Age, Gender, Race/Ethnicity, and Income-Related Patterns in Ideal Cardiovascular Health Among Adolescents and Adults in the United States

Abstract

Introduction:

Ideal cardiovascular health (CVH) is present in <50% of children and <1% of adults; yet its prevalence from adolescence through adulthood has not been fully evaluated. This study characterizes the association of age with ideal CVH and to compare these associations across gender, race/ethnicity, and socioeconomic status subgroups.

Methods:

This study conducted in 2020 analyzed adolescents and adults aged 12 to 79 years from the cross-sectional survey NHANES 2005–2016 (n=38,706). Polynomial models were used to model the association of age with ideal CVH, defined using the AHA’s Life’s Simple 7 criteria (scale 0–14 with higher values indicating better CVH).

Results:

Mean CVH was lower with increasing ages starting in early adolescence and dropping to a nadir by age 60 before stabilizing. At age 20, only 45% of adults had ideal CVH (≥5 ideal CVH metrics) and >50% of adults had poor CVH (≤2 ideal CVH metrics) at age 53. Women had higher mean CVH than men in early-life but lower mean CVH from age 60 onward. Mean CVH scores were highest for non-Hispanic white and higher income adults, and lowest for non-Hispanic blacks and low-income adults across all ages. Mean CVH scores fell from intermediate to poor levels ~30 years earlier for non-Hispanic blacks compared with non-Hispanic whites, and ~35 years earlier for low-income compared with higher-income adults.

Conclusions:

CVH scores are lower with increasing ages from early adolescence through adulthood. Race/ethnicity and income disparities in CVH were observed at young ages and were more profound at older ages.

INTRODUCTION

Ideal cardiovascular health (CVH) in adulthood is exceedingly rare.^1,2 Ideal CVH, defined by 4 health factors (body mass index (BMI), blood pressure, blood glucose, and total cholesterol) and 3 health behaviors (smoking, dietary intake, and physical activity) leads to reduced all-cause mortality and chronic diseases of aging.^1,3 In 2010, the American Heart Association released its new strategic impact goals which emphasized primordial prevention and maintenance of CVH into adulthood.³ As such, it becomes important to understand the relationship of age with CVH from youth to adulthood in order to identify ages where CVH losses might be common.

Prior longitudinal cohort studies have reported declines in CVH with age.^4–10 These studies have yielded insights on age-related trajectories in CVH and their association with clinical outcomes. However, the estimates presented in these studies often span multiple ages due to small sample sizes and follow participants for short study intervals. To date, no study has reported single-year, age-specific estimates of CVH across the entire age spectrum to better understand variations in this relationship. Furthermore, some studies have documented sizeable disparities in CVH by race/ethnicity and socioeconomic status (SES),^11–14 but how these factors modify the relationship of age with CVH remains unknown. Such information can offer insight into how to optimize CVH across all ages and may help clinicians target risk factors for specific age, race/ethnicity, and socioeconomic subgroups.

Accordingly, the goals of this study were to characterize the association of age with CVH and individual CVH metrics from early adolescence through older adulthood and to compare these associations across gender, race/ethnicity, and income subgroups. We used a nationally representative, cross-sectional sample of U.S. adolescents and adults to estimate the age-specific prevalence of CVH.

METHODS

Study Sample

This study conducted in 2020 analyzed data from the 2005–2016 National Health and Nutrition Examination Survey (NHANES).¹⁵ NHANES uses a complex, multistage, probability sampling design to collect health-related information on US adults and children through in-home interviews and examinations performed at mobile centers. This analysis included all non-pregnant participants aged 12–79 years.

Sociodemographic Characteristics

Participant age, gender, race/ethnicity, and annual household income were collected during the home interview. Self-reported race/ethnicity was categorized as non-Hispanic white, non-Hispanic black, and Mexican American (Supplemental Methods). Annual household income adjusted for household size and location was categorized in relation to the federal poverty line (FPL) as low-income (≤130% FPL) and higher income (>130% FPL).

Definition of Cardiovascular Health

Definitions of the individual CVH components are outlined in Table S1 and summarized in the Supplemental Methods.³ In brief, cardiovascular health behaviors included smoking status, self-reported physical activity, and diet quality as assessed using the Healthy Eating Index 2010 determined from dietary information collected by a single 24-hour recall.^16,17 Cardiovascular health factors included body mass index (BMI), total cholesterol, blood pressure, and fasting plasma glucose. Detailed descriptions about blood collection and processing are provided in the NHANES Laboratory Procedures Manuals.¹⁸

All components were categorized as poor, intermediate, or ideal as outlined in the AHA 2020 Strategic Impact Goals and were assigned scores of 0, 1, or 2, respectively. The overall CVH score was calculated as the sum of the individual components (range 0 to 14). For population trends evaluating mean CVH, we defined optimal CVH as having a total CVH score ≥10, intermediate CVH as having a score 7–10, and low CVH as having a score <7. In addition, we plotted the percentage of participants with ideal CVH (defined as having ideal levels on ≥5 components) and poor CVH (defined as having ideal levels on ≤2 components).

Missing Data

Approximately 15% of the sample was missing data on any CVH metric with the highest percentage of missing for any one variable being 7% for total cholesterol. For CVH components derived from continuous variables (e.g. systolic blood pressure), we used a Markov chain Monte Carlo methods assuming an arbitrary missing data pattern and specifying a noninformative Jeffreys prior with 200 bun-in iterations to impute missing values from which we recalculated the CVH component scores. For CVH components derived from categorical variables (e.g. smoking), we used a logistic regression model assuming a monotone missing data pattern to impute the scores. PROC MI and PROC MIANALYZE were used to create and analyze 10 imputed dataset. Missing income data (~8% of sample) were not imputed. Sensitivity analyses restricted to those with complete CVH metric data were also performed.

Statistical Analyses

Data were analyzed using the survey procedures in SAS version 9.4 (SAS Institute, Cary, NC) to incorporate the complex, multistage sampling design of NHANES.¹⁹ Specifically, we used PROC SURVEYFREQ, SURVEYMEANS, and SURVEYREG with MEC weights on each imputed data set and then used PROC MIANALYZE to combine results with adjusted variances on the 10 imputed datasets. Cubic or quadratic polynomial models were used to model the association of mean CVH scores with age. The polynomial degree of each model was selected by comparing adjusted r-square, Akaike’s information criteria, and Mallows’ Cp statistics across models.

These models were then repeated stratified by gender, race/ethnicity, and income. Differences in the age-related curves were assessed by comparing nested models with and without interaction terms between the sociodemographic variables (gender, race/ethnicity, and income) and the polynomial age terms using F-tests. P-values from these models represented global tests of significance for all interaction terms (e.g. female*age + female*age² + female*age³). We also compared the ages at which mean CVH fell below optimal (<10) and intermediate (<7) levels for each gender, race/ethnicity, and income subgroup.

Age-specific CVH component scores were plotted for each gender, race/ethnicity, and income subgroup to identify factors contributing to the observed differences in overall CVH. Components were plotted on the CVH scale (0–2) and on the scales of the individual health factors when possible.

Finally, we plotted the weighted percentage of participants with ideal (≥5 ideal components) and poor (≤2 ideal components) CVH by age in order to understand how the overall CVH score was distributed across its components.²³ As a sensitivity analysis to differentiate between age and secular trends, we compared mean CVH estimates and the percentage of participants with ideal and poor CVH for 12–19, 20–39, 40–59, and 60–79 year-olds in the 2005–2006 and 2015–2016 cohorts using survey-weighted student’s t-tests for means and chi-square tests for percentages.

RESULTS

The sample included 38,706 NHANES participants representing 203 million adults (20–79 years) and 33 million adolescents (12–19 years). The sample was well balanced by age and gender (Table 1, Figure S1). Most participants were non-Hispanic white with non-Hispanic blacks and Mexican Americans representing 12% (standard error [SE] 1%) and 15% (SE 1%), respectively. Approximately 23% (SE 1%) of the sample lived below 130% of the FPL.

Table 1.

Weighted characteristics of study sample

Characteristic	Weighted % (SE) or mean (SE)
Age, mean (SE)	41.7 (0.2)
Female, % (SE)	50.6 (0.2)
Race/ethnicity, % (SE)
Non-Hispanic white	65.6 (1.5)
Non-Hispanic black	12.1 (0.8
Hispanic	14.8 (1.0)
Other	7.5 (0.4)
Household income <130% of federal poverty line, % (SE)	22.8 (0.8)

Abbreviations: SE, standard error.

Overall Cardiovascular Health by Age

Mean CVH scores declined with increasing ages starting with scores just below 11 in early teen years and dropping to a nadir just below 7 at age 60 before stabilizing (Figure 1A). Cubic polynomial models explained 98% of the variance in mean values by age (Table S2). Similar age-related patterns were seen when analyses were limited to participants with complete data on CVH scores (Figure S2). When the percentage of the U.S. population with ideal and poor CVH was plotted by age, there was a precipitous decrease in the percentage of participants with ideal CVH beginning in the early 20s and a steady increase in the percentage of participants with poor CVH by age (Figure 1B).

Figure 1.

Weighted cardiovascular health scores plotted by age modeled using A) means with cubic polynomial function, and B) percentage of participants with ideal (≥5 ideal metrics) and poor (≤2 ideal metrics).

The association of mean CVH component scores by age is provided in Figure 2 and Figure S3. The greatest decrease in the prevalence of ideal smoking behavior was observed in the late teens and early 20s. The prevalence of ideal BMI dropped precipitously around age 20 years; however, this was likely explained by different definitions of ideal BMI for adolescents and adults. Physical activity, blood pressure, and total cholesterol scores decreased gradually by age whereas diet scores increased with age. Glucose scores were stable until around age 40 when they gradually decreased among older individuals. Similar patterns were observed when CVH component scores were modeled as percentages with ideal and poor values (Figure S4).

Figure 2.

Weighted mean cardiovascular health component scores plotted by age modeled with a cubic polynomial function.

Cardiovascular Health by Age and Gender

The relationship between CVH and age differed for men and women (p for gender*age interactions <0.001). Women had higher mean CVH scores than men in early- to mid-life but lower scores from age 60 onward (Figure 3A). Mean CVH scores decreased from ideal to intermediate levels at ages 19 and 22 years and from intermediate to poor levels at ages >79 and 62 years in men and women, respectively (Table S3).

Figure 3.

Weighted mean cardiovascular health scores plotted by age modeled with cubic polynomial functions stratified by A) gender, B) race/ethnicity, and C) income.

Differences in age-specific patterns of mean CVH for men and women were largely explained by differences in smoking, blood pressure, total cholesterol, and glucose (Figure S5). Women consistently had higher mean scores for these 4 components early in life but poorer total cholesterol control later in life. In contrast, men had higher physical activity scores throughout adulthood.

Cardiovascular Health by Age and Race/Ethnicity

Mean CVH scores were highest for non-Hispanic whites and lowest for non-Hispanic blacks across all ages (Figure 3B). Mean CVH scores decreased from ideal to intermediate levels at similar ages for all races/ethnicities; however, mean CVH scores decreased from intermediate to poor levels approximately 30 years earlier in non-Hispanic blacks and 20 years earlier in Mexican Americans compared with non-Hispanic whites (Table S3). Global F-tests comparing age-related curves in CVH for white and black participants and black and Hispanic participants were significant (p <0.001) but not for white and Hispanic participants (p=0.95).

Differences in age-specific patterns of BMI, physical activity, blood pressure, and glucose control largely accounted for the observed differences in mean CVH by race/ethnicity (Figure S6). Non-Hispanic whites consistently had more ideal BMI, physical activity, and glucose scores than non-Hispanic blacks and Mexican Americans. Non-Hispanic blacks had poorer blood pressure scores across all ages.

Cardiovascular Healthy by Age and Income

Mean CVH scores were markedly lower for low-income adolescents and adults compared with those with higher income (Figure 3C). Mean CVH decreased from ideal to intermediate levels at ages 18 and 21 years and from intermediate to poor levels at ages 45 and >79 years for low-income and higher income participants, respectively (Table S3). The relationship between age and CVH differed significantly for low- and higher income participants (p for age*poverty interactions <0.001). The largest differences in mean CVH components by poverty were observed for smoking and physical activity (Figure S7).

Sensitivity Analyses

To assess for secular changes in CVH, we compared CVH estimates for 12–19, 20–39, 40–59, and 60–79 year-olds in the 2005–2006 and 2015–2016 cohorts (Table S4). Mean CVH was similar between the two cohorts for adults aged 40–59 and 60–79 years, but adolescents aged 12–19 years and young adults aged 20–39 years in the 2015–2016 cohort had significantly higher mean CVH compared to those in the 2005–2006 cohort. Similar findings were identified for comparisons of the proportion of adults with ideal and poor CVH.

DISCUSSION

In this nationally representative, cross-sectional sample of the U.S. adolescent and adult population, we found lower CVH scores with increasing age, a pattern that emerged in early adolescence and then stabilized in late adulthood. At age 20 years, fewer than half of adults had ideal CVH (≥5 components), and by age 53 more than half of adults had poor CVH (≤2 components). The relationship of age with CVH differed by gender, race/ethnicity, and income status with different CVH components accounting for the sociodemographic differences observed. Women had higher CVH compared with men at younger ages, but these differences were less prominent at older ages. Non-Hispanic blacks and low-income adults had consistently lower CVH scores compared with their non-Hispanic white and higher income counterparts. Mean CVH fell to poor levels nearly 30 years earlier for non-Hispanic blacks compared with non-Hispanic whites and 35 years earlier for low-income compared with higher income adults.

Our findings reinforce those of longitudinal cohort studies demonstrating declines in ideal CVH over the lifespan.^4–8,20 For example, data from the Atherosclerosis Risk in Communities study documented racial differences in the rate of decline for 6 CVH metrics between ages 40 to 90 years.⁴ However, all of these longitudinal studies have been limited by small sample sizes and infrequent follow-up visits necessitating reports of CVH across broadly defined age groups. The larger and nationally representative sample size of NHANES allows us to extend this work to examine age-specific estimates of CVH from adolescence to adulthood and to quantify disparities across gender, race/ethnicity, and income in the US.

We found early decreases in CVH across every subgroup with mean CVH scores falling from optimal to intermediate levels between ages 19 and 23 years. These findings suggest that adolescence and young adulthood are times of heightened vulnerability for developing poor CVH behaviors and highlight the importance of primordial prevention efforts. Among younger participants, the inverse relationship between CVH and age appeared to be due to increased smoking, weight gain, and poorer blood pressure control. These factors predominated across all subgroups suggesting avenues for intervention when trying to maintain optimal levels of CVH in a population. Specifically, early reinforcement of healthy eating and exercising habits coupled with anti-smoking campaigns may help to preserve CVH early in life.^21–23 Such effects may also have long-lasting benefits later in life, if behaviors are maintained, and may ultimately delay the onset of cardiovascular diseases. Several studies have shown that diet and physical activity behaviors established in adolescence persist into later adulthood.^24,25 However, previous interventions to promote these behaviors in young adults have been met with varied success.^26–31 Lessons from these programs suggest that future interventions need to target multiple CVH factors, tailor intervention content to retain participants, and focus on long-term goals in order to establish lasting effects.

Interestingly, women had higher mean CVH scores than men in early to mid-adulthood which reversed after age 60. We hypothesized that this reversal might be explained by a healthy survivor effect in men given gender differences in life expectancy, or alternatively by the onset of menopause in women given changes in lipid profiles, blood pressure, and weight gain that occur with menopause.^32–34 Poorer CVH in men at younger ages was largely explained by worse blood pressure control whereas poorer CVH in women at older ages was explained by worse cholesterol control. These findings may help target blood pressure and lipid screening efforts in men and women across the age spectrum.

Non-Hispanic blacks and Mexican Americans consistently had lower mean CVH scores than non-Hispanic whites across all ages – a phenomenon that was explained by higher BMI, lower physical activity, and poorer blood pressure and glucose control in these groups. Prior studies have shown that the prevalence of hypertension and diabetes in non-Hispanic blacks is almost twice that in non-Hispanic whites owing to earlier onset and poorer control despite increased awareness and pharmacologic treatment.^35–40 Pediatric data suggest that disparities in blood pressure and glucose control begin well before adult life.^41–43 Early and persistent racial disparities in CVH likely account, in part, for the earlier onset of coronary artery disease and stroke in African Americans, with some studies estimating that hypertension alone explains ~50% of the excess risk of stroke in blacks.^35,44

Finally, we identified profound disparities in CVH by income, which were attributable to higher smoking and lower physical activity in low-income adults. Low-income adolescents are nearly 3 times more likely to start smoking than high-income adolescents, and early nicotine dependence predicts smoking regularity and quantity into young adulthood.^45,46 Tobacco control measures, such as taxation, smoke-free legislation, and anti-tobacco media campaigns, have proven effective in reducing socioeconomic inequalities in smoking,^47,48 but low-income adults still account for the majority of U.S. smokers.⁴⁹ Interventions targeting tobacco cessation in addition to prevention may be needed to reduce socioeconomic disparities in CVH further.

Although we analyzed CVH scores by gender, race/ethnicity, and income separately, there is likely significant overlap in the relationships identified with age. Poverty rates in black and Hispanic people living in the U.S. are over twice those in non-Hispanic whites, and many more minorities live on the margins.⁵⁰ As such, the poorer CVH observed in black and Hispanic participants may be explained, in part, by the poorer CVH observed in low-income participants. Nevertheless, different CVH components appeared to mediate the differences in CVH by race/ethnicity (e.g. BMI, physical activity, blood pressure, and glucose control) and income (e.g. smoking and physical activity).

The major limitation of this study is the use of cross-sectional data to examine age-related patterns in CVH. This type of study design does not allow us to distinguish between age, cohort, and period effects, and thus, we are unable to comment on CVH trajectories. Instead, we are only able to evaluate the association of age with CVH. The sensitivity analyses showed declines in ideal CVH for certain metrics and improvements in others between the beginning of the cohort (2005–2006) to the end (2015–2016). These findings suggest that cohort and/or period effects may be present, particularly in younger cohorts where the change was most marked, but longitudinal study designs are better suited to evaluate these questions. Nevertheless, our results are consistent with those of past longitudinal cohort studies, and the granular, age-specific CVH estimates reported here can only be accomplished with a large, comprehensive dataset like NHANES.

Additional limitations of this study pertain to the measurement of certain variables. First, we measured income at a single timepoint and thus are unable to account for the influences of poverty earlier in life or for changes in income over time. Second, dietary quality can be challenging to measure due in part to differences in nutritional recommendations across organizations and populations.^51,52 We chose not to use the AHA score due to discrepancies in the AHA and USDA’s dietary recommendations for sodium; however, we still indexed the HEI2010 dietary score to values calculated with the AHA recommendations. This approach may lead to more people in our sample being classified as having poor diet quality than if the less stringent USDA threshold of <2300 mg daily had been used. Third, we were unable to account for the use of insulin or diabetes medication in determining ideal, intermediate, and poor glucose levels.

CONCLUSIONS

In summary, we found a pattern of lower CVH scores with increasing age across all gender, race/ethnicity, and income subgroups that began in early adolescence and persisted throughout early and mid-adulthood. Racial/ethnic and income disparities in CVH began early and were more profound at older ages. These findings underscore the importance of primordial prevention efforts and highlight the need for interventions that target specific CVH risk factors within racial and income subgroups.