410 (25.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; Female\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e2,688 (62.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e2,077 (77.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e611 (22.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.046ᵃ\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eAge, median (IQR)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e32 (24\u0026ndash;42)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e32 (24\u0026ndash;42)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e42 (31\u0026ndash;57)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt; 0.001ᶜ\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"5\"\u003e\n \u003cp\u003e\u003cstrong\u003eResidence, n (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; Rural\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e2,658 (61.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e2,081 (78.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e577 (21.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; Urban\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1,644 (38.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1,200 (73.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e444 (27.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt; 0.001ᵃ\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"5\"\u003e\n \u003cp\u003e\u003cstrong\u003eEducation level, n (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; No primary\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1,546 (36.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1,159 (75.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e387 (25.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; Primary\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1,036 (24.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e791 (76.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e245 (23.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; Lower secondary\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e826 (19.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e667 (80.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e159 (19.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; Higher secondary\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e564 (13.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e438 (77.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e126 (22.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; College\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e327 (7.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e226 (69.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e101 (30.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt; 0.001ᵃ\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"5\"\u003e\n \u003cp\u003e\u003cstrong\u003eMarital status, n (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; Single\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e934 (21.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e767 (82.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e167 (17.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp;Married/cohabiting\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e2,627 (61.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e2,020 (76.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e607 (23.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; Other\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e731 (17.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e491 (67.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e240 (32.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt; 0.001ᵃ\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"5\"\u003e\n \u003cp\u003e\u003cstrong\u003eCurrent smoker, n (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; Yes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e475 (11.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e354 (74.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e121 (25.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; No\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e3,826 (89.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e2,927 (76.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e899 (23.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.339ᵃ\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"5\"\u003e\n \u003cp\u003e\u003cstrong\u003eSalt intake, n (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp;Rarely/never\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1,352 (31.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1,024 (75.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e328 (24.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; Sometimes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1,319 (30.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1,036 (78.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e283 (21.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp;Often/always\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1,618 (37.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1,211 (74.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e407 (25.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.055ᵃ\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"5\"\u003e\n \u003cp\u003e\u003cstrong\u003eAlcohol use, n (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; Yes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1,398 (32.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1,026 (73.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e372 (26.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; No\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e2,903 (67.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e2,255 (77.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e648 (22.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.002ᵃ\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003eᵃChi-square test; ᵇIndependent samples t-test; ᶜWilcoxon rank-sum test. BP \u0026ndash; blood pressure; IQR \u0026ndash; interquartile range; SD \u0026ndash; standard deviation. p-values \u0026lt; 0.05 considered statistically significant.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSex-Stratified Anthropometric Differences\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAmong participants with high blood pressure, women had higher mean BMI (26.28 \u0026plusmn; 6.28 vs 23.87 \u0026plusmn; 5.14 kg/m\u0026sup2;), waist circumference (87.29 \u0026plusmn; 16.65 vs 82.95 \u0026plusmn; 13.0 cm), and waist-to-height ratio (0.55 \u0026plusmn; 0.10 vs 0.50 \u0026plusmn; 0.08) than men. Waist-to-hip ratio was similar between sexes. The same pattern held among normotensive participants. All differences between hypertensive and normotensive groups were significant within each sex (all p \u0026lt; 0.001; Table 2).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e2\u003c/strong\u003e\u003cstrong\u003e.\u003c/strong\u003e Anthropometric Index Values According to Blood Pressure Status by Sex Among Adult Residents of Zambia.\u003c/p\u003e\n\u003cdiv align=\"\"\u003e\n \u003ctable style=\"width: 3.7e+2pt;border: none;\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eAnthropometric Index\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eHigh BP\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eHigh BP\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003ep-value\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"4\"\u003e\n \u003cp\u003e\u003cstrong\u003eMale\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp; n\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003e353\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003e1,212\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; BMI (kg/m\u0026sup2;)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e23.87 (5.14)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e22.00 (3.18)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; WstC (cm)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e82.95 (13.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e77.77 (8.24)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; WHpR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.86 (0.09)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.85 (0.06)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; WHtR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.50 (0.08)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.47 (0.05)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"4\"\u003e\n \u003cp\u003e\u003cstrong\u003eFemale\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp; n\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003e480\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003e1,960\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; BMI (kg/m\u0026sup2;)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e26.28 (6.28)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e23.46 (4.91)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; WstC (cm)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e87.29 (16.65)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e79.40 (11.10)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; WHpR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.86 (0.10)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.83 (0.08)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; WHtR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.55 (0.10)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.50 (0.07)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"4\"\u003e\n \u003cp\u003e\u003cstrong\u003eBoth sexes combined\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp; n\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003e833\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003e3,172\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; BMI (kg/m\u0026sup2;)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e25.26 (5.94)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e22.90 (4.39)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; WstC (cm)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e85.45 (15.36)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e78.80 (10.13)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; WHpR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.86 (0.10)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.84 (0.08)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp; WHtR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.53 (0.10)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.49 (0.07)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003eData presented as mean (SD). p-values derived from independent samples t-test. BMI \u0026ndash; body mass index; WstC \u0026ndash; waist circumference; WHpR \u0026ndash; waist-to-hip ratio; WHtR \u0026ndash; waist-to-height ratio; BP \u0026ndash; blood pressure; SD \u0026ndash; standard deviation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDiscriminatory Ability of Anthropometric Indices\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe ROC analysis showed that all four indices had fair discriminatory ability for high blood pressure (AUC range: 0.59\u0026ndash;0.66). Among men, waist circumference had the highest AUC (0.63, 95% CI 0.60\u0026ndash;0.67), followed by waist-to-height ratio (0.62), BMI (0.60), and waist-to-hip ratio (0.59). Among women, waist-to-height ratio and waist circumference performed best (both AUC 0.66), followed by BMI (0.64) and waist-to-hip ratio (0.60). Central adiposity measures consistently matched or outperformed BMI in both sexes (Figure 2).\u003c/p\u003e\n\u003cp\u003eOptimal cutoff values derived from Youden\u0026apos;s Index are shown in Table 3. The BMI cutoff was 21.97 kg/m\u0026sup2; for men (sensitivity 58%, specificity 58%) and 24.12 kg/m\u0026sup2; for women (sensitivity 58%, specificity 67%). For waist circumference, cutoffs were 80.35 cm (men) and 81.85 cm (women). For waist-to-height ratio, the cutoff was 0.48 for men and 0.53 for women, with the latter yielding the best combination of sensitivity (55%) and specificity (71%) among all indices in women.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 3.\u003c/strong\u003e Area Under the Curve (AUC), Optimal Cutoff Values, Sensitivity, and Specificity of Selected Anthropometric Indices for Predicting High Blood Pressure by Sex Among Adult Residents of Zambia.\u003c/p\u003e\n\u003ctable style=\"width: 110%;border: none;\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eIndices\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eSex\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eAUC (95% CI)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eCut off value\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eSensitivity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eSpecificity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eYounden Index\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eHigh blood pressure\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eBMI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.60 [0.57 - \u0026nbsp;0.64]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e21.97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.58\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.58\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.16\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.64 \u0026nbsp;[0.61 - 0.67]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e24.12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.58\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.67\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.25\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003ewstc\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.63 [0.60 - 0.67]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e80.35\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.52\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.69\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.21\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.66 [0.63 - 0.69]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e81.85\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.61\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.64\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.25\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003ewhpr\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.58 [ 0.55 - 0.62]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.86\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.52\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.61\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.13\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.60 [0.57 - 0.63]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.84\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.59\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.56\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.15\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003ewhtr\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.62 [0.59 - 0.66]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.48\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.52\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.68\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.66 [0.63 - 0.69]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.53\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.55\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.71\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.26\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eSystolic bp \u0026gt; 140 mmHg\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eBMI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.61 [0.57 - \u0026nbsp;0.65]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e22.21\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.57\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.17\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.65 [0.62 - 0.69]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e24.12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.61\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.66\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.27\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003ewstc\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.64 [0.60 - 0.68]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e80.35\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.54\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.68\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.22\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.68 [0.65 - 0.71]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e82.45\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.61\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.66\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.27\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003ewhpr\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.59 [0.55 - 0.63]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.86\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.55\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.15\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.61 [0.58 - 0.65]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.84\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.62\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.55\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.17\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003ewhtr\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.62 [0.58 - 0.66]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.48\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.54\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.67\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.21\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.68 [0.65 - 0.71]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.53\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.59\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.29\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eDiastolic bp \u0026gt; 90 mmHg\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eBMI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.63 [0.59 - \u0026nbsp;0.67]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e21.78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.67\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.54\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.21\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.63 [0.60 - 0.67]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e24.12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.59\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.66\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.25\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003ewstc\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.65 [0.60 - 0.69]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e81.45\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.53\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.73\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.26\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.65 [0.61 - \u0026nbsp;0.68]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e81.85\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.61\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.62\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.23\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003ewhpr\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.58 [ 0.53 - 0.62]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.87\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.65\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.15\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.58 [0.54 - 0.61]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.84\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.59\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.55\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.14\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003ewhtr\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.64 [ 0.59 - 0.68]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.49\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.51\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.26\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.64 [0.61 - 0.67]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.54\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.52\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.72\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\"\u003e\n \u003cp\u003e0.24\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cem\u003eAUC values presented with 95% confidence intervals (CI) in parentheses. Optimal cutoff values determined by the Youden Index.BMI \u0026ndash; body mass index; WstC \u0026ndash; waist circumference; WHpR \u0026ndash; waist-to-hip ratio; WHtR \u0026ndash; waist-to-height ratio; AUC \u0026ndash; area under the receiver operating characteristic curve; BP \u0026ndash; blood pressure.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExploratory Feature Analysis and Mutual Information (MI)\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe feature correlation matrix (Figure 3) showed moderate correlations among anthropometric indices (e.g., BMI with waist circumference, r = 0.70), reflecting shared adiposity constructs. Pairwise correlations between individual predictors and high blood pressure were modest (r = 0.20\u0026ndash;0.23), confirming that no single variable dominates hypertension risk.\u003c/p\u003e\n\u003cp\u003eMutual information scores shown in Figure 4 indicated that age contributed the most standalone information toward hypertension (MI = 0.031), followed by waist-to-hip ratio (MI = 0.031), waist-to-height ratio (MI = 0.026), waist circumference (MI = 0.025), and BMI (MI = 0.022). Behavioural variables such as smoking (MI = 0.015), salt intake (MI = 0.011), and cholesterol (MI = 0.008) contributed modestly, while physical activity, sex, education, and residence contributed minimal information (MI \u0026lt; 0.002).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMachine Learning Model Performance\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTable 4 summarizes model performance across all five classifiers. Logistic Regression and Linear SVM achieved the highest ROC AUC (both 0.74) and the highest recall (0.70 and 0.69, respectively), indicating superior identification of hypertensive individuals. Their F1 scores (both 0.51) reflected a reasonable balance between sensitivity and precision.\u003c/p\u003e\n\u003cp\u003eTree-based models (Random Forest, Gradient Boosting, and XGBoost) achieved higher overall accuracy (0.78\u0026ndash;0.79) but substantially lower recall (0.22\u0026ndash;0.26), indicating that these models favoured the majority normotensive class. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e4\u003c/strong\u003e\u003cstrong\u003e.\u003c/strong\u003e Machine Learning Model Performance for Predicting High Blood Pressure Using the Zambia WHO STEPS Dataset\u003c/p\u003e\n\u003cdiv align=\"center\"\u003e\n \u003ctable style=\"width: 100%;\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eModel\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eROC AUC\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003ePR AUC\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eAccuracy\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003ePrecision\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eRecall\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eF1 Score\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eLinear SVM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003e0.74\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003e0.50\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.69\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.41\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.69\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003e0.51\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eLogistic Regression\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003e0.74\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.49\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.69\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.40\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003e0.70\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003e0.51\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eRandom Forest\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.73\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.49\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003e0.79\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003e0.64\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.22\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.32\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eGradient Boosting\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.70\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.47\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.60\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.23\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.33\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eXGBoost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.69\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.46\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.57\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.26\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.36\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003ePerformance metrics evaluated on a held-out test set. Bold values indicate the best-performing model for each metric. ROC AUC \u0026ndash; area under the receiver operating characteristic curve; PR AUC \u0026ndash; area under the precision-recall curve; F1 \u0026ndash; harmonic mean of precision and recall; SVM \u0026ndash; support vector machine.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe ROC curves in Figure 5 show moderate and closely aligned discrimination, with Logistic Regression and Linear SVM achieving the highest ROC AUC values (about 0.74), followed by Random Forest (about 0.73). These patterns indicate that traditional and machine-learning models achieved similar discriminatory performance, with no substantial advantage from increased model complexity in this exploratory analysis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFeature Importance\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe SHAP analysis in Figure 6 from the best performing logistic regression model identifies age as the strongest predictor of high blood pressure among Zambian adults aged 18\u0026ndash;69 years in the WHO STEPS Survey, with risk increasing steadily with advancing age. Measures of adiposity, particularly waist circumference and body mass index (BMI), also show strong positive associations, highlighting the importance of central and overall obesity in hypertension risk. Waist-to-height ratio demonstrates a more complex, potentially non-linear relationship, while civil or marital status and sex contribute moderate effects. Fasting blood glucose shows strong positive effects among individuals with very high levels, consistent with the link between dysglycaemia and hypertension, whereas total cholesterol shows some inverse associations in specific subgroups. Other factors, including salt intake, physical activity, diet, education, and smoking, contribute minimally after accounting for age and obesity measures. Overall, age and key adiposity, particularly waist circumference, indicators emerge as the strongest predictors, supporting their use in screening for high blood pressure in this population.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eCentral obesity markers identified high blood pressure more accurately than BMI in our study. In women, waist-to-height ratio and waist circumference outperformed BMI, while waist-to-hip ratio performed poorly in men. The optimal cut-offs for all measures fell below WHO thresholds. BMI showed moderate sensitivity and specificity, while waist-to-height ratio and waist circumference performed similarly, and waist-to-hip ratio remained less consistent. The models showed moderate discrimination, with random forest achieving the highest accuracy but lower recall, while logistic regression and linear SVM achieved better recall and overall balance, and age ranked as the strongest predictor followed by BMI, waist circumference, waist-to-height ratio, and cholesterol.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAnthropometric Differences and the Role of Central Obesity in High Blood Pressure\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIndividuals with high blood pressure in this study demonstrated significantly higher body mass index (BMI), waist circumference (WC), waist-to-height ratio (WHtR), and waist-to-hip ratio (WHpR) across both sexes and all blood pressure categories. This pattern aligns with findings reported in various global regions. Central obesity indicators, particularly WC and WHtR, exhibited greater discriminatory power for high blood pressure status compared to BMI. This distinction was especially evident among women, as the area under the curve (AUC) values for WHtR and WC approached 0.66, whereas BMI reached 0.64. These findings are consistent with studies conducted in rural India, Nigeria, and Ethiopia, which also identified strong associations between both BMI and central adiposity measures and elevated blood pressure.\u003csup\u003e33\u0026ndash;36\u003c/sup\u003e . Similarly, research from China and Korea demonstrates that central obesity indices (WC, WHR, WHtR) are superior to BMI in predicting hypertension\u003csup\u003e33\u0026ndash;36\u003c/sup\u003e . The biological credibility of these associations is well established, with mechanisms including insulin resistance, sodium retention, sympathetic nervous system activation, and vascular dysfunction contributing to the link between adiposity and hypertension\u0026nbsp;\u003csup\u003e33,37,38\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDiscriminatory Performance and Optimal Cutoff Values of Obesity Indices\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOur results reinforce that BMI, waist circumference (WC), waist-to-hip ratio (WHpR), and waist-to-height ratio (WHtR) can all help identify high blood pressure in Zambian adults. However, our ROC analysis shows that WHtR and WC were more accurate than BMI, while WHpR was less reliable, especially for men (AUC = 0.59). This trend, where central obesity measures outperform BMI,has also been seen in studies from Malaysia, China, and Vietnam, where WHtR and WC offered higher AUC values and were better at predicting hypertension, particularly for women.\u003c/p\u003e\n\u003cp\u003e\u003csup\u003e34,39\u0026ndash;41\u003c/sup\u003e. Notably, the optimal cut-off values for these indices in our Zambian sample were generally lower than the WHO-recommended thresholds, a trend also observed in Ethiopia, several Asian populations, and among Albanians, where BMI and WC cut-offs for hypertension risk were consistently below international standards\u003csup\u003e33,42\u0026ndash;44\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFor example, our study identified BMI cut-offs of 21.97 kg/m\u0026sup2; for men and 24.12 kg/m\u0026sup2; for women, compared to the WHO\u0026rsquo;s 25 kg/m\u0026sup2;, and WC cut-offs of 80.35 cm for men and 81.85 cm for women, both lower than the WHO\u0026rsquo;s 94 cm and 80 cm, respectively. Similar lower thresholds have been reported in Ethiopia, Hong Kong, Shandong (China), and Taiwan, particularly among men, while studies in Albania and Korea also found optimal BMI and WC cut-offs below WHO recommendations\u0026nbsp;\u003csup\u003e12,33,35,42,43,45,46\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSensitivity and Specificity of \u0026nbsp;Index Cutoffs and Implications for Population Screening\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe performance of WHpR in our study, with cut-offs of 0.86 for men and 0.84 for women, was also lower than the WHO definition of abdominal obesity (\u0026ge;0.90 for men and \u0026ge;0.85 for women), mirroring findings from Ethiopia and Hong Kong, though sensitivity estimates were generally lower than those observed in Chinese populations\u003csup\u003e33,44,45\u003c/sup\u003e. In contrast, WHtR demonstrated the most consistent performance, with cut-off values of 0.48 for men and 0.53 for women, closely aligning with findings from Hong Kong, Taiwan, Shandong (China), Western Ethiopia, and Vietnam, where optimal WHtR thresholds ranged between 0.47 and 0.55\u003csup\u003e33,39,45\u003c/sup\u003e. \u0026nbsp;While specificity estimates for WHtR were comparable across studies, sensitivity remained modest in both sexes, underscoring the need for population-specific thresholds.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePopulation and Sex Differences in Optimal Anthropometric Thresholds\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eComparative studies across regions highlight the variability in optimal anthropometric cut-offs, often influenced by population structure, urbanization, and ethnic differences in body composition. For instance, urban Ethiopian cohorts reported higher BMI cut-offs than rural populations, and studies in Nigeria and Iran found higher WC cut-offs for women than men, diverging from WHO standards\u003csup\u003e12,33,47\u003c/sup\u003e. \u0026nbsp;In South Africa and Cambodia, WC and WHtR were better predictors of hypertension in women, while BMI and WC were more appropriate for men, emphasizing the importance of sex-specific and regionally tailored screening tools\u003csup\u003e37,46\u003c/sup\u003e.\u0026nbsp;Furthermore, meta-analyses and large-scale studies in China, Korea, and the US confirm that central obesity measures (WC, WHtR, WHR) are generally superior to BMI for hypertension prediction, though the magnitude of association and optimal thresholds vary by ethnicity, age, and sex\u0026nbsp;\u003csup\u003e34,48\u0026ndash;51\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMachine Learning\u0026ndash;Based Prediction of Hypertension Using Obesity Indices\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOur study also evaluated the predictive performance of several machine learning models, including Logistic Regression, Linear SVM, Random Forest, and Gradient Boosting, for hypertension classification using anthropometric indices and additional demographic and biochemical factors. Consistent with prior research, our results showed that Logistic Regression and Linear SVM achieved the highest discriminatory performance (ROC AUC = 0.74), closely followed by Random Forest (AUC = 0.73) and Gradient Boosting (AUC = 0.72). Ensemble models such as Random Forest demonstrated higher overall accuracy (up to 79.0%) but lower recall (0.22\u0026ndash;0.26), reflecting challenges with class imbalance, while Logistic Regression and Linear SVM provided more balanced recall (0.68\u0026ndash;0.69) and F1 scores (0.51), indicating better sensitivity to hypertensive cases. Feature importance analysis confirmed that age was the dominant predictor, followed by BMI, waist circumference, waist-to-height ratio, and cholesterol, supporting the continued relevance of these indices in multivariable models\u003csup\u003e19,52\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eComparative Performance of Traditional and Machine Learning Models for Prediction\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOur findings align with previous studies that have compared traditional and machine learning approaches for hypertension prediction. For example, ensemble models such as Random Forest and XGBoost have been shown to achieve high accuracy and AUC in various populations, with feature importance analyses consistently highlighting age, BMI, waist circumference, and related anthropometric measures as key predictors\u003csup\u003e53,54\u003c/sup\u003e. However, as in our study, no single obesity index or model consistently outperformed others across all metrics or populations. The modest improvements observed when integrating multiple indices and demographic variables into predictive models underscore the value of combining simple, non-invasive measures for population-level risk stratification, as also demonstrated in large-scale studies using WHO STEPS and similar datasets\u003csup\u003e12,19\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eClinical Implications\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA key clinical implication for hypertension screening in Zambia is the need to use locally derived anthropometric cut-off values, rather than relying solely on WHO thresholds, for measures such as waist circumference (WC), waist-to-height ratio (WHtR), and body mass index (BMI). Evidence from Zambia and neighboring countries shows that standard WHO cut-offs may not accurately identify all individuals at risk, potentially leading to underdiagnosis and missed opportunities for early intervention. Adopting population-specific thresholds, informed by local data, can improve the sensitivity and specificity of screening programs, ensuring that more at-risk individuals are identified and managed appropriate\u003csup\u003e55,56\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eSecondly, integrating simple, validated predictive models that combine age and anthropometric indicators into primary health care is especially important in resource-limited settings. These models, when embedded within existing primary care infrastructure and supported by community health workers, can expand the reach of hypertension screening, improve case detection, and facilitate timely linkage to care. Community-based screening, particularly when integrated with other chronic disease services, has been shown to be both effective and cost-efficient, leading to significant reductions in cardiovascular morbidity and mortality across Africa\u0026nbsp;\u003csup\u003e57,58\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eSex differences in hypertension risk prediction also permit attention. Central obesity measures like WC and WHtR are particularly strong predictors of hypertension in women, suggesting that sex-specific cut-off values or screening protocols may further enhance detection rates\u003csup\u003e56,59\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThese recommendations are urgent given Zambia\u0026rsquo;s rising burden of non-communicable diseases, driven by urbanization and lifestyle changes, and the persistently low rates of hypertension awareness, treatment, and control, especially in rural and underserved communities. Addressing these gaps requires expanding universal screening, adopting locally relevant anthropometric thresholds, integrating risk prediction tools into primary care, and ensuring equitable access to follow-up and treatment\u0026nbsp;\u003csup\u003e56,60\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePublic Health Relevance and Future Directions for Hypertension Risk Prediction\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eImportantly, our models were not tuned, as this was an exploratory analysis, yet their performance was comparable to or slightly better than those reported in earlier studies using similar datasets and methods\u0026nbsp;\u003csup\u003e19,53,54\u003c/sup\u003e. This suggests that even untuned, interpretable models can provide valuable insights for public health screening and risk stratification, particularly in resource-limited settings.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFuture research should focus on model tuning, external validation, and the development of country- or region-specific models to further enhance predictive accuracy and generalizability. Additionally, the integration of machine learning-based prediction tools into health systems could support targeted interventions and more efficient allocation of resources for hypertension prevention and control.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStrengths and Limitations\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOur study has several limitations. Its cross-sectional design precludes causal inference between anthropometric indicators and hypertension. Blood pressure and behavioural variables were measured at a single point in time, which may not fully capture long-term exposure or variability. Although the WHO STEPS methodology ensures standardized measurements, some variables relied on self-reported information and may be subject to recall or social desirability bias. In addition, the machine-learning models were not extensively tuned or externally validated, and their predictive performance should therefore be interpreted as exploratory. Despite these limitations, the large, nationally representative sample and standardized measurements strengthen the reliability and public-health relevance of the findings.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eCentral obesity indices, particularly WHtR and WC, outperformed BMI in discriminating high blood pressure among Zambian adults, with optimal cutoff values consistently lower than WHO recommended thresholds. Simple, interpretable models combining age and anthropometric indices achieved discriminatory performance comparable to more complex machine learning algorithms. These findings support the adoption of locally derived, sex specific anthropometric cutoffs for hypertension screening in Zambia and the integration of simple prediction tools within primary care and community health programmes to strengthen early detection of undiagnosed hypertension.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eAUC\u0026emsp;\u0026emsp;Area Under the Curve\u003c/p\u003e\n\u003cp\u003eBMI\u0026emsp;\u0026emsp;Body Mass Index\u003c/p\u003e\n\u003cp\u003eCI\u0026emsp;\u0026emsp;\u0026emsp;Confidence Interval\u003c/p\u003e\n\u003cp\u003eLMICs\u0026emsp;Low- and Middle-Income Countries\u003c/p\u003e\n\u003cp\u003eNCDs\u0026emsp;\u0026emsp;Non-Communicable Diseases\u003c/p\u003e\n\u003cp\u003eOR\u0026emsp;\u0026emsp;\u0026emsp;Odds Ratio\u003c/p\u003e\n\u003cp\u003eROC\u0026emsp;\u0026emsp;Receiver Operating Characteristic\u003c/p\u003e\n\u003cp\u003eSSA\u0026emsp;\u0026emsp;Sub-Saharan Africa\u003c/p\u003e\n\u003cp\u003eSTEPS\u0026emsp;STEPwise Approach to NCD Risk Factor Surveillance\u003c/p\u003e\n\u003cp\u003eWC\u0026emsp;\u0026emsp;\u0026emsp;Waist Circumference\u003c/p\u003e\n\u003cp\u003eWHpR\u0026emsp;\u0026emsp;Waist-to-Hip Ratio\u003c/p\u003e\n\u003cp\u003eWHtR\u0026emsp;\u0026emsp;Waist-to-Height Ratio\u003c/p\u003e\n\u003cp\u003eWHO\u0026emsp;\u0026emsp;World Health Organization\u003c/p\u003e\n"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors thank the World Health Organization and the Ministry of Health, Zambia for providing access to the 2017 Zambia WHO STEPwise Approach to Surveillance (STEPS) dataset for research purposes. The authors also acknowledge the survey teams and all study participants for their valuable contributions to the collection of this nationally representative dataset.\u003c/p\u003e\n\u003cp\u003eWe sincerely acknowledge the US\u0026ndash;Zambia NCD Risk Project for its support and collaboration. We thank mentors at The George Washington University Milken Institute School of Public Health, including Dr. Heather Rosen and Dr. Nino Paichadze, as well as colleagues at The University of Zambia, including Dr. Cosmas Zyambo, Prof. Choolwe Jacobs, and Dr. Adam Silumbwe, for their guidance and mentorship throughout this study. The authors confirm that these institutions had no role in the design, analysis, interpretation, or writing of this manuscript. Any errors or omissions remain the responsibility of the authors\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSM led the conceptualisation of the study, conducted the statistical analysis, and drafted the manuscript. CM, GM, WM and PM contributed to study conceptualisation, interpretation of the results, and critically reviewed the analysis and manuscript. All authors reviewed and approved the final version of the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study received no specific funding.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data analyzed in our study are derived from the 2017 Zambia WHO STEPwise Approach to NCD Risk Factor Surveillance (STEPS) survey and are publicly available through the World Health Organization NCD Microdata Repository. Access to the data is subject to World Health Organization data access approval procedures.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics Approval and Consent to Participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe 2017 Zambia WHO STEPwise Approach to Surveillance (STEPS) survey was conducted in accordance with the ethical principles outlined in the Declaration of Helsinki. Ethical approval for the original survey was obtained from the relevant national ethics authorities in Zambia, with technical oversight from the World Health Organization.\u003c/p\u003e\n\u003cp\u003eThis study was a secondary analysis of de-identified, publicly available STEPS data accessed through an approved World Health Organization data request. Ethical clearance for the secondary analysis was obtained from the University of Zambia Biomedical Research Ethics Committee (UNZABREC) with approval No. REF. 6-492-2025 and registered with the National Health Research Authority (NHRA) with registration number NHRA-2135/15/04/2025. The analysis involved no direct contact with participants and utilized anonymized data.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for Publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable. This manuscript does not contain any individual-level identifiable data.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting Interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interest\u003cstrong\u003es\u003c/strong\u003e\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eBl\u0026uuml;her, M. Obesity: global epidemiology and pathogenesis. \u003cem\u003eNat. Rev. Endocrinol.\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 288\u0026ndash;298 (2019).\u003c/li\u003e\n\u003cli\u003eHruby, A. \u0026amp; Hu, F. B. The Epidemiology of Obesity: A Big Picture. \u003cem\u003ePharmacoEconomics\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 673\u0026ndash;689 (2015).\u003c/li\u003e\n\u003cli\u003eMills, K. T., Stefanescu, A. \u0026amp; He, J. The global epidemiology of hypertension. \u003cem\u003eNat. Rev. Nephrol.\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 223\u0026ndash;237 (2020).\u003c/li\u003e\n\u003cli\u003eAttia, A. \u003cem\u003eet al.\u003c/em\u003e Abstract TH122: Prevalence and Associated Risk Factors of Hypertension in Urban and Rural Communities of Sub-Saharan Africa. \u003cem\u003eHypertension\u003c/em\u003e \u003cstrong\u003e82\u003c/strong\u003e, hyp.82.suppl_1.TH122 (2025).\u003c/li\u003e\n\u003cli\u003eGuwatudde, D. \u003cem\u003eet al.\u003c/em\u003e The burden of hypertension in sub-Saharan Africa: a four-country cross sectional study. \u003cem\u003eBMC Public Health\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 1211 (2015).\u003c/li\u003e\n\u003cli\u003eNwankwo, M. \u003cem\u003eet al.\u003c/em\u003e Hypertension in West Africa: a systematic review and meta-analysis of prevalence and associated risk factors. \u003cem\u003eBMC Public Health\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 3269 (2025).\u003c/li\u003e\n\u003cli\u003eHendriks, M. E. \u003cem\u003eet al.\u003c/em\u003e Hypertension in Sub-Saharan Africa: Cross-Sectional Surveys in Four Rural and Urban Communities. \u003cem\u003ePLoS ONE\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, e32638 (2012).\u003c/li\u003e\n\u003cli\u003ePrice, A. J. \u003cem\u003eet al.\u003c/em\u003e Prevalence of obesity, hypertension, and diabetes, and cascade of care in sub-Saharan Africa: a cross-sectional, population-based study in rural and urban Malawi. \u003cem\u003eLancet Diabetes Endocrinol.\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 208\u0026ndash;222 (2018).\u003c/li\u003e\n\u003cli\u003eSchutte, A. E. Hypertension in Low- and Middle-Income Countries. \u003cem\u003eHHS Pub Lic Access Physiol. Behav.\u003c/em\u003e \u003cstrong\u003e176\u003c/strong\u003e, 139\u0026ndash;148 (2021).\u003c/li\u003e\n\u003cli\u003ePengpid, S. \u0026amp; Peltzer, K. Prevalence and correlates of multiple non-communicable disease risk factors among adults in zambia: Results of the first national steps survey in 2017. \u003cem\u003ePan Afr. Med. J.\u003c/em\u003e \u003cstrong\u003e37\u003c/strong\u003e, 1\u0026ndash;12 (2020).\u003c/li\u003e\n\u003cli\u003eHaregu, T. N., Oti, S., Egondi, T. \u0026amp; Kyobutungi, C. Measurement of overweight and obesity an urban slum setting in sub-Saharan Africa: a comparison of four anthropometric indices. \u003cem\u003eBMC Obes.\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, 46 (2016).\u003c/li\u003e\n\u003cli\u003ePatel, S. A. \u003cem\u003eet al.\u003c/em\u003e Obesity and its Relation With Diabetes and Hypertension. \u003cem\u003eGlob. Heart\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 71-79.e4 (2016).\u003c/li\u003e\n\u003cli\u003eZhou, B. \u003cem\u003eet al.\u003c/em\u003e Worldwide trends in hypertension prevalence and progress in treatment and control from 1990 to 2019: a pooled analysis of 1201 population-representative studies with 104 million participants. \u003cem\u003eThe Lancet\u003c/em\u003e \u003cstrong\u003e398\u003c/strong\u003e, 957\u0026ndash;980 (2021).\u003c/li\u003e\n\u003cli\u003eAkpa, O. M. \u003cem\u003eet al.\u003c/em\u003e Regional Patterns and Association Between Obesity and Hypertension in Africa: Evidence From the H3Africa CHAIR Study. \u003cem\u003eHypertension\u003c/em\u003e \u003cstrong\u003e75\u003c/strong\u003e, 1167\u0026ndash;1178 (2020).\u003c/li\u003e\n\u003cli\u003eHaregu, T. N., Oti, S., Egondi, T. \u0026amp; Kyobutungi, C. Measurement of overweight and obesity an urban slum setting in sub-Saharan Africa: a comparison of four anthropometric indices. \u003cem\u003eBMC Obes.\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, 46 (2016).\u003c/li\u003e\n\u003cli\u003ePatel, S. A. \u003cem\u003eet al.\u003c/em\u003e Obesity and its Relation With Diabetes and Hypertension: A Cross-Sectional Study Across 4 Geographical Regions. \u003cem\u003eGlob. Heart\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 71 (2016).\u003c/li\u003e\n\u003cli\u003eJovel, J. \u0026amp; Greiner, R. An Introduction to Machine Learning Approaches for Biomedical Research. \u003cem\u003eFront. Med.\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 771607 (2021).\u003c/li\u003e\n\u003cli\u003eSidey-Gibbons, J. A. M. \u0026amp; Sidey-Gibbons, C. J. Machine learning in medicine: a practical introduction. \u003cem\u003eBMC Med. Res. Methodol.\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 64 (2019).\u003c/li\u003e\n\u003cli\u003eLatifa, A. A., Id, L. S. A., Al, M. F., Id, A. \u0026amp; Id, M. E. A. Predicting hypertension using machine learning : Findings from Qatar Biobank Study. 1\u0026ndash;17 (2020) doi:10.1371/journal.pone.0240370.\u003c/li\u003e\n\u003cli\u003eLopez-Jimenez, F. \u003cem\u003eet al.\u003c/em\u003e Obesity and cardiovascular disease: mechanistic insights and management strategies. A joint position paper by the World Heart Federation and World Obesity Federation. \u003cem\u003eEur. J. Prev. Cardiol.\u003c/em\u003e \u003cstrong\u003e29\u003c/strong\u003e, 2218\u0026ndash;2237 (2022).\u003c/li\u003e\n\u003cli\u003eZhao, H. \u003cem\u003eet al.\u003c/em\u003e Predicting the Risk of Hypertension Based on Several Easy-to-Collect Risk Factors: A Machine Learning Method. \u003cem\u003eFront. Public Health\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 619429 (2021).\u003c/li\u003e\n\u003cli\u003eIslam, Md. M. \u003cem\u003eet al.\u003c/em\u003e Predicting the risk of hypertension using machine learning algorithms: A cross sectional study in Ethiopia. \u003cem\u003ePLOS ONE\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, e0289613 (2023).\u003c/li\u003e\n\u003cli\u003eSeptian, E. \u003cem\u003eet al.\u003c/em\u003e Prediction of Personalised Hypertension Using Machine Learning in Indonesian Population. \u003cem\u003eJ. Med. Syst.\u003c/em\u003e \u003cstrong\u003e49\u003c/strong\u003e, 137 (2025).\u003c/li\u003e\n\u003cli\u003eSifat, I. K. \u0026amp; Kibria, Md. K. Optimizing hypertension prediction using ensemble learning approaches. \u003cem\u003ePLOS ONE\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, e0315865 (2024).\u003c/li\u003e\n\u003cli\u003eBisong, E. \u003cem\u003eet al.\u003c/em\u003e Predicting high blood pressure using machine learning models in low- and middle-income countries. \u003cem\u003eBMC Med. Inform. Decis. Mak.\u003c/em\u003e \u003cstrong\u003e24\u003c/strong\u003e, 234 (2024).\u003c/li\u003e\n\u003cli\u003eBujang, M. A. Requirements for Minimum Sample Size for Sensitivity and Specificity Analysis. \u003cem\u003eJ. Clin. Diagn. Res.\u003c/em\u003e https://doi.org/10.7860/JCDR/2016/18129.8744 (2016) doi:10.7860/JCDR/2016/18129.8744.\u003c/li\u003e\n\u003cli\u003eEgan, B. M. \u0026amp; Zhao, Y. Different Definitions of Prevalent Hypertension Impact: The Clinical Epidemiology of Hypertension and Attainment of Healthy People Goals. \u003cem\u003eJ. Clin. Hypertens.\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 154\u0026ndash;161 (2013).\u003c/li\u003e\n\u003cli\u003eHaibo He \u0026amp; Garcia, E. A. Learning from Imbalanced Data. \u003cem\u003eIEEE Trans. Knowl. Data Eng.\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, 1263\u0026ndash;1284 (2009).\u003c/li\u003e\n\u003cli\u003eStataCorp LLC. \u003cem\u003eStata Statistical Software: Release 17\u003c/em\u003e. (StataCorp LLC, College Station, TX, USA, 2021).\u003c/li\u003e\n\u003cli\u003ePython Software Foundation. Python: A Programming Language. (2023).\u003c/li\u003e\n\u003cli\u003ePedregosa, F. \u003cem\u003eet al.\u003c/em\u003e Scikit-learn: Machine learning in Python. \u003cem\u003eJ. Mach. Learn. Res.\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 2825\u0026ndash;2830 (2011).\u003c/li\u003e\n\u003cli\u003eChen, T. \u0026amp; Guestrin, C. Xgboost: A scalable tree boosting system. in \u003cem\u003eProceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining\u003c/em\u003e 785\u0026ndash;794 (2016).\u003c/li\u003e\n\u003cli\u003eGutema, B. T. \u003cem\u003eet al.\u003c/em\u003e Predictive capacity of obesity indices for high blood pressure among southern Ethiopian adult population: A WHO STEPS survey. \u003cem\u003eBMC Cardiovasc. Disord.\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 1\u0026ndash;9 (2020).\u003c/li\u003e\n\u003cli\u003eLee, X., Gao, Y., Zhang, Y. \u0026amp; Feng, Y. Comparison of obesity-related indices for predicting hypertension based on ROC analysis in Chinese adults. \u003cem\u003eFront. Public Health\u003c/em\u003e https://doi.org/doi:%2010.3389/fpubh.2022.1042236 (2022) doi:doi: 10.3389/fpubh.2022.1042236.\u003c/li\u003e\n\u003cli\u003eLee, J., Lim, N., Baek, T., Park, S. \u0026amp; Park, H. Anthropometric indices as predictors of hypertension among men and women aged 40 \u0026ndash; 69 years in the Korean population : the Korean Genome and Epidemiology Study. 1\u0026ndash;7 (2015) doi:10.1186/s12889-015-1471-5.\u003c/li\u003e\n\u003cli\u003eAdegoke, O., Ozoh, O. B., Odeniyi, I. A., Bello, B. T. \u0026amp; Okubadejo, N. U. Prevalence of obesity and an interrogation of the correlation between anthropometric indices and blood pressures in urban. \u003cem\u003eSci. Rep.\u003c/em\u003e 1\u0026ndash;12 (2021) doi:10.1038/s41598-021-83055-w.\u003c/li\u003e\n\u003cli\u003eModjadji, P., Salane, M. C., Mokwena, K. E., Mudau, T. S. \u0026amp; Mphekgwana, P. M. Utility of Obesity Indicators for Predicting Hypertension among Older Persons in Limpopo Province, South Africa. \u003cem\u003eAppl. Sci. Switz.\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, (2022).\u003c/li\u003e\n\u003cli\u003eChen, L. \u003cem\u003eet al.\u003c/em\u003e Association of different obesity patterns with hypertension in US male adults : a cross ‑ sectional study. 1\u0026ndash;10 (2023) doi:10.1038/s41598-023-37302-x.\u003c/li\u003e\n\u003cli\u003eZhang, Y. \u003cem\u003eet al.\u003c/em\u003e Association between anthropometric indicators of obesity and cardiovascular risk factors among adults in Shanghai , China. 1\u0026ndash;9 (2019).\u003c/li\u003e\n\u003cli\u003eQuan, N. M. \u0026amp; Nguyen, M. H. Anthropometric Indexes for Predicting High Blood Pressure in Vietnamese Adults : A Cross-Sectional Study. \u003cem\u003eDovepress\u003c/em\u003e 181\u0026ndash;186 (2020).\u003c/li\u003e\n\u003cli\u003eYing, J., Tee, H., Gan, W. Y. \u0026amp; Lim, P. Y. Comparisons of body mass index , waist a body shape index ( ABSI ) in predicting high blood pressure among Malaysian adolescents : a cross- \u0026shy; sectional study. 1\u0026ndash;9 (2020) doi:10.1136/bmjopen-2019-032874.\u003c/li\u003e\n\u003cli\u003eGupta, R. Das \u0026amp; Hashan, M. R. Association Between Abdominal Obesity , Body Mass Index , and Hypertension in India : Evidence From a Large Nationally Representative Data. \u003cem\u003eJ. Clin. Hypertens.\u003c/em\u003e https://doi.org/10.1111/jch.70034 (2025) doi:10.1111/jch.70034.\u003c/li\u003e\n\u003cli\u003eHu, J., Zhong, Y., Ge, W., Lv, H. \u0026amp; Ding, Z. Comparisons of tri-ponderal mass index and body mass index in discriminating hypertension at three separate visits in adolescents : A retrospective cohort study. 1\u0026ndash;11 (2022) doi:10.3389/fnut.2022.1028861.\u003c/li\u003e\n\u003cli\u003eKawamoto, R., Kikuchi, A., Akase, T., Ninomiya, D. \u0026amp; Kumagi, T. Usefulness of waist-to-height ratio in screening incident hypertension among Japanese community-dwelling middle-aged and elderly individuals. \u003cem\u003eClin. Hypertens.\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 1\u0026ndash;9 (2020).\u003c/li\u003e\n\u003cli\u003eIslam, M. R., Md, M. \u0026amp; Samaha, M. S. Relationship of Anthropometric Indicators of General and Abdominal Obesity with Hypertension and Their Predictive Performance among Albanians : A Nationwide. \u003cem\u003eMDPI Nutr.\u003c/em\u003e (2021).\u003c/li\u003e\n\u003cli\u003eShimotake, Y. \u003cem\u003eet al.\u003c/em\u003e Heliyon Comparative evaluation of anthropometric measurements and prevalence of hypertension : community based cross-sectional study in rural male and female Cambodians. \u003cem\u003eHeliyon\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, e04432 (2020).\u003c/li\u003e\n\u003cli\u003eAdegoke, O. \u003cem\u003eet al.\u003c/em\u003e Prevalence of obesity and an interrogation of the correlation between anthropometric indices and blood pressures in urban Lagos, Nigeria. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 1\u0026ndash;12 (2021).\u003c/li\u003e\n\u003cli\u003eGui, J. \u003cem\u003eet al.\u003c/em\u003e Obesity ‑ and lipid ‑ related indices as a risk factor of hypertension in mid ‑ aged and elderly Chinese : a cross ‑ sectional study. \u003cem\u003eBMC Geriatr.\u003c/em\u003e 1\u0026ndash;13 (2024) doi:10.1186/s12877-023-04650-2.\u003c/li\u003e\n\u003cli\u003eJiang, X. \u003cem\u003eet al.\u003c/em\u003e Gender differences in associations between obesity and hypertension , diabetes , dyslipidemia : evidence from electronic health records of 3 . 5 million Chinese senior population. (2025).\u003c/li\u003e\n\u003cli\u003eChao, Wanga. ;, Fub., W. \u0026amp; Caoa, S. Association of Adiposity Indicators with Hypertension among Chinese Adults Running title : Obesity indicators and hypertension Chao Wang. \u003cem\u003eSciencedirect\u003c/em\u003e (2021).\u003c/li\u003e\n\u003cli\u003eGe, Q. \u003cem\u003eet al.\u003c/em\u003e Comparison of different obesity indices related with hypertension among different sex and age groups in China. \u003cem\u003eNutr. Metab. Cardiovasc. Dis. 313 793-801\u003c/em\u003e \u003cstrong\u003e31(3), 793\u003c/strong\u003e, (2021).\u003c/li\u003e\n\u003cli\u003eSifat, I. K. \u0026amp; Kaderi, K. Optimizing hypertension prediction using ensemble learning approaches. 1\u0026ndash;17 (2024) doi:10.1371/journal.pone.0315865.\u003c/li\u003e\n\u003cli\u003eSilva, G. F. S., Fagundes, T. P. \u0026amp; Teixeira, B. C. Machine Learning for Hypertension Prediction: a Systematic Review. \u003cem\u003ehttps://link.springer.com/\u003c/em\u003e (2022).\u003c/li\u003e\n\u003cli\u003eAhmed, F., Ali, S., Rahman, J. \u0026amp; Roy, D. C. Predicting the risk of hypertension using machine learning algorithms : A cross sectional study in Ethiopia. 1\u0026ndash;20 (2023) doi:10.1371/journal.pone.0289613.\u003c/li\u003e\n\u003cli\u003eShah, P. \u003cem\u003eet al.\u003c/em\u003e Abstract P1057: Hypertension Screening in Children and Adolescents: A Global Policy Analysis. \u003cem\u003eCirculation\u003c/em\u003e \u003cstrong\u003e151\u003c/strong\u003e, cir.151.suppl_1.P1057 (2025).\u003c/li\u003e\n\u003cli\u003eTateyama, Y. \u003cem\u003eet al.\u003c/em\u003e Hypertension, its correlates and differences in access to healthcare services by gender among rural Zambian residents: A cross-sectional study. \u003cem\u003eBMJ Open\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, (2022).\u003c/li\u003e\n\u003cli\u003eHickey, M. D. \u003cem\u003eet al.\u003c/em\u003e Cost-effectiveness of leveraging existing HIV primary health systems and community health workers for hypertension screening and treatment in Africa: An individual-based modeling study. \u003cem\u003ePLOS Med.\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, e1004531 (2025).\u003c/li\u003e\n\u003cli\u003eMutelo, L. \u003cem\u003eet al.\u003c/em\u003e Prevalence and correlates of microalbuminuria in high-risk persons with hypertension, diabetes, or HIV at a tertiary hospital in Zambia. \u003cem\u003ePLOS One\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, e0328529 (2025).\u003c/li\u003e\n\u003cli\u003eKim, J.-H. \u003cem\u003eet al.\u003c/em\u003e Examining geospatial and temporal distribution of invasive non-typhoidal \u003cem\u003eSalmonella\u003c/em\u003e disease occurrence in sub-Saharan Africa: a systematic review and modelling study. \u003cem\u003eBMJ Open\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, e080501 (2024).\u003c/li\u003e\n\u003cli\u003eLucinde, R. K. \u003cem\u003eet al.\u003c/em\u003e Diagnostic Performance of Unattended Automated Office Blood Pressure Measurement for Hypertension Screening Among People With and Without HIV. \u003cem\u003eJ. Am. Heart Assoc.\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, e043957 (2025).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"University of Zambia","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"hypertension, blood pressure, obesity indices, anthropometric indices, machine learning, prediction","lastPublishedDoi":"10.21203/rs.3.rs-9585939/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9585939/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackground:\u003c/strong\u003e Hypertension is a leading contributor to cardiovascular morbidity in sub-Saharan Africa. BMI, waist circumference (WC), waist-to-height ratio (WHtR), and waist-to-hip ratio (WHpR) are widely used for screening, yet their optimal thresholds and predictive utility vary by population. This study evaluated these indices for hypertension prediction in Zambian adults using traditional and machine learning approaches.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods:\u003c/strong\u003e This cross-sectional study assessed hypertension in 4,302 adults aged 18–69 years from the 2017 Zambia WHO STEPS survey. ROC analysis with Youden's Index identified optimal sex-stratified cut-offs for BMI, WC, WHtR, and WHpR. Five machine learning classifiers — logistic regression, linear SVM, random forest, gradient boosting, and XGBoost — were then trained on anthropometric, sociodemographic, and behavioural covariates using stratified 5-fold cross-validation, with recall prioritised as the primary metric given the imbalanced outcome.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults:\u003c/strong\u003e Hypertension prevalence was 23.7% (n=1,021), higher in men (25.4% vs. 22.7%; p=0.046) and urban than rural residents (27.0% vs. 21.7%; p\u0026lt;0.001). WC and WHtR showed the highest discrimination (men: AUC 0.63/0.62; women: 0.66/0.66); locally derived cut-offs were below WHO thresholds (WC: 80.35/81.85 cm; WHtR: 0.48/0.53; BMI: 21.97/24.12 kg/m²). Logistic regression and linear SVM achieved the best ROC-AUC (0.74; recall 0.69–0.70; F1=0.51); ensemble models showed higher accuracy (0.78–0.79) but lower recall (0.22–0.26).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion:\u003c/strong\u003e Central adiposity measures (WC and WHtR) outperformed BMI for hypertension identification in Zambian adults, with locally derived cut-offs consistently below WHO thresholds. Logistic regression and linear SVM demonstrated the best predictive performance, supporting their utility in risk stratification using routinely collected data.\u003c/p\u003e","manuscriptTitle":"Evaluating the Predictive Capacity of Obesity Indices for High Blood Pressure Among Zambians Aged 18–69 Years Using Machine Learning: Evidence From WHO STEPS Survey","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-04 11:45:08","doi":"10.21203/rs.3.rs-9585939/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"1e595f2f-d05c-42a0-a17b-781918c1aba3","owner":[],"postedDate":"May 4th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":67373381,"name":"Artificial Intelligence and Machine Learning"}],"tags":[],"updatedAt":"2026-05-04T11:45:08+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-04 11:45:08","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9585939","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9585939","identity":"rs-9585939","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}