Abstract
Background
Identifying young adults at high risk of hypertension can improve blood pressure screening recommendations.
Methods
We developed models to predict incident hypertension using diverse contemporary cohorts of young adults aged 18 to 39 years from Kaiser Permanente Southern California (derivation and internal validation) and 3 cohort studies (CARDIA [Coronary Artery Risk Development in Young Adults], FHS [Framingham Heart Study], HCHS/SOL [Hispanic Community Health Study/Study of Latinos]; external validation). Predictors included age, systolic and diastolic blood pressure, body mass index, smoking, social determinants of health, comorbidities, high‐ and low‐density lipoprotein cholesterol, and pregnancy‐related hypertensive disorders. We used Cox elastic net and random survival forests to develop sex‐specific models and compared their performance to 2021 US Preventive Services Task Force blood pressure screening recommendations.
Results
Among 355 524 adults from Kaiser Permanente Southern California (mean age, 29 years; mean systolic/diastolic blood pressure, 115/70 mm Hg), 11.7% developed hypertension in a median of 5.4 years. External validation showed good discrimination and calibration (Harrell's C‐statistic, 0.76 and 0.82; Integrated Brier Score, 0.04 and 0.02 for men and women, respectively). Compared with US Preventive Services Task Force, a 10‐year risk for hypertension of ≥15% in the external cohort using the new model showed similar sensitivity (men, 0.90 versus 0.89; women, 0.81 versus 0.81) and moderate improvement in specificity (men, 0.35 versus 0.25; women, 0.66 versus 0.44). Using the National Health and Nutrition Examination Survey, the prediction model estimated 28.5 million US young adults being at high risk of hypertension compared with 45.3 million by the US Preventive Services Task Force.
Conclusions
Compared with the US Preventive Services Task Force, the hypertension risk prediction model may be a more efficient tool to identify high‐risk young adults for early intervention.
Keywords: hypertension, prediction model, young adults
Subject Categories: Hypertension
Nonstandard Abbreviations and Acronyms
- CARDIA
Coronary Artery Risk Development in Young Adults
- FHS
Framingham Heart Study
- HCHS/SOL
Hispanic Community Health Study/Study of Latinos
- KPSC
Kaiser Permanente Southern California
- NHANES
National Health and Nutrition Examination Survey
- USPSTF
US Preventive Services Task Force
Clinical Perspective.
What Is New?
We developed a predictive model to calculate 10‐year hypertension risk among young adults.
What Are the Clinical Implications?
In this cohort of 355 524 young adults, a prediction model using age, systolic and diastolic blood pressure, body mass index, smoking, dyslipidemia, diabetes, high‐ and low‐density lipoprotein cholesterol, and hypertensive pregnancy disorders showed good discrimination and calibration in predicting hypertension over 10 years. Compared with US Preventive Services Task Force recommendations, the model shows similar sensitivity and improved specificity.
This prediction model may help clinicians identify young adults who may benefit from more frequent blood pressure screening and early interventions to prevent hypertension and premature cardiovascular events.
Overall reduction in the rate of cardiovascular disease over the past decades in the US has not extended to young adults, aged 18–39 years. 1 , 2 , 3 , 4 Increased prevalence of cardiovascular disease risk factors in young adults, including high blood pressure (BP), 5 , 6 , 7 may have contributed to increased cardiovascular disease rates in this population. About 20% of US young adults have hypertension defined by systolic BP ≥130 or diastolic BP ≥80 mm Hg or antihypertensive medication use. Hypertension at a young age contributes to early onset end‐organ damage and increased risk of premature and lifetime cardiovascular disease. 8 Despite this, about 37% of US young adult men have not visited a health care professional's office in the past 12 months 9 and >25% of US young adults with hypertension are unaware of their condition. 10
The 2021 US Preventive Services Task Force (USPSTF) considered young adults with systolic BP of 130 to 139 or diastolic BP of 85 to 89 mm Hg or with hypertension risk factors (eg, overweight, obesity, or Black race) as being at high risk of developing hypertension and recommended annual BP screening. 11 However, these recommendations were based on a small number of studies with short‐term follow‐up. 12 Also, important hypertension risk factors such as smoking or pregnancy‐related hypertensive disorders were not considered. 13 , 14 While hypertension prediction models have been proposed, 15 , 16 , 17 , 18 , 19 , 20 most were developed on the basis of middle‐aged and older adults, and none were for contemporary US young adults. Identifying young adults at high risk for developing hypertension can potentially guide clinicians to reach those who may benefit from more frequent BP screening and earlier lifestyle intervention.
To address this knowledge gap, we used data from multiethnic and racial contemporary cohorts to develop and validate models to predict incident hypertension (systolic BP ≥140 or diastolic BP ≥90 mm Hg) among US young adults. We compared the accuracy in predicting incident hypertension between the newly developed risk prediction models and 2021 USPSTF BP screening recommendations. Further, we estimated the number of US young adults who would be categorized as having a high risk of developing hypertension using the newly developed risk prediction models and 2021 USPSTF recommendations.
Methods
For Kaiser Permanente Southern California (KPSC), deidentified data that support the findings of this study may be made available from the corresponding author pursuant to a written request with appropriate approvals and assurances to maintain data in accordance with security requirements and documented evidence of human subject protections. Data from the pooled cohort cannot be shared publicly due to the privacy of individuals who participated in the study. Requests to access these data sets from qualified researchers can be sent directly to the coordinating center of each cohort. Limited versions of the data sets may also be available through the Biologic Specimen and Data Repository Information Coordinating Center.
Derivation Cohort
We developed a prediction model for incident hypertension among a contemporary, large multiethnic and racial, insured young adult population from KPSC (Figure 1). Using the electronic health records from KPSC, we included adults aged 18 to 39 years with ≥1 outpatient BP measurement in 2009 and followed them from their first BP measurement date (index) until death, membership disenrollment, incidence of hypertension (systolic BP ≥140 or diastolic BP ≥90 mm Hg), initiation of antihypertensive medication, or December 31, 2019, whichever occurred first (maximum 11 years of follow‐up). We excluded young adults who did not have ≥2 outpatient BP measurements on separate days anytime during follow‐up (including index) and those who had cardiovascular or kidney disease, untreated systolic BP ≥140 or diastolic BP ≥90 mm Hg, or a diagnosis of hypertension or used antihypertensive medication 3 years before or on the index date (Figure S1).
Figure 1. Overview of study sample.
CARDIA indicates Coronary Artery Risk Development in Young Adults; FHS, Framingham Heart Study; HCHS/SOL, Hispanic Community Health Study/Study of Latinos; and KPSC, Kaiser Permanente Southern California.
External Validation Cohort
We used a pooled cohort of 3 US epidemiologic studies to externally validate the prediction model: (1) CARDIA (Coronary Artery Risk Development in Young Adults), 21 (2) FHS (Framingham Heart Study; including the Offspring, Third Generation, Omni 1, and Omni 2 cohorts), 22 , 23 , 24 and (3) HCHS/SOL (Hispanic Community Health Study/Study of Latinos) (Figure 1). 25 The pooled epidemiologic cohort represents a racial, ethnically, and geographically diverse young adult population with clinical information collected in standardized study settings during regularly scheduled visits (Data S1). We included adults aged 18 to 39 years with ≥2 BP measures on separate visits and excluded participants with a history of cardiovascular or kidney disease, untreated systolic BP ≥140 or diastolic BP ≥90 mm Hg or antihypertensive medication use at baseline, or missing values for hypertension risk factors (Figure S2). We followed participants from their baseline visit until death, hypertension development (systolic BP ≥140 or diastolic BP ≥90 mm Hg), initiation of antihypertensive medication, or end of follow‐up (each cohort has a different follow‐up; see details in Data S1), whichever occurred first.
US Population Estimates
To project US national estimates, we used data from January 2017 to March 2020 from the US National Health and Nutrition Examination Survey (NHANES). 26 We included adults aged 18 to 39 years and excluded those with systolic BP ≥140 or diastolic BP ≥90 mm Hg or antihypertensive medication use, with a self‐reported history of cardiovascular or kidney disease, or with missing values for hypertension risk factors used in the prediction models (Figure S3).
Ethical Statement
All study protocols were approved by the Institutional Review Boards at participating institutions. All participants in the pooled cohort provided written informed consent. In the KPSC cohort, informed consent was waived given the retrospective nature of the study.
Candidate Variables for Hypertension Prediction and Outcomes
In the KPSC derivation cohort, we considered candidate variables in the prediction model that have been associated with hypertension development (see Data S1 for a list of candidate variables). The candidate variables from pooled cohort data sets were harmonized with KPSC variables (Table S1). In KPSC, we defined the outcome of incident hypertension as the average of 2 outpatient untreated BP values from 2 separate days meeting the criteria (systolic BP ≥140 or diastolic BP ≥90 mm Hg) following major hypertension guidelines. 27 , 28 , 29 The maximum duration between the 2 BP measurements on separate days was 2 years, with the later date meeting criteria as the date of incident hypertension (Data S1). In the pooled cohort, we defined incident hypertension as the first visit during follow‐up when untreated systolic BP was ≥140 or diastolic BP was ≥90 mm Hg.
Statistical Analysis
Prediction Model Development
We followed the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis checklist. 30 We built separate models for men and women to address sex differences in BP determinants. 31 , 32 We randomly split KPSC data into 80% derivation and 20% test data set. To avoid potential overfitting, we used 10‐fold cross‐validation to develop and compare models using the training set. We used (1) Cox proportional hazards regression with elastic net 33 and (2) random survival forests for model building. 34 See Data S1 for a description of the model tuning process.
Internal and External Validation
The final models were internally validated in the KPSC test data set, then externally validated in the pooled cohort. We assessed model performance using Harrell's C‐statistic 35 for discrimination and Integrated Brier Score 36 for calibration. Both measures evaluate model performance across the study time frame, equally prioritizing all time points, including 3‐, 5‐, or 10‐year outcomes. We also estimated 10‐year predicted risk of hypertension by patient risk profiles used in USPSTF recommendations. 37
Comparison of Model Performance Based on the Risk Prediction Models Versus USPSTF Recommendations
We compared the accuracy of detecting true incident hypertension by calculating sensitivity, specificity, positive predictive values, and negative predictive values accounting for censoring 38 using the final risk prediction model and 2021 USPSTF BP screening recommendations. For the risk prediction model‐based approach, we considered young adults with a predicted 10‐year hypertension risk above various prespecified thresholds (ie, 7.5%, 10%, 15%, and 20%) as at high risk of developing hypertension on the basis of the distribution of 10‐year risk score for men and women in the training data. For 2021 USPSTF BP screening recommendations, we considered young adults with BP of 130 to 139/85 to 89 mm Hg, overweight, obesity, or Black race as at high risk of hypertension. We evaluated the same performance metrics stratified by body mass index and self‐reported race and ethnicity (non‐Hispanic Asian/Pacific Islander, non‐Hispanic Black, non‐Hispanic White, and Hispanic).
Potential US Young Adult Population at High Risk of Hypertension
We estimated proportion and number of US young adults aged 18 to 39 years who would be categorized as being at high risk of hypertension according to the new prediction models and 2021 USPSTF BP screening recommendations using data from NHANES. All analyses were weighted to account for NHANES's complex sampling design.
Dr An had full access to all the KPSC data in the study and takes responsibility for the integrity of those data and accuracy of the analysis of KPSC data. Dr. Zhang had full access to all the pooled cohort data in the study and takes responsibility for the integrity of those data and accuracy of the analysis of the pooled cohort data.
Results
Study Cohort, Baseline Characteristics, and Outcomes
We included 355 524 young adults from KPSC and 9518 from the pooled cohort (4339 from CARDIA, 2682 from FHS, 2497 from HCHS/SOL; Table 1, Table S2). Mean±SD age was 29.1±6.5 and 27.6±5.9 years for KPSC and the pooled cohort, respectively, and mean±SD systolic/diastolic BP was 115±10/70±8 and 111±11/69±8 mm Hg for KPSC and the pooled cohort, respectively (Table 1). During median (interquartile range) follow‐up of 5.4 (2.4–10.3) years for KPSC and 11.0 (6.4–11.0) years for the pooled cohort, the percentage of young adults who developed incident hypertension was 11.7% for KPSC and 5.4% for the pooled cohort. From the NHANES data, we included 1857 young adults (Table S3).
Table 1.
Baseline Characteristics and Hypertension Outcomes of Young Adults From Development and External Validation Cohorts
| Development cohort (KPSC) | External validation cohort (pooled cohort) | |||||
|---|---|---|---|---|---|---|
| Overall | Men | Women | Overall | Men | Women | |
| No. | 355 524 | 123 783 | 231 741 | 9518 | 4737 | 4781 |
| Age, y, mean±SD | 29.1±6.5 | 29.1±6.6 | 29.0±6.4 | 27.6±5.9 | 27.7±5.8 | 27.6±5.9 |
| Race and ethnicity, n (%) | ||||||
| Non‐Hispanic White | 103 048 (29) | 38 995 (31.5) | 64 053 (27.6) | 4681 (49.2) | 2705 (57.1) | 1976 (41.3) |
| Non‐Hispanic Black | 29 712 (8.4) | 8815 (7.1) | 20 897 (9.0) | 2232 (23.5) | 1005 (21.2) | 1227 (25.7) |
| Hispanic | 159 425 (44.8) | 52 788 (42.6) | 106 637 (46.0) | 2562 (26.9) | 999 (21.1) | 1563 (32.7) |
| Non‐Hispanic Asian/Pacific Islander | 34 041 (9.6) | 8069 (6.5) | 9396 (4.1) | 30 (0.3) | 20 (0.4) | 10 (0.2) |
| Other | 7847 (0.2) | 12 335 (10.0) | 24 286 (10.5) | 8 (0.1) | 6 (0.1) | 2 (0.0) |
| Missing | 17 465 (4.9) | 2781 (2.2) | 6472 (2.8) | 5 (0.1) | 2 (0.0) | 3 (0.1) |
| Blood pressure, mm Hg, mean±SD | ||||||
| Systolic | 115±10 | 121±9 | 113±10 | 111±11 | 116±10 | 106±9 |
| Diastolic | 70±8 | 71±8 | 69±8 | 70±9 | 72±9 | 67±8 |
| Systolic and diastolic blood pressure, category, n (%) | ||||||
| <120 and <80 mm Hg | 226 920 (63.8) | 53 816 (43.5) | 173 104 (74.7) | 6792 (71.4) | 2633 (55.6) | 4159 (87.0) |
| 120–129 and <80 mm Hg | 73 972 (20.8) | 38 968 (31.5) | 35 004 (15.1) | 1055 (11.1) | 812 (17.1) | 243 (5.1) |
| 130–139 or 80–89 mm Hg | 54 632 (15.4) | 30 999 (25.0) | 23 633 (10.2) | 1671 (17.6) | 1292 (27.3) | 379 (7.9) |
| Smoking status, n (%) | ||||||
| Never | 279 552 (78.6) | 88 023 (71.1) | 191 529 (82.6) | 5731 (60.2) | 2575 (54.4) | 3156 (66.0) |
| Former | 43 282 (12.2) | 17 899 (14.5) | 25 383 (11.0) | 1317 (13.8) | 665 (14.0) | 652 (13.6) |
| Current | 32 331 (9.1) | 17 598 (14.2) | 14 733 (6.4) | 2470 (26.0) | 1497 (31.6) | 973 (20.4) |
| Missing | 359 (0.1) | 263 (0.2) | 96 (0.0) | 0 | 0 | 0 |
| Body mass index, kg/m2, mean±SD | 27.3±6.1 | 27.9±5.5 | 26.9±6.4 | 25.9±5.5 | 25.8±4.6 | 25.9±6.3 |
| Body mass index, category, n (%) | ||||||
| Underweight (<18.5) | 7766 (2.2) | 6389 (2.8) | 7766 (2.2) | 0 | 0 | 0 |
| Normal (18.5 to <25) | 138 183 (38.9) | 100 611 (43.4) | 138 183 (38.9) | 4917 (51.7) | 2295 (48.4) | 2622 (54.8) |
| Overweight (25 to <30) | 111 177 (31.3) | 63 965 (27.6) | 111 177 (31.3) | 2864 (30.1) | 1712 (36.1) | 1152 (24.1) |
| Obesity (≥30) | 95 822 (27) | 59 794 (25.8) | 95 822 (27.0) | 1737 (18.2) | 730 (15.4) | 1007 (21.1) |
| Missing | 2576 (0.7) | 982 (0.4) | 2576 (0.7) | 0 | 0 | 0 |
| Total cholesterol, mg/dL, mean±SD | 181.0±35.5 | 186.5±39.2 | 178.4±33.3 | 179.7±35.3 | 182.4±37.0 | 177.1±33.4 |
| Missing, n (%) | 137 424 (38.7) | 52 189 (42.2) | 85 235 (36.8) | 68 (0.7) | 29 (0.6) | 39 (0.8) |
| HDL‐C, mg/dL, mean±SD | 51.9±13.2 | 45.4±10.7 | 55.1±13.2 | 51.2±13.4 | 47.4±12.3 | 55.0±13.3 |
| Missing, n (%) | 140 892 (39.6) | 53 493 (43.2) | 87 399 (37.7) | 68 (0.7) | 29 (0.6) | 39 (0.8) |
| LDL‐C, mg/dL, mean±SD | 108.9±30.3 | 116.3±32.6 | 105.2±28.4 | 110.5±32.3 | 115.3±33.6 | 105.8±30.2 |
| Missing, n (%) | 156 886 (44.1) | 58 115 (46.9) | 98 771 (42.6) | 128 (1.3) | 77 (1.6) | 51 (1.1) |
| Dyslipidemia diagnosis, n (%) | 12 261 (3.4) | 6641 (5.4) | 5620 (2.4) | 892 (9.4) | 589 (12.4) | 303 (6.3) |
| Diabetes, n (%) | 6334 (1.8) | 1543 (1.2) | 4791 (2.1) | 131 (1.4) | 58 (1.2) | 73 (1.5) |
| Prediabetes, n (%) | 23 564 (6.6) | 10 502 (8.5) | 13 062 (5.6) | 1211 (12.7) | 1012 (21.4) | 199 (4.2) |
| AHRQ SES index, mean±SD | 56.5±2.4 | 56.5±2.5 | 56.4±2.4 | NA | NA | NA |
| Missing, n (%) | 24 613 (0.1) | 8384 (6.8) | 16 229 (7.0) | NA | NA | NA |
| Among women only, n (%) | ||||||
| Pregnancy | 20 316 (5.7) | NA | 20 316 (5.7) | 6 (0.1) | NA | 6 (0.1) |
| Hypertensive disorders of pregnancy | 1198 (0.5) | NA | 1198 (0.5) | 352 (7.4) | NA | 352 (7.4) |
| Gestational diabetes | 3006 (0.8) | NA | 3006 (0.8) | 104 (2.2) | NA | 104 (2.2) |
| Median (IQR) follow‐up, y | 5.4 (2.4–10.3) | 5.4 (2.4–10.2) | 5.5 (2.4–10.3) | 11.0 (6.4–11.0) | 11.0 (6.9–11.0) | 11.0 (6.1–11.0) |
| Incident hypertension (≥140/≥90 mm Hg) during follow‐up, n (%) | 41 567 (11.7) | 19 189 (16.0) | 22 378 (9.7) | 512 (5.4) | 336 (7.1) | 176 (3.7) |
AHRQ indicates Agency for Healthcare Research and Quality; HDL‐C, high‐density lipoprotein cholesterol; IQR, interquartile range; LDL‐C, low‐density lipoprotein cholesterol; NA, not applicable; and SES, socioeconomic status. “Other” Specifies Multi‐race, Native American, Alaskan, Self‐reported other.
Derivation of Prediction Models
The final prediction model included variables selected in either overall or lipid stratified models across sex, and excluded variables consistently not selected as predictors (Table S4). Predictors for the final prediction models are age (natural log), age squared (natural log), body mass index category, systolic BP (natural log), diastolic BP (natural log), current smoking, history of smoking, dyslipidemia diagnosis, diabetes, low‐density lipoprotein cholesterol (natural log and missing indicator), high‐density lipoprotein cholesterol (natural log and missing indicator), and hypertensive disorders of pregnancy (Tables S5 through S7). Applying the final prediction model to KPSC test data, the overall 10‐year risk for hypertension was 22.7% for men and 15.4% for women (Table S8). Baseline BP measures contributed the largest share to the estimated 10‐year risk for hypertension, with a predicted risk estimated to be 10.3% and 9.3% for men and women with normal baseline BP (110/65 mm Hg), respectively, compared with a 10‐year risk of 41.1% and 44.0% for men and women with prehypertension at baseline (130/85 mm Hg), while holding all other risk factors constant (Table S8). The estimated 10‐year risk for hypertension was 27.0% for men with obesity and 18.9% for women with obesity, compared with 19.6% for normal‐weight men and 12.9% for normal‐weight women. The estimated 10‐year risk for hypertension was 22.0% for non‐Hispanic Black men, and 17.5% for non‐Hispanic Black women.
Internal and External Validation
In internal validation using the KPSC test data set, Harrell's C‐statistics were 0.75 for men and 0.82 for women, and Integrated Brier Scores were 0.10 for men and 0.06 for women (Table 2). External validation showed that Harrell's C‐statistics were 0.76 for men and 0.82 for women, and Integrated Brier Score were 0.04 for men and 0.02 for women.
Table 2.
Discrimination and Calibration of the Hypertension Risk Prediction Models Stratified by Race and Ethnicity
| Men | Women | |||
|---|---|---|---|---|
| Internal validation | Harrell's C‐statistic | Overall | 0.75 | 0.82 |
| Non‐Hispanic Asian/Pacific Islander | 0.75 | 0.85 | ||
| Non‐Hispanic Black | 0.72 | 0.81 | ||
| Hispanic | 0.75 | 0.82 | ||
| Non‐Hispanic White | 0.74 | 0.82 | ||
| Integrated Brier Score | Overall | 0.10 | 0.06 | |
| Non‐Hispanic Asian/Pacific Islander | 0.09 | 0.05 | ||
| Non‐Hispanic Black | 0.12 | 0.09 | ||
| Hispanic | 0.09 | 0.06 | ||
| Non‐Hispanic White | 0.11 | 0.06 | ||
| External validation | Harrell's C‐statistic | Overall | 0.76 | 0.82 |
| Non‐Hispanic Black | 0.80 | 0.79 | ||
| Hispanic | 0.79 | 0.82 | ||
| Non‐Hispanic White | 0.76 | 0.88 | ||
| Integrated Brier Score | Overall | 0.04 | 0.02 | |
| Non‐Hispanic Black | 0.04 | 0.03 | ||
| Hispanic | 0.06 | 0.04 | ||
| Non‐Hispanic White | 0.04 | 0.01 |
Comparison of Model Performance Based on the Risk Prediction Models Versus USPSTF Recommendations
We evaluated risk prediction thresholds using the internal validation data set. Using 10‐year hypertension risk thresholds of 7.5%, 10%, 15%, and 20% based on the newly developed risk prediction model categorized 91.0%, 83.4%, 66.5%, and 51.0% of men and 67.4%, 55.5%, 37.9%, and 26.2% of women as being at high risk of hypertension (Table 3). Using 2021 USPSTF BP screening recommendations, 73.7% of men and 58.1% of women were categorized as being at high risk of hypertension. Compared with USPSTF BP screening recommendations, using a 10‐year risk threshold of 15% showed similar sensitivity (men, 0.90 versus 0.89; women, 0.81 versus 0.81) and improved specificity (men, 0.35 versus 0.25; women, 0.66 versus 0.44), positive predictive values (men, 0.29 versus 0.26; women, 0.30 versus 0.21), and negative predictive values (men, 0.92 versus 0.88; women, 0.95 versus 0.93) while reducing the number of young adults categorized as being high risk from 73.7% to 66.5% in men and from 58.1% to 37.9% in women.
Table 3.
Percentage of High Risk, Sensitivity, Specificity, PPV, and NPV at 10 Years According to the Prediction Model Versus USPSTF Recommendations in the KPSC Cohort
| Prediction model | High risk, % | Sensitivity | Specificity | PPV | NPV |
|---|---|---|---|---|---|
| Men | |||||
| 10‐y 7.5% | 91.0 | 0.99 | 0.09 | 0.24 | 0.96 |
| 10‐y 10% | 83.4 | 0.96 | 0.17 | 0.25 | 0.94 |
| 10‐y 15% | 66.5 | 0.90 | 0.35 | 0.29 | 0.92 |
| 10‐y 20% | 51.0 | 0.79 | 0.53 | 0.33 | 0.89 |
| USPSTF | 73.7 | 0.89 | 0.25 | 0.26 | 0.88 |
| Women | |||||
| 10‐y 7.5% | 67.4 | 0.95 | 0.33 | 0.21 | 0.98 |
| 10‐y 10% | 55.5 | 0.91 | 0.46 | 0.24 | 0.97 |
| 10‐y 15% | 37.9 | 0.81 | 0.66 | 0.30 | 0.95 |
| 10‐y 20% | 26.2 | 0.67 | 0.78 | 0.36 | 0.93 |
| USPSTF | 58.1 | 0.81 | 0.44 | 0.21 | 0.93 |
KPSC indicates Kaiser Permanente Southern California; NPV, negative predictive value; PPV, positive predictive value; and USPSTF, The US Preventive Services Task Force.
In subgroup analyses of normal‐weight individuals at the 10‐year risk cutoff of 15%, sensitivity increased from 0.33 to 0.72 for men and from 0.20 to 0.63 for women, while the percentage of high‐risk population increased from 17.0% to 44.4% for men and 9.1% to 19.2% for women compared with USPSTF recommendations (Figure 2, Table S9). Among overweight individuals, the specificity increased from 0 to 0.38 for men and from 0 to 0.66 for women, while the percentage of high‐risk population decreased from 100% to 66.9% for men and from 100% to 38.1% for women. Among young adults with obesity, the sensitivity decreased by up to 0.08 while the specificity increased by up to 0.33 and the percentage of high risk decreased by up to 28.5%. For non‐Hispanic Black young adults, the sensitivity decreased by up to 0.16, while the specificity increased by up to 0.58 and the percentage of high risk decreased by up to 55.0%. Across different racial and ethnic groups other than non‐Hispanic Black young adults, the prediction model showed improved or similar sensitivity and specificity using a 10‐year risk cutoff of 15%.
Figure 2. High risk, sensitivity, specificity using the prediction model vs USPSTF by subgroups, KPSC cohort.
A, Normal‐weight men, (B) normal‐weight women, (C) overweight men, (D) overweight women, (E) men with obesity, (F) women with obesity, (G) Asian and Pacific Islander men, (H) Asian and Pacific Islander women, (I) Black men, (J) Black women, (K) Hispanic men, (L) Hispanic women, (M) White men, and (N) White women. KPSC indicates Kaiser Permanente Southern California; PI, Pacific Islander; and USPSTF, US Preventive Services Task Force.
US Young Adult Population at High Risk of Hypertension
The new prediction model with a 10‐year risk cutoff of ≥15% estimated 43.6% (28.5 million [95% CI, 25.4–31.6 million]) of US young adults without a history of cardiovascular or kidney disease were categorized as high risk for hypertension, while 2021 USPSTF recommendations estimated 69.3% (45.3 million [95% CI, 41.9, 48.7 million]) having high risk. Additionally, 5.7% (3.8 million [95% CI, 3.7–3.8 million]) were categorized as high risk of hypertension by the new prediction model but not by 2021 USPSTF recommendations, and 37.9% (24.7 million [95% CI, 21.7–27.8 million]) were categorized as high risk by both the new prediction model and USPSTF recommendations.
Discussion
We developed and validated risk models to predict the development of incident hypertension among young adults to inform BP screening and earlier intervention strategies. The new prediction model showed good discrimination and calibration in men and women and across racial and ethnic groups. Compared with 2021 USPSTF recommendations, the new prediction model showed similar sensitivity and moderate improvement in specificity, with greater improvement in sensitivity and specificity among normal‐weight and overweight young adults. The new prediction model reduced the percentage of young adults being categorized as having a high risk of developing hypertension compared with USPSTF recommendations. These findings suggest that the new prediction model may be more efficient than 2021 USPSTF recommendations at identifying young adults at high risk for developing hypertension.
The 2021 USPSTF guidelines recommended annual BP screening for young adults with systolic/diastolic BP 130 to 139/85 to 89 mm Hg or who are overweight, obese, and of Black race. 39 However, hypertension risk factors such as smoking and pregnancy‐related hypertensive disorders were not considered in USPSTF guidelines. Unlike USPSTF recommendations, the prediction model accounts for hypertension risk factors concurrently; the effect of a given predictor on hypertension risk is nonlinear and dependent on other risk predictors. Therefore, although estimated hazard ratios for systolic and diastolic BP were highest in magnitude compared with other predictors, risk predictors with smaller relative hazard ratios may have an important effect on hypertension risk. This is evident, for example, in the final model estimates showing a greater increase in 10‐year hypertension risk for a person who is obese versus normal weight given a baseline BP measurement at the prehypertension level (eg, 130/85 mm Hg, which shows a 12.2% risk increase for men and 14.4% for women), compared with a person with a normal baseline BP (eg, 120/65 mm Hg, which shows a 5.7% risk increase for men and 6.0% increase for women; Table S8).
Compared with 2021 USPSTF BP screening recommendations, the risk prediction model using 10‐year risk cutoff ≥15% showed the greatest improvement in sensitivity among normal‐weight individuals (∆0.39 for men and ∆0.43 for women). Using NHANES data, we estimated 5.7% (3.8 million) US young adults were not considered as high risk by USPSTF recommendations but were categorized as high risk using the risk prediction model. While the risk prediction model requires more clinical information (age, smoking, diabetes, lipids, pregnancy‐related hypertensive disorders), identifying individuals overlooked by current USPSTF recommendations may be important. Moreover, the risk prediction model reduced the number of young adults who were categorized as being at high risk of developing hypertension. The estimated US young adults were reduced from 69.3% (45.3 million) on the basis of USPSTF recommendations to 43.6% (28.5 million) on the basis of the risk prediction model. Given that young adults have unique challenges in health care access, 40 the newly developed risk model could be effective in directing an intervention toward young adults with the highest risk of hypertension by providing more frequent BP screening and early lifestyle intervention.
Improvement in sensitivity using the risk prediction model versus 2021 USPSTF recommendations was more prominent among racial and ethnic minority groups other than non‐Hispanic Black young adults. The 2021 USPSTF recommendations recognize Black populations as at‐risk but largely ignore racial and ethnic minority groups that experience hypertension disparities. The new prediction tool may help identify at‐risk Hispanic and Asian Americans who have lower BP control rates and less awareness of hypertension compared with White Americans. 41 All Black individuals were considered high risk according to 2021 USPSTF recommendations; therefore, sensitivity could not improve when applying the newly developed risk prediction model. However, the new risk prediction model improved specificity and reduced the number of non‐Hispanic Black individuals categorized as having high risk.
The main strengths of the study are development of a new hypertension risk prediction model among an ethnically and racially diverse young adult population, and demonstration of the benefits of this risk prediction tool to identify high‐risk young adults for more frequent BP screening and early interventions. However, this study has several limitations. Although we explored an extensive list of hypertension risk factors in the prediction model development, diet, physical activity, alcohol consumption, and family history of hypertension or cardiovascular disease data were unavailable. A self‐reported exercise variable was tested but dropped due to large missingness and lack of intensity information. Moreover, electronic health records data on healthier young adults who are less likely to interact with the health care system may not be reflected. BP measures were also dependent on data availability at KPSC, and automated BP measurement machines have changed over time. BP measurement protocols were inconsistent between and even within the cohorts over time. All BP measurements were from office or clinic visits, which includes white‐coat hypertension or may have missed masked hypertension. The results on sensitivity, specificity, positive predictive value, and negative predictive value applying the prediction model are likely optimistic, as analysis was conducted in the same data set used to develop the risk prediction model. We chose to perform these analyses on the internal validation data set because it reflects a more contemporary rate of hypertension and data collection in electronic health records mirror other modern health care systems that may implement this risk equation. It will be valuable to validate these results in other contemporary US populations using electronic health records.
Conclusions
Using data from a large diverse contemporary cohort of young adults, we developed and validated risk models to predict incident hypertension in young adults, which showed good discrimination and calibration in internal and external validation. Compared with current USPSTF recommendations, the new prediction model may be more efficient in identifying high‐risk young adults who may benefit from more frequent BP screening and earlier interventions to prevent hypertension.
Sources of Funding
This work was supported by NIH R01HL158790 (Drs An and Zhang). The funder/sponsor had no role in the design and conduct of the study; the collection, management, analysis, and interpretation of the data; the preparation, review, or approval of the manuscript; or the decision to submit the manuscript for publication. The CARDIA study is conducted and supported by the National Heart, Lung, and Blood Institute in collaboration with the University of Alabama at Birmingham (HHSN268201800005I and HHSN268201800007I), Northwestern University (HHSN268201800003I), University of Minnesota (HHSN268201800006I), and Kaiser Foundation Research Institute (HHSN268201800004I). The FHS was supported by the National Heart, Lung, and Blood Institute (HHSN268201500001I) and the Boston University Chobanian and Avedisian School of Medicine. The HCHS/SOL is a collaborative study supported by contracts from the National Heart, Lung, and Blood Institute to the University of North Carolina (HHSN268201300001I/N01‐HC‐65233), University of Miami (HHSN268201300004I/N01‐HC‐65234), Albert Einstein College of Medicine (HHSN268201300002I/N01‐HC‐65235), University of Illinois at Chicago (HHSN268201300003I/N01‐HC‐65236 Northwestern University), and San Diego State University (HHSN268201300005I/N01‐HC‐65237). The following Institutes/Centers/Offices have contributed to the HCHS/SOL through a transfer of funds to the the National Heart, Lung, and Blood Institute: National Institute on Minority Health and Health Disparities, National Institute on Deafness and Other Communication Disorders, National Institute of Dental and Craniofacial Research, National Institute of Diabetes and Digestive and Kidney Diseases, National Institute of Neurological Disorders and Stroke, National Institutes of Health Institution Office of Dietary Supplements.
Disclosures
Dr An and Ms Choi receive research support from Bayer AG unrelated to this work. Dr Reynolds receives research support from Merck Sharp & Dohme LLC unrelated to this work. Dr Colantonio receives research support from Amgen unrelated to this work. The remaining authors have no disclosures to report.
Supporting information
Data S1
Tables S1–S9
Figures S1–S3
Acknowledgments
The authors thank the investigators, staff, and participants of all the cohorts for their valuable contributions.
This manuscript was sent to Ajay K. Gupta, MD, MSc, PhD, FRCP, FESC, Senior Associate Editor, for review by expert referees, editorial decision, and final disposition.
Supplemental Material is available at https://www.ahajournals.org/doi/suppl/10.1161/JAHA.124.040769
For Sources of Funding and Disclosures, see page 10.
References
- 1. Wilmot KA, O'Flaherty M, Capewell S, Ford ES, Vaccarino V. Coronary heart disease mortality declines in the United States from 1979 through 2011: evidence for stagnation in young adults, especially women. Circulation. 2015;132:997–1002. doi: 10.1161/circulationaha.115.015293 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Gupta A, Wang Y, Spertus JA, Geda M, Lorenze N, Nkonde‐Price C, D'Onofrio G, Lichtman JH, Krumholz HM. Trends in acute myocardial infarction in young patients and differences by sex and race, 2001 to 2010. J Am Coll Cardiol. 2014;64:337–345. doi: 10.1016/j.jacc.2014.04.054 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Arora S, Stouffer GA, Kucharska‐Newton AM, Qamar A, Vaduganathan M, Pandey A, Porterfield D, Blankstein R, Rosamond WD, Bhatt DL, et al. Twenty year trends and sex differences in young adults hospitalized with acute myocardial infarction. Circulation. 2019;139:1047–1056. doi: 10.1161/circulationaha.118.037137 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Ekker MS, Verhoeven JI, Vaartjes I, van Nieuwenhuizen KM, Klijn CJM, de Leeuw FE. Stroke incidence in young adults according to age, subtype, sex, and time trends. Neurology. 2019;92:e2444–e2454. doi: 10.1212/wnl.0000000000007533 [DOI] [PubMed] [Google Scholar]
- 5. Vikulova DN, Grubisic M, Zhao Y, Lynch K, Humphries KH, Pimstone SN, Brunham LR. Premature atherosclerotic cardiovascular disease: trends in incidence, risk factors, and sex‐related differences, 2000 to 2016. J Am Heart Assoc. 2019;8:e012178. doi: 10.1161/jaha.119.012178 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Mayer‐Davis EJ, Lawrence JM, Dabelea D, Divers J, Isom S, Dolan L, Imperatore G, Linder B, Marcovina S, Pettitt DJ, et al. Incidence trends of type 1 and type 2 diabetes among youths, 2002–2012. N Engl J Med. 2017;376:1419–1429. doi: 10.1056/NEJMoa1610187 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Aigner A, Grittner U, Rolfs A, Norrving B, Siegerink B, Busch MA. Contribution of established stroke risk factors to the burden of stroke in young adults. Stroke. 2017;48:1744–1751. doi: 10.1161/strokeaha.117.016599 [DOI] [PubMed] [Google Scholar]
- 8. Bibbins‐Domingo K, Pletcher MJ, Lin F, Vittinghoff E, Gardin JM, Arynchyn A, Lewis CE, Williams OD, Hulley SB. Racial differences in incident heart failure among young adults. N Engl J Med. 2009;360:1179–1190. doi: 10.1056/NEJMoa0807265 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. QuickStats: Percentage* of adults aged≥ 18 years who have seen or talked to a doctor or other health care professional about their own health in the past 12 months, by sex and age group—National Health Interview Survey, United States, 2015. MMWR Morb Mortal Wkly Rep. 2017;66:65. doi: 10.15585/mmwr.mm6602a12 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Zhang Y, Moran AE. Trends in the prevalence, awareness, treatment, and control of hypertension among young adults in the United States, 1999 to 2014. Hypertension. 2017;70:736–742. doi: 10.1161/hypertensionaha.117.09801 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. US Preventive Services Task Force , AH K, KW D, CM M, M C, AB C, EM D, KE D, CA D, M K, et al. Screening for hypertension in adults: US preventive services task force reaffirmation recommendation statement. JAMA. 2021;325:1650–1656. doi: 10.1001/jama.2021.4987 [DOI] [PubMed] [Google Scholar]
- 12. Guirguis‐Blake JM, Evans CV, Webber EM, Coppola EL, Perdue LA, Weyrich MS. Screening for Hypertension in Adults: An Updated Systematic Evidence Review for the US Preventive Services Task Force . 2021. U.S. Preventive Services Task Force Evidence Syntheses, formerly Systematic Evidence Reviews. [PubMed]
- 13. van Oort S, Beulens JWJ, van Ballegooijen AJ, Grobbee DE, Larsson SC. Association of cardiovascular risk factors and lifestyle behaviors with hypertension: a Mendelian randomization study. Hypertension. 2020;76:1971–1979. doi: 10.1161/HYPERTENSIONAHA.120.15761 [DOI] [PubMed] [Google Scholar]
- 14. Brown MC, Best KE, Pearce MS, Waugh J, Robson SC, Bell R. Cardiovascular disease risk in women with pre‐eclampsia: systematic review and meta‐analysis. Eur J Epidemiol. 2013;28:1–19. doi: 10.1007/s10654-013-9762-6 [DOI] [PubMed] [Google Scholar]
- 15. Kanegae H, Suzuki K, Fukatani K, Ito T, Harada N, Kario K. Highly precise risk prediction model for new‐onset hypertension using artificial intelligence techniques. J Clin Hypertens (Greenwich). 2020;22:445–450. doi: 10.1111/jch.13759 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Chowdhury MZI, Leung AA, Sikdar KC, O'Beirne M, Quan H, Turin TC. Development and validation of a hypertension risk prediction model and construction of a risk score in a Canadian population. Sci Rep. 2022;12:12780. doi: 10.1038/s41598-022-16904-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Pearson TA, LaCroix AZ, Mead LA, Liang KY. The prediction of midlife coronary heart disease and hypertension in young adults: the Johns Hopkins multiple risk equations. Am J Prev Med. 1990;6:23–28. doi: 10.1016/S0749-3797(19)30122-9 [DOI] [PubMed] [Google Scholar]
- 18. Echouffo‐Tcheugui JB, Batty GD, Kivimaki M, Kengne AP. Risk models to predict hypertension: a systematic review. PLoS One. 2013;8:e67370. doi: 10.1371/journal.pone.0067370 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Parikh NI, Pencina MJ, Wang TJ, Benjamin EJ, Lanier KJ, Levy D, D'Agostino RB Sr, Kannel WB, Vasan RS. A risk score for predicting near‐term incidence of hypertension: the Framingham heart study. Ann Intern Med. 2008;148:102–110. doi: 10.7326/0003-4819-148-2-200801150-00005 [DOI] [PubMed] [Google Scholar]
- 20. Carson AP, Lewis CE, Jacobs DR Jr, Peralta CA, Steffen LM, Bower JK, Person SD, Muntner P. Evaluating the Framingham hypertension risk prediction model in young adults: the coronary artery risk development in young adults (CARDIA) study. Hypertension. 2013;62:1015–1020. doi: 10.1161/HYPERTENSIONAHA.113.01539 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Friedman GD, Cutter GR, Donahue RP, Hughes GH, Hulley SB, Jacobs DR Jr, Liu K, Savage PJ. CARDIA: study design, recruitment, and some characteristics of the examined subjects. J Clin Epidemiol. 1988;41:1105–1116. doi: 10.1016/0895-4356(88)90080-7 [DOI] [PubMed] [Google Scholar]
- 22. Feinleib M, Kannel WB, Garrison RJ, McNamara PM, Castelli WP. The Framingham offspring study. Design and preliminary data. Prev Med. 1975;4:518–525. doi: 10.1016/0091-7435(75)90037-7 [DOI] [PubMed] [Google Scholar]
- 23. Splansky GL, Corey D, Yang Q, Atwood LD, Cupples LA, Benjamin EJ, D'Agostino RB Sr, Fox CS, Larson MG, Murabito JM, et al. The third generation cohort of the National Heart, Lung, and Blood Institute's Framingham heart study: design, recruitment, and initial examination. Am J Epidemiol. 2007;165:1328–1335. doi: 10.1093/aje/kwm021 [DOI] [PubMed] [Google Scholar]
- 24. Tsao CW, Vasan RS. Cohort profile: the Framingham heart study (FHS): overview of milestones in cardiovascular epidemiology. Int J Epidemiol. 2015;44:1800–1813. doi: 10.1093/ije/dyv337 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Sorlie PD, Avilés‐Santa LM, Wassertheil‐Smoller S, Kaplan RC, Daviglus ML, Giachello AL, Schneiderman N, Raij L, Talavera G, Allison M, et al. Design and implementation of the Hispanic community health study/study of Latinos. Ann Epidemiol. 2010;20:629–641. doi: 10.1016/j.annepidem.2010.03.015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. US Centers for Disease Control and Prevention National Health and Nutrition Examination Survey . Accessed April 8, 2024. https://www.cdc.gov/nchs/nhanes/index.htm.
- 27. Unger T, Borghi C, Charchar F, Khan NA, Poulter NR, Prabhakaran D, Ramirez A, Schlaich M, Stergiou GS, Tomaszewski M, et al. 2020 International Society of Hypertension Global Hypertension Practice Guidelines. Hypertension. 2020;75:1334–1357. doi: 10.1161/HYPERTENSIONAHA.120.15026 [DOI] [PubMed] [Google Scholar]
- 28. Whelton PK, Carey RM, Aronow WS, Casey DE Jr, Collins KJ, Dennison Himmelfarb C, DePalma SM, Gidding S, Jamerson KA, Jones DW, et al. 2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA guideline for the prevention, detection, evaluation, and Management of High Blood Pressure in adults: a report of the American College of Cardiology/American Heart Association task force on clinical practice guidelines. Hypertension. 2018;71:e13–e115. doi: 10.1161/HYP.0000000000000065 [DOI] [PubMed] [Google Scholar]
- 29. Williams B, Mancia G, Spiering W, Agabiti Rosei E, Azizi M, Burnier M, Clement DL, Coca A, de Simone G, Dominiczak A, et al. 2018 ESC/ESH guidelines for the management of arterial hypertension: the task force for the management of arterial hypertension of the European Society of Cardiology and the European Society of Hypertension: the task force for the management of arterial hypertension of the European Society of Cardiology and the European Society of Hypertension. J Hypertens. 2018;36:1953–2041. doi: 10.1097/HJH.0000000000001940 [DOI] [PubMed] [Google Scholar]
- 30. Collins GS, Reitsma JB, Altman DG, Moons KG. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement. BMJ. 2015;350:g7594. doi: 10.1136/bmj.g7594 [DOI] [PubMed] [Google Scholar]
- 31. Ji H, Kim A, Ebinger JE, Niiranen TJ, Claggett BL, Bairey Merz CN, Cheng S. Sex differences in blood pressure trajectories over the life course. JAMA Cardiol. 2020;5:19–26. doi: 10.1001/jamacardio.2019.5306 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Connelly PJ, Currie G, Delles C. Sex differences in the prevalence, outcomes and management of hypertension. Curr Hypertens Rep. 2022;24:185–192. doi: 10.1007/s11906-022-01183-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Tay K, Simon N, Friedman J, Hastie T, Tibshirani R, Narasimhan B. Regularized Cox Regression . Accessed April 25, 2023. https://glmnet.stanford.edu/articles/Coxnet.html.
- 34. Hemant I, Udaya BK, Eugene HB, Michael SL. Random survival forests. Ann Appl Stat. 2008;2:841–860. doi: 10.1214/08-AOAS169 [DOI] [Google Scholar]
- 35. Harrell FE. Regression Modeling Strategies: with Applications to Linear Models, Logistic Regression, and Survival Analysis. Springer; 2015. [Google Scholar]
- 36. Gerds TA, Schumacher M. Consistent estimation of the expected brier score in general survival models with right‐censored event times. Biom J. 2006;48:1029–1040. doi: 10.1002/bimj.200610301 [DOI] [PubMed] [Google Scholar]
- 37. Austin PC. Absolute risk reductions and numbers needed to treat can be obtained from adjusted survival models for time‐to‐event outcomes. J Clin Epidemiol. 2010;63:46–55. doi: 10.1016/j.jclinepi.2009.03.012 [DOI] [PubMed] [Google Scholar]
- 38. Blanche P, Dartigues JF, Jacqmin‐Gadda H. Estimating and comparing time‐dependent areas under receiver operating characteristic curves for censored event times with competing risks. Stat Med. 2013;32:5381–5397. doi: 10.1002/sim.5958 [DOI] [PubMed] [Google Scholar]
- 39. Siu AL. Screening for high blood pressure in adults: U.S. preventive services task force recommendation statement. Ann Intern Med. 2015;163:778–786. doi: 10.7326/m15-2223 [DOI] [PubMed] [Google Scholar]
- 40. Gooding HC, Gidding SS, Moran AE, Redmond N, Allen NB, Bacha F, Burns TL, Catov JM, Grandner MA, Harris KM, et al. Challenges and opportunities for the prevention and treatment of cardiovascular disease among young adults: report from a National Heart, Lung, and Blood Institute working group. J Am Heart Assoc. 2020;9:e016115. doi: 10.1161/JAHA.120.016115 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Aggarwal R, Chiu N, Wadhera RK, Moran AE, Raber I, Shen C, Yeh RW, Kazi DS. Racial/ethnic disparities in hypertension prevalence, awareness, treatment, and control in the United States, 2013 to 2018. Hypertension. 2021;78:1719–1726. doi: 10.1161/HYPERTENSIONAHA.121.17570 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data S1
Tables S1–S9
Figures S1–S3