Cardiovascular Disease Projections in the United States Based on the 2020 Census Estimates

DEFINITIONS OF TERMS.

In NHANES, prevalence of DM was defined by a glycosylated hemoglobin value ≥6.5% or self-reported past diagnosis of DM or taking glucose-lowering medications. The prevalence of hypertension was defined by using blood pressure measurements during physical examination (systolic blood pressure ≥130 mm Hg or diastolic blood pressure ≥80 mm Hg) or the responses to interview questions about being told of having high blood pressure or taking blood pressure medications. Dyslipidemia was defined based on self-report or taking cholesterol-lowering medication. Obesity was defined by a body mass index ≥30 kg/m². The prevalence of IHD was based on patient self-report during an interview that asked about IHD, angina, and MI. The prevalence of HF and stroke was also based on patient self-report.

STATISTICAL METHODS.

The “survey” package in R (R Foundation for Statistical Computing) was used to account for the complex sampling design of the NHANES study. Sample weights were used in the required analyses. Using NHANES data from 2013 to 2018, logistic regression models were developed adjusting for age, sex, race, ethnicity, and survey years as independent variables and each CV risk factor and CV disease as a dependent variable. Furthermore, we assessed and included any significant interactions between demographic characteristics in the model. To account for stability and generaliz-ability of our results, we calculated the 95% CI of the prevalence estimates by performing bootstrap resampling 1,000 times and providing CIs for the projections. Using predictions from these logistic regression models, the prevalence of each CV risk factor or disease was estimated in each of 40 groups of sex (male and female), age category (18–44, 45–64, 65–79, and ≥80 years), and race and ethnicity (Asian, Black, Hispanic, White, and Other [Alaska Native, Native American, Native Hawaiian, Other Pacific Islanders, and those with 2 or more races]).

We next combined these prevalence estimates with U.S. Census projections of population counts (provided for each age/sex/race and ethnicity group) for years 2025–2060 using data released by the U.S. Census Bureau in 2020, to project the number of people with prevalent CV risk factors or CV disease through 2060.⁸ The projections are based on assumptions about future births, deaths, and net international immigration.⁹ To integrate prevalence estimates with U.S. Census projection counts, the predicted prevalence of each CV risk factor or condition in each sex/age/race and ethnicity group was multiplied by the projected population counts in the corresponding groups for years 2025–2060 to project the number of people with CV risk factors or disease in each cell according to 5-year blocks. To consider the percentage of change in both CV risk factors and disease, the following was used:

All P values reported are 2-sided. A P value <0.05 was considered statistically significant. All statistical analyses were performed by using R version 3.6.2.

CV RISK FACTORS.

Figure 1 shows the projected overall prevalence of CV risk factors from 2025 to 2060 among the general U.S. population. By the year 2060, the prevalence of CV risk factors is estimated to rise: DM, 39.3% increase compared with 2025 (16.8% of the population or 54.6 million persons); hypertension, 27.1% increase compared with 2025 (50.1% of the population or 162.5 million persons); dyslipidemia, 27.6% increase compared with 2025 (38.7% of the population or 125.7 million persons); and obesity, 18.3% increase compared with 2025 (38.8% of the population or 125.7 million persons). The fastest rate of increase in DM, hypertension, and dyslipidemia is projected to occur between the years 2025 and 2030. The prevalence of obesity is projected to decrease slightly from 39.4% in 2025 to 38.8% in 2060, although the absolute number of individuals with obesity is expected to rise. As shown in Figure 1, apart from obesity (for which women are projected to continue to have higher prevalence), differences in projected prevalence of other CV risk factors are generally negligible between women and men.

The projected change in the prevalence of each CV risk factor during the 5-year blocks shows a noteworthy pattern across decades: the most rapid increase in prevalence is projected between 2025 and 2030, then slows until 2045 or 2050, after which the increase in prevalence grows more quickly again for CV risk factors (except obesity) (Figure 2).

CV DISEASES.

Figure 3 details the projected prevalence of CV diseases by 2060. The prevalence is projected to rise for IHD (30.7% increase compared with 2025; 8.8% of the population or 28.7 million persons), HF (33.4% increase compared with 2025; 4.0% of the population or 12.9 million persons), MI (16.9% increase compared with 2025; 4.9% of the population or 16.0 million persons), and stroke (33.8% increase compared with 2025; 4.5% of the population or 14.5 million persons). As shown in Figure 3, men will have a higher prevalence of IHD and MI compared with women, but the projected change in CV disease prevalence among both sexes over time is similar. The change in each CV disease prevalence during 5-year blocks exhibits a noteworthy pattern across decades: between 2040 and 2050, projections show a state of little or no change before rising again in 2050. The highest rate of increase is projected to occur between 2025 and 2030 (Figure 4).

PROJECTED CHANGE ACCORDING TO RACE AND ETHNICITY.

Distribution of projected prevalence and race and ethnic-specific prevalence (within one’s group) of CV risk factors according to race and ethnicity through 2060 are shown in Table 1. By 2060, the highest prevalence of CV risk factors will remain among the White population due to its overall size, despite a gradual decrease in prevalence of CV risk factors. Conversely, the projected prevalence of CV risk factors is expected to rise among all other race and ethnicities. As an example, the projected race- and ethnic-specific prevalence of DM, hypertension, and obesity will be highest in the Black population starting from 2025 and increasing further through 2060; in 2060, 19.8% of all Black adults are projected to have DM, 59.9% to have hypertension, 35.9% to have dyslipidemia, and 45.6% to have obesity. The percent change in race- and ethnic-specific prevalence of CV risk factors and CV disease is further detailed in Supplemental Table 4, which shows a rise in race- and ethnic-specific CV risk factor burden in racial and ethnic minority groups. For DM, hypertension, and dyslipidemia, Hispanic persons are projected to have the largest increase in prevalence by 2060.

Projected race and ethnicity distribution of prevalence and race- and ethnic-specific CV disease prevalence is detailed in Table 2. The projected prevalence of CV diseases will decrease for the White population and rise in all other race/ethnicities, leading to higher race- and ethnic-specific prevalence and greater rates of increase for IHD, MI, HF, and stroke, particularly in the Black and Hispanic populations.

STUDY LIMITATIONS.

Importantly, in generating predictions for future CV disease, the approach used assumes that patterns of CV risk factors and CV disease according to age/sex/race and ethnicity group will continue for the duration of the time examined. This conventional method assumes that any current estimate of the disease will cease at the beginning of the forecast period, and the prevalence is held constant when projecting numbers of cases into the future. This method has been used by several high-impact efforts to estimate future disease, including from the International Diabetes Federation,²⁷ Heidenreich et al,²¹ Huovinen et al,²⁸ Odden et al,²⁹ Huffman et al,³⁰ Heidenreich et al,³¹ Ovbiagele et al,³² and Nelson et al.³³ Lending credence to the results of this study, the near-term projections in our estimates for the year 2030 were entirely consistent with the outer horizon of estimates from Heidenreich et al²⁸; although we have derived our CVD prevalence estimate from a different NHANES data set and different U.S. census counts, our projection numbers were similar. This implies the accuracy and reproducibility of the method. Importantly, in multivariable logistic regression, age, sex, and ethnicity (but not survey year) were associated with CV disease. This implies that demographic factors are determinants of CV disease rather than the survey year. Second, we did not take into account the effect of coronavirus disease-2019 (COVID-19) on our estimates; nearly 1 million deaths caused by COVID-19 have occurred so far in the United States, and many individuals have been left with morbid complications of the diagnosis. The long-term consequences of COVID-19 on the CV system remain unclear, but this is an obvious concern.³⁴ Importantly, minoritized populations have been disproportionality affected by COVID-19, which may further influence our projections.³⁵ Third, CV diseases were defined based on self-report, which is important as some populations with limited health literacy may not self-report past illness with the same frequency.³⁶ Moreover, self-reporting can potentially lead to many epidemiologic biases, including misclassification, ascertainment, and survivor bias, which can ultimately affect our projection results. This is a fundamental bias embedded in NHANES. However, NHANES is one of the most accurate survey data sets available, and the methodology used is consistent with other efforts similar to ours.^37–42 Clair et al⁴³ compared 2 methods of measuring disease prevalence: 1) using self-reported disease from survey responses; and 2) using disease diagnosis codes from administrative data. Multiple sources were included: Medicare claims data, NHANES, and the Health and Retirement Study. The investigators concluded that surveys provide less biased samples and may be the preferred method because they are nationally representative. Conversely, claim data provide more accurate diagnosis and may be more suited for research on specific disease likely to be less well-defined in surveys. Lastly, we need to mention that we used history of treatment as part of the CV risk definition. This may lead to underestimation of prevalence of CV risk factors in minoritized individuals with limited health access.