Abstract
Introduction:
Epidemiologic studies relating health outcomes to dietary patterns captured by diet quality indices have shown better quality scores associated with lower mortality and chronic disease incidence. However, changing chronic disease risk factors only alters population health over time, and initial diet quality systematically varies across the population by sociodemographic status. This study uses microsimulation to examine 30-year impacts of improved diet quality by sociodemographic group.
Methods:
Diet quality across 12 sex-, race/ethnicity-, and education-defined subgroups was estimated from the 2011–2012 National Health and Nutrition Examination Survey. In 2021, the Future Adults (dynamic microsimulation) Model was used to simulate population health and economic outcomes over 30 years for these subgroups and all adults. The modeled pathway was through lowering risk for heart disease by following U.S. Dietary Guidelines.
Results:
Diet quality varied across the sociodemographic subgroups and half of U.S. adults had diet quality that would be classified as poor. Improving U.S. diet quality to that reported for the top 20% in 2 large health professionals’ samples could reduce incidence of heart disease by 9.9% (7.6%–13.8% across the 12 sociodemographic groups) after 30 years. Year 30 would also have 37,000 fewer deaths, 694,000 more quality-adjusted life years, and healthcare cost savings of $59.6 billion (2019 U.S. dollars).
Conclusions:
Dynamic microsimulation enables predictions of socially important outcomes of prevention efforts, most of which are many years in the future and beyond the scope of trials. This paper estimates the 30-year population health and economic impact of poor diet quality by sociodemographic group.
INTRODUCTION
Diet quality is a major determinant of health.1, 2 The concept of diet quality recognizes that people do not consume nutrients and foods in isolation, but as part of a dietary pattern synergistically affecting health.2
The best known estimates of the social burden of dietary risk factors come from the Global Burden of Disease (GBD) project,3 which attributes 1 in 5 deaths globally and half of cardiovascular disease (CVD) deaths in developed countries to dietary factors, independent of obesity.4, 5 The GBD relies on associations between individual food groups and health, and recognizes the shortcomings of the food group approach: Intake of healthy dietary factors are generally positively correlated with each other and inversely correlated with harmful dietary factors, leading to exaggerated effect sizes for individual dietary factors.5
A variety of diet quality indices have been created to summarize complete diet patterns and systematic reviews of epidemiologic studies relating health outcomes to these indices have shown that diets with higher quality scores were associated with lower all-cause mortality, CVD, cancer, type 2 diabetes, and neurodegenerative diseases with risk ratios ranging from 0.80 to 0.86 when comparing the scores in the highest to the lowest quintile.6–9 The Healthy Eating Index (HEI) captures diet quality according to U.S. Dietary Guidelines for Americans,10–12 and the National Cancer Institute and the U.S. Department of Agriculture have collaborated to update it (HEI-2015).
The static macro-level calculations used by the GBD and others13, 14 use attributable or relative risk calculations, a popular approach to cost-of-illness/burden-of-disease estimates. However, eliminating a risk factor now does not immediately change the prevalence of chronic diseases or healthcare needs. These changes occur only over time through disease incidence and mortality. When a new equilibrium is reached, possibly after decades, population characteristics have changed as well, partly in response.
Microsimulation models can capture the dynamics of population health while simultaneously tracking how individuals or subgroups with different characteristics fare over time. Two microsimulation models have recently been applied to diet quality, the Future Adult Model (FAM)15 and the CVD-PREDICT model.16–18 These 2 models have complementary strengths and weaknesses: FAM is well suited for outcomes such as medical expenditures, disability, work, and government programs, but lacks clinical details beyond chronic disease indicators. CVD- PREDICT was built for CVD but lacks economic or social outcome information in individual transitions—for example, it uses average healthcare costs from the literature for a condition regardless of patient characteristics or disease stage/duration.18 Diet quality was also conceptualized differently: The FAM study used the Alternative HEI-2010 (AHEI-2010) and CVD-PREDICT, like the GBD, used a food group approach—that is, optimal intake of fruits, vegetables, nuts/seeds, whole grains, red meat, processed meat, sugar-sweetened beverages, fatty acids, omega-3 fats, and sodium.
This paper builds on the previous FAM approach, but also addresses several of its limitations.15 In particular, the earlier publication assumed that the entire population started at average diet quality as measured by the AHEI-2010 and then improved to the top quintile,19 but this ignores that diet quality varies across the population and there are systematic differences by sociodemographic status.20 This study updates to a newer index (HEI-2015), estimates the distribution of diet quality within and across sociodemographic groups, links those to corresponding hazard ratios for heart disease, and then simulates the potential health and economic changes over time associated with diet quality improvements using the FAM.
METHODS
Study Sample
This study integrated estimates from the National Health and Nutrition Examination Survey (NHANES) public use files of diet quality by sociodemographic subgroup, published estimates on the relationship between disease incidence and diet quality, and the FAM microsimulation model to determine the 30-year impact from hypothetical changes in diet quality on health and economic outcomes. Results were summarized separately for 12 sociodemographic subgroups defined by sex (male, female), education (no college, some college), and race/ethnicity (Hispanic, non-Hispanic Black, all other). The RAND Human Subjects Protection Committee determined that this project does not involve human subjects.
The FAM is a dynamic microsimulation model that simulates individual life trajectories for a cohort of U.S. adults aged ≥25 years. The primary engine of the FAM is a set of transition equations between health states and economic outcomes primarily estimated from the Panel Study of Income Dynamics 2001–2015. The model runs in 2-year steps during which individuals experience events according to their risk equations. At every step, a new group of individuals aged 25–26 years is added through a replenishing module.
Measures
Risk factors, chronic diseases, functional limitations, and economic outcomes change probabilistically over time according to the estimated transition equations. Some characteristics of the individual do not change over time (e.g., sex, race/ethnicity) or are deterministic (e.g., age, Medicare eligibility). Some chronic diseases affect other chronic diseases’ incidence models (e.g., heart disease or hypertension affects stroke incidence). In this paper, diet is considered a permanent individual characteristic rather than something that changes—that is, a hypothetical improvement in diet quality remains for the rest of a person’s life.
The prevalence of chronic disease is based on has a doctor ever told you that you have questions and the heart disease question includes coronary heart disease (CHD), angina, and congestive heart failure. General health status used a scale from 1=poor to 5=excellent and disability included living in a nursing home or difficulty doing any of a list of activities of daily living or instrumental activities of daily living by yourself. Economic outcomes include total healthcare costs (government, private insurance, and patients’ out-of-pocket) and whether an individual was working for pay and using Social Security or other income support programs. Further details on the FAM are found in its Technical Documentation Report.21
The NHANES, conducted by the Centers for Disease Control and Prevention, examines about 5,000 U.S. residents every 2 years, which after weighting produces estimates representative of non-institutionalized civilian population in the U.S.22 The NHANES includes 24-hour dietary recalls (2 total) from which the HEI-2015 (total score of 0–100 across 13 dietary components)11, 12 was calculated using the Markov Chain Monte Carlo approach developed by the National Cancer Institute.23 The NHANES 2011–2012 cycle was used to enable replication of previously published HEI-2015 and NHANES results.12 Although results are summarized for 12 subgroups, HEI-2015 scores vary within each group and this distribution is incorporated in the simulations.
Shan et al.24 reported hazard ratios for the incidence of CVD by quintile of the HEI-2015 score distribution seen in the first Nurses’ Health Study (NHS) for women and the Health Professionals Follow-up Study (HPFS) for men. Separate estimates of the impact of diet quality on CHD and stroke were obtained from the authors (Appendix Table 1). The relationship between diet quality and stroke was not statistically significant and that direct causal effect was not modeled (even though it is possible that one exists). Their multivariate-adjusted hazard ratios for CHD were not applied to the diet quality scores directly, but to quintiles of scores within the NHS or HPFS. This use of quintiles reduces the comparability problems inherent between HEI- 2015 scores calculated from 24-hour recall (NHANES) and scores calculated from food frequency questionnaires (NHS/HPFS samples).25 Individuals in the simulation were assigned to the cut points for a quintile in Shan and colleagues24 using a normal distribution approximation. Hazard ratios were rescaled to match the risks within subgroups.
Statistical Analysis
Each simulated individual was assigned a diet quality (expressed as being in 1 of the 5 NHS/HPFS quintiles). In the baseline scenario, this assignment used the distribution of scores from NHANES. Two substantial, but possible counterfactual scenarios assumed that diet quality improved in 2019 and continued at the improved level for 30 years. The first scenario assumed that everybody improved to the top NHS/HPFS quintile (no change to those already assigned to the top level). The second scenario assumed that individuals in the first 3 quintiles improved by 2 quintiles and those in the fourth quintile improved by 1. Medical costs were adjusted to 2019 U.S. dollars using the Consumer Price Index (U.S. Bureau of Labor Statistics) medical cost component.
Uncertainty in the FAM transition equations was assessed with a nonparametric bootstrap (which re-estimates transition equations). One hundred bootstrap replications were used to estimate 95% CIs for model outcomes. All results were calculated in 2021 using the FAM, version 202.
RESULTS
The median HEI-2015 scores (range=0–100) for adults aged >25 years ranged from 54.7 (male, non-Hispanic White or other race, no college) to 63.2 (female, non-Hispanic White or other race, some college) (Table 1).
Table 1.
HEI-2015 Scores and Heart Disease Reductions in Years 10 and 30 Under 2 Diet- Quality Improvement Scenarios
| Variable | HEI-2015 score, Median (IQR) | Percent reduction in heart disease (reduction in heart disease cases in thousands) | |||
|---|---|---|---|---|---|
| Scenario 1 - All to quintile 5 | Scenario 2 - Up 2 quintiles | ||||
| Year 10: 2029 | Year 30: 2049 | Year 10: 2029 | Year 30: 2049 | ||
| All U.S. adults aged ≥25 years | 58.9 (51.5, 66.3) | −6.5 (−2,860.6) | −9.9 (−5,471.6) | −3.8 (−1,660.4) | −5.7 (−3,157.9) |
| (1) Male, non-Hispanic White or other race, no college | 54.7 (47.9, 62.3) | −6.6 (−412.0) | −9.1 (−561.6) | −3.8 (−236.8) | −5.2 (−318.9) |
| (2) Male, non-Hispanic White or other race, some college | 60.7 (53.1, 68.1) | −5.3 (−592.6) | −8.5 (−1,126.1) | −3.2 (−359.0) | −5.1 (−685.1) |
| (3) Male, Hispanics, no college | 55.8 (50.3, 61.1) | −7.6 (−101.9) | −12.1 (−294.7) | −4.3 (−57.4) | −6.8 (−165.3) |
| (4) Male, Hispanics, some college | 60.2 (54.2, 65.9) | −6.6 (−62.2) | −9.8 (−190.5) | −4.0 (−37.9) | −5.9 (−114.8) |
| (5) Male, non-Hispanic Black, no college | 55.5 (49.5, 61.3) | −7.5 (−84.4) | −10.1 (−156.4) | −4.3 (−47.7) | −5.7 (−88.0) |
| (6) Male, non-Hispanic Black, some college | 59.5 (50.6, 68.4) | −4.7 (−57.7) | −7.6 (−142.2) | −2.8 (−34.7) | −4.5 (−85.2) |
| (7) Female, non-Hispanic White or other race, no college | 60.1 (52.8, 67.0) | −7.6 (−414.5) | −11.7 (−502.5) | −4.4 (−237.1) | −6.7 (−286.7) |
| (8) Female, non-Hispanic White or other race, some college | 63.2 (54.4, 70.9) | −6.1 (−675.8) | −9.5 (−1,390.8) | −3.6 (−399.1) | −5.5 (−814.6) |
| (9) Female, Hispanics, no college | 60.0 (53.9, 65.8) | −8.0 (−112.2) | −11.6 (−261.8) | −4.6 (−64.4) | −6.7 (−151.7) |
| (10) Female, Hispanics, some college | 57.1 (50.8, 63.5) | −8.8 (−102.5) | −12.4 (−352.2) | −4.8 (−55.4) | −6.6 (−187.0) |
| (11) Female, non-Hispanic Black, no college | 56.4 (50.2, 62.6) | −8.9 (−124.4) | −13.8 (−236.4) | −4.7 (−66.0) | −7.1 (−122.4) |
| (12) Female, non-Hispanic Black, some college | 58.3 (49.6, 66.7) | −7.8 (−120.3) | −11.5 (−256.4) | −4.2 (−64.8) | −6.2 (−138.2) |
HEI-2015, Healthy Eating Index 2015 version.
Table 1 also shows the percentage and absolute reductions in cases of CHD estimated using FAM for 2 hypothetical diet improvement scenarios in Year 10 and Year 30 for the U.S. adult population and by subgroup. The reductions seen for Scenario 1 (all improve to Quintile 5) were larger than seen for Scenario 2 (improving by 2 quintiles), and the reductions were larger in Year 30 than in Year 10. In Year 30, across subgroups the percentage reduction in CHD cases (compared with cases expected with no diet improvement) ranged from 7.6% (male, non- Hispanic Black, some college) to 13.8% (female, non-Hispanic Black, no college) for Scenario 1 and from 4.5% (male, non-Hispanic Black, some college) to 7.1% (female, non-Hispanic Black, no college) for Scenario 2.
The first column of Table 2 shows the percentage of the U.S. adult population represented by each subgroup according to U.S. Census data. The rest of the table shows the case reductions for CHD and death for the U.S. population and by subgroup by scenario in terms of absolute numbers and percentage of total case reductions. Comparing each subgroup’s percentage contribution to total case reductions to that subgroup’s percentage of the U.S. population indicates the subgroups where diet quality improvements have more or less impact. In general, the reductions in CHD cases in each subgroup were roughly what would be expected from their proportion of the U.S. population. However, there was some variation—for example, diet improvement in the subgroup female, Hispanic, some college had a larger impact on total CHD reductions than would be the case if all groups benefited equally. The impact of improved diet quality on reductions in deaths also varied, reflecting the interaction with other mortality risks.
Table 2.
Year-30 Reductions in Heart Disease Cases and Deaths by Group and Diet-Quality Improvement Scenario
| Number of cases in thousands (% of total reductions) | % of U.S. population | Scenario 1 - All to quintile 5 | Scenario 2 - Up 2 quintiles | ||
|---|---|---|---|---|---|
| Heart disease cases (% of total cases) | Deaths (% of total deaths) | Heart disease cases (% of total cases) | Deaths (% of total deaths) | ||
| All U.S. adults aged ≥25 years | 100% | −5,471.6 (100.0) | −36.6 (100.0) | −3,157.9 (100.0) | −21.2 (100.0) |
| (1) Male, non-Hispanic White or other race, no college | 11.5 | −561.6 (10.3) | −2.7 (7.5) | −318.9 (10.1) | −1.5 (7.0) |
| (2) Male, non-Hispanic White or other race, some college | 23.3 | −1,126.1 (20.6) | −6.4 (17.5) | −685.1 (21.7) | −4.1 (19.3) |
| (3) Male, Hispanics, no college | 4.9 | −294.7 (5.4) | −2.3 (6.2) | −165.3 (5.2) | −1.2 (5.7) |
| (4) Male, Hispanics, some college | 2.9 | −190.5 (3.5) | −2.0 (5.4) | −114.8 (3.6) | −1.2 (5.7) |
| (5) Male, non-Hispanic Black, no college | 2.7 | −156.4 (2.9) | −1.5 (4.2) | −88.0 (2.8) | −0.7 (3.4) |
| (6) Male, non-Hispanic Black, some college | 2.9 | −142.2 (2.6) | −2.2 (6.1) | −85.2 (2.7) | −1.4 (6.5) |
| (7) Female, non-Hispanic White or other race, no college | 11.7 | −502.5 (9.2) | −3.0 (8.3) | −286.7 (9.1) | −2.5 (11.6) |
| (8) Female, non-Hispanic White or other race, some college | 25.5 | −1,390.8 (25.4) | −9.9 (27.1) | −814.6 (25.8) | −5.6 (26.5) |
| (9) Female, Hispanics, no college | 4.4 | −261.8 (4.8) | −1.7 (4.7) | −151.7 (4.8) | −1.0 (4.5) |
| (10) Female, Hispanics, some college | 3.4 | −352.2 (6.4) | −2.2 (6.0) | −187.0 (5.9) | −1.1 (5.0) |
| (11) Female, non-Hispanic Black, no college | 2.8 | −236.4 (4.3) | −1.0 (2.6) | −122.4 (3.9) | −0.3 (1.5) |
| (12) Female, non-Hispanic Black, some college | 4.0 | −256.4 (4.7) | −1.7 (4.6) | −138.2 (4.4) | −0.7 (3.5) |
Table 3 shows Year 30 results by diet quality improvement scenario for various health and economic measures. The benefits of Scenario 1 (diet quality improvement to the top quintile) were larger than the benefits of Scenario 2 (improvement by 2 quintiles). In addition to reducing the number of cases of CHD, improving diet quality reduced the number of people reporting fair or poor health, disability, and having 1 (morbidity) or ≥2 (comorbidity) chronic diseases. Improved diet quality also resulted in 21,000–37,000 fewer deaths in Year 30, and increased quality-adjusted life years. The reductions in CHD resulted in reductions of $34.0 billion to $59.6 billion in healthcare costs (even with fewer deaths resulting in more people using the healthcare system) as well as reductions in Supplemental Social Security Income and Disability Insurance claims, while increasing participation in the workforce by 125,000 to >200,000 that year. People living longer because of improved diet quality also increased those receiving Social Security in Year 30 by 198,000 to 338,000.
Table 3.
Year-30 Future Adult Model Results by Diet-Quality Improvement Scenario, Each Compared to the Status Quo
| Variable | Year-30 Results for Scenario 1 (All to Quintile 5) | Year-30 Results for Scenario 2 (Up 2 Quintiles) | ||
|---|---|---|---|---|
| Mean (95% CI) | Difference from status quo (%) | Mean (95% CI) | Difference from status quo (%) | |
| Prevalence of heart disease (millions) | 49.9 (46.9, 52.9) | −5.5 (−9.9) | 52.2 (49.1, 55.4) | −3.2 (−5.7) |
| Self-report of health (fair to poor, millions) | 60.8 (58.6, 63.1) | −0.5 (−0.7) | 61.0 (58.8, 63.2) | −0.3 (−0.4) |
| Quality-adjusted life-years (QALYs) gained (millions) | 227.4 (225.5, 229.3) | 0.7 (0.3) | 227.1 (225.2, 229.0) | 0.4 (0.2) |
| Disability (difficulties with ADL/IADL or nursing home, millions) | 68.1 (65.2, 70.9) | −0.1 (−0.2) | 68.1 (65.3, 71.0) | −0.1 (−0.1) |
| Morbidity (millions)a | 162.7 (160.7, 164.8) | −0.4 (−0.3) | 162.9 (160.9, 164.9) | −0.2 (−0.2) |
| Co-morbidity (millions)b | 87.1 (84.4, 89.7) | −1.6 (−1.8) | 87.7 (85.1, 90.4) | −0.9 (−1.0) |
| Deaths (thousands) | 8,478.3 (8,405.1, 8,551.5) | −36.6 (−0.4) | 8,493.7 (8,421.7, 8,565.7) | −21.2 (−0.2) |
| Economic outcomes in Year 30 | ||||
| Total medical spending ($billions, 2019USD) | 8,481.4 (8,386.4, 8,576.5) | −59.6 (−0.7) | 8,507.1 (8,602.9, 8,411.3) | −34.0 (−0.4) |
| People working for pay (millions) | 172.5 (170.1, 175.0) | 0.2 (0.1) | 172.4 (170.0, 174.9) | 0.1 (0.1) |
| Social Security (OASI) claims (millions) | 75.4 (69.0, 81.9) | 0.3 (0.5) | 75.3 (68.9, 81.7) | 0.2 (0.3) |
| Supplemental Social Security Income claims (thousands) | 5,064.8 (4,228.3, 5,901.3) | −80.7 (−1.6) | 5,100.1 (4,260.3, 5,939.9) | −45.4 (−0.9) |
| Disability insurance claims (thousands) | 8,449.9 (6,043.5, 10,856.4) | −86.7 (−1.0) | 8,486.7 (6,082.9, 10,890.5) | −50.0 (−0.6) |
Presence of any of these chronic diseases: diabetes, heart disease, stroke, cancer, lung disease, or hypertension.
Presence of 2 or more of these 6 chronic diseases: diabetes, heart disease, stroke, cancer, lung disease, or hypertension.
ADL, activities of daily living – i.e., daily self-care activities; IADL, instrumental activities of daily living – i.e., those activities that allow an individual to live independently in a community; OASI, Old Age and Survivor’s Insurance; USD, U.S. Dollars.
DISCUSSION
Poor diet quality is a major risk factor for heart disease and other chronic diseases, and diet quality differs significantly across subgroups of U.S. adults by sex, race/ethnicity, and especially by level of education. In this microsimulation, diet quality improvement to the top quintile (Scenario 1) resulted in a reduction of 2.9 million cases (6.5% reduction in incidence) of CHD in Year 10 and 5.5 million cases (9.9%) in Year 30. The smaller diet quality improvement of 2 quintiles (Scenario 2) resulted in a reduction of 1.7 million cases (3.8%) of CHD in Year 10 and million cases (5.7%) in Year 30. Although the largest reductions in U.S. CHD cases and deaths from diet quality improvement would come from targeting the largest group (female, non- Hispanic White or other race, some college), other groups show larger proportional benefits.
Under Scenario 1, the female, non-Hispanic Black, no college group would have the largest proportional reduction in CHD (13.8%). This was not the group with the worst diet quality and such a nonlinearity is not entirely surprising in microsimulation models because interactions between risk factors are allowed—for example, changing diet will have differential outcomes depending on smoking or inactivity.
This simulation takes into account that diet quality differs across individuals, even within a narrowly defined subgroup. This is an improvement over the previous FAM study15 that assumed everyone started at the same average diet quality and moved up an identical amount (to the top quintile).19 Despite its limitations, the previous study’s estimated reductions of 2.2 million cases (5.0%) of CHD in Year 10 and 4.0 million cases (7.2%) in Year 30 fall squarely within this study’s estimates. The current Scenario 1 resulted in larger reductions (as most of the population started with below-average diet quality), and Scenario 2 resulted in smaller reductions in CHD (some people already had a high level of diet quality and could not improve by 2 quintiles).
This paper also has a narrower scope than the previous FAM study, which included published diet quality effects on diabetes, stroke, and cancer,19 even though most were statistically insignificant under that study’s average-to-top-20% diet quality assumption.15 Including diet quality impacts on multiple chronic condition pathways resulted in estimates of other health and economic impacts that were 2–3 times larger than this study. Restricting this analysis to CHD is likely to underestimate the health and economic effects of improved diet quality. Nevertheless, the association between diet quality improvements as measured by HEI-2015 and CVD is robust, making this a credible pathway and starting point.24, 26, 27
Many studies have shown differences in diet quality by sex, race/ethnicity, and education and this paper takes this heterogeneity into account.20, 28–31 The subgroups’ median diet quality estimates are lower and differ substantially from that of the NHS/HPFS samples: 67 for women (NHS) and 68 for men (HPFS).24 The HEI-2015 scores in this study were calculated from NHANES using the methods recommended by the National Cancer Institute as most appropriate for estimating the distribution of scores in the U.S. adult population (repeated 24-hour dietary recalls and Markov Chain Monte Carlo calculation).23, 25 The NHS/HPFS sample scores are based on food frequency questionnaires, which in a validation study gave higher scores.25 However, the more important difference when comparing scores may be that the NHS/HPFS samples are not representative of the U.S. population. The NHS is made up of female nurses where >97% identify as non-Hispanic White and the HPFS is male health professionals with >90% identifying as non-Hispanic White.24 This is likely why the median scores of 2 groups (male and female, non-Hispanic White or other race, some college) have diet quality scores closest to the NHS/HPFS medians.
The National Cancer Institute and the U.S. Department of Agriculture researchers who developed the HEI-2015 proposed the following letter grades to interpret scores in terms of qualitatively describing adherence to the Dietary Guidelines:11 F for scores 0–59, D for scores 60–69, C for scores 70–79, B for scores 80–89, and A for scores 90–100. Using these letter grades, half of U.S. adults have diet quality graded as F. Thus, there is a large potential for diet improvements among the U.S. adult population. Scores vary across individuals and any sociodemographic group includes individuals with both excellent and poor diets, but there are systematic differences across subgroups. Slightly more than half of some groups (men with some college, non-Hispanic White or other race women, and Hispanic women with no college) have diet quality graded D or better, but only 1 group (non-Hispanic White or other race women with some college) had ≥25% with diet quality graded C or better.
Limitations
A major limitation of this work is that scientific uncertainty remains about the health effects of diet quality. The HEI-2015 has received extensive work on measurement, validation of instruments, and alignment with dietary recommendations (which themselves are based on extensive research). Nevertheless, even for this well-studied approach to measure diet quality, the evidence for specific disease pathways is limited and this report focused on only one, CHD. Estimates of risk reduction are not representative of the U.S. population as these come from the NHS/HPFS.
There are also limitations in the simulation approach. FAM is a well-established, internally consistent model built on nationally representative panel data that included personal and household economic indicators as well as health states and sociodemographics. Transition functions are based on historical data and assume that those relationships remain constant during the simulation period—an assumption hard to sustain over 30 years. Two hypothetical improvements in diet quality were modeled, which were assumed to begin in 2019 and be maintained for 30 years. The paper does not address how those changes could come about, nor their costs. In that respect, it is like a cost-of-illness study that shows the potential, but not an operational intervention.
CONCLUSIONS
A microsimulation approach captures the dynamics of population health, which is more realistic than static calculations where interventions immediately result in equilibrium outcomes. Improving a health behavior cannot instantly change the prevalence of chronic disease or retroactively prevent deaths. Therefore, the health and economic benefits of better diet will only accrue over time, and those benefits will not be evenly distributed across a population. Benefits to individuals and groups depend on their current diet quality, the improvement they make, and their other risk factors. In this study, the impact of 2 diet quality improvement scenarios on the variation seen in current diet quality within and across 12 sociodemographic subgroups was examined using a model that also accounts for other determinants of health. Dynamic microsimulation allowed predictions of prevention efforts’ socially important outcomes, most of which occur many years in the future and beyond the scope of trials.
Supplementary Material
ACKNOWLEDGMENTS
We are grateful to Bryan Tysinger for maintaining the microsimulation software and to PhuongGiang Nguyen for assisting Annie Chen in using the microsimulation software.
Funding for the current study was provided through a grant from Pharmavite LLC, a dietary supplements company. The maintenance of the Future Adult Model at RAND is supported by NIH grant R01HD087257. Neither Pharmavite or NIH had any role in the design, analysis, interpretation of results, writing of this article, or the decision to submit for publication.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Credit Author Statement
Conceptualization; Data curation; Formal analysis; Funding acquisition; Investigation; Methodology; Project administration; Resources; Software; Supervision; Validation; Visualization; Roles/Writing – original draft; Writing – review & editing.
Patricia M Herman: Conceptualization; Funding acquisition; Methodology; Project administration; Supervision; Visualization; Roles/Writing – original draft; Writing – review & editing.
Annie Yu-An Chen: Data curation; Formal analysis; Methodology; Software; Validation; Writing – review & editing.
Roland Sturm: Conceptualization; Formal analysis; Funding acquisition; Methodology; Supervision; Validation; Writing – review & editing.
An abstract based on this work was presented at a virtual poster session on October 21, 2021, for the American Public Health Association’s 2021 Annual Meeting and Expo.
No financial disclosures were reported by the authors of this paper.
REFERENCES
- 1.2015‒2020 Dietary Guidelines for Americans 8th Edition. Washington, DC: HHS, U.S. Department of Agriculture; 2015. [Google Scholar]
- 2.2020‒2025 Dietary Guidelines for Americans 9th Edition. Washington, DC: HHS, U.S. Department of Agriculture; 2020. [Google Scholar]
- 3.Micha R, Kalantarian S, Wirojratana P, et al. Estimating the global and regional burden of suboptimal nutrition on chronic disease: methods and inputs to the analysis. Eur J Clin Nutr 2012;66(1):119–129. 10.1038/ejcn.2011.147. [DOI] [PubMed] [Google Scholar]
- 4.Meier T, Gräfe K, Senn F, et al. Cardiovascular mortality attributable to dietary risk factors in 51 countries in the WHO European Region from 1990 to 2016: a systematic analysis of the Global Burden of Disease Study. Eur J Epidemiol 2019;34(1):37–55. 10.1007/s10654-018-0473-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Global Burden of Disease Diet Collaborators. Health effects of dietary risks in 195 countries, 1990‒2017: a systematic analysis for the Global Burden of Disease Study 2017. Lancet 2019;393(10184):1958–1972. 10.1016/S0140-6736(19)30041-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Onvani S, Haghighatdoost F, Surkan P, Larijani B, Azadbakht L. Adherence to the Healthy Eating Index and Alternative Healthy Eating Index dietary patterns and mortality from all causes, cardiovascular disease and cancer: a meta‐analysis of observational studies. J Hum Nutr Diet 2017;30(2):216–226. 10.1111/jhn.12415. [DOI] [PubMed] [Google Scholar]
- 7.Schwingshackl L, Bogensberger B, Hoffmann G. Diet quality as assessed by the Healthy Eating Index, Alternate Healthy Eating Index, Dietary Approaches to Stop Hypertension Score, and Health outcomes: an updated systematic review and meta-analysis of cohort studies. J Acad Nutr Diet 2018;118(1):74–100.e11. 10.1016/j.jand.2017.08.024. [DOI] [PubMed] [Google Scholar]
- 8.Schwingshackl L, Hoffmann G. Diet quality as assessed by the Healthy Eating Index, the Alternate Healthy Eating Index, the Dietary Approaches to Stop Hypertension score, and health outcomes: a systematic review and meta-analysis of cohort studies. J Acad Nutr Diet 2015;115(5):780–800.e5. 10.1016/j.jand.2014.12.009. [DOI] [PubMed] [Google Scholar]
- 9.Morze J, Danielewicz A, Hoffmann G, Schwingshackl L. Diet quality as assessed by the Healthy Eating Index, Alternate Healthy Eating Index, Dietary Approaches to Stop Hypertension Score, and Health Outcomes: a second update of a systematic review and meta-analysis of cohort studies. J Acad Nutr Diet 2020;120(12):1998–2031.e15 10.1016/j.jand.2020.08.076. [DOI] [PubMed] [Google Scholar]
- 10.Guenther PM, Kirkpatrick SI, Reedy J, et al. The Healthy Eating Index-2010 is a valid and reliable measure of diet quality according to the 2010 Dietary Guidelines for Americans. J Nutr 2014;144(3):399–407. 10.3945/jn.113.183079. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Krebs-Smith SM, Pannucci TE, Subar AF, et al. Update of the Healthy Eating Index: HEI-2015. J Acad Nutr Diet 2018;118(9):1591–1602. 10.1016/j.jand.2018.05.021. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Reedy J, Lerman JL, Krebs-Smith SM, et al. Evaluation of the Healthy Eating Index- 2015. J Acad Nutr Diet 2018;118(9):1622–1633. 10.1016/j.jand.2018.05.019. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Abdullah MM, Jones JP, Jones PJ. Economic benefits of the Mediterranean-style diet consumption in Canada and the United States. Food Nutr Res 2015;59:27541. 10.3402/fnr.v59.27541. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Scrafford CG, Bi X, Multani JK, Murphy MM, Schmier JK, Barraj LM. Health economic evaluation modeling shows potential health care cost savings with increased conformance with healthy dietary patterns among adults in the United States. J Acad Nutr Diet 2019;119(4):599–616. 10.1016/j.jand.2018.10.002. [DOI] [PubMed] [Google Scholar]
- 15.Herman PM, Nguyen P, Sturm R. Diet quality improvement and 30-year population health and economic outcomes: a microsimulation study. Public Health Nutr In press. Online January 13, 2021. 10.1017/S136898002100015X. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Lee Y, Mozaffarian D, Sy S, Huang Y, Liu J, Wilde PE, et al. Cost-effectiveness of financial incentives for improving diet and health through Medicare and Medicaid: a microsimulation study. PLoS Med 2019;16(3):e1002761. 10.1371/journal.pmed.1002761. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Mozaffarian D, Liu J, Sy S, et al. Cost-effectiveness of financial incentives and disincentives for improving food purchases and health through the US Supplemental Nutrition Assistance Program (SNAP): a microsimulation study. PLoS Med 2018;15(10):e1002661. 10.1371/journal.pmed.1002661. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Jardim TV, Mozaffarian D, Abrahams-Gessel S, et al. Cardiometabolic disease costs associated with suboptimal diet in the United States: a cost analysis based on a microsimulation model. PLoS Med 2019;16(12):e1002981. 10.1371/journal.pmed.1002981. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Chiuve SE, Fung TT, Rimm EB, et al. Alternative dietary indices both strongly predict risk of chronic disease. J Nutr 2012;142(6):1009–1018. 10.3945/jn.111.157222. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Hiza HA, Casavale KO, Guenther PM, Davis CA. Diet quality of Americans differs by age, sex, race/ethnicity, income, and education level. J Acad Nutr Diet 2013;113(2):297–306. 10.1016/j.jand.2012.08.011. [DOI] [PubMed] [Google Scholar]
- 21.Goldman DP, Leaf DE, Tysinger B. The Future Americans Model: Technical Documentation Los Angeles, CA: Roybal Center for Health Policy Simulation; University of Southern California; 2021. [Google Scholar]
- 22.Chen T-C, Parker JD, Clark J, Shin H-C, Rammon JR, Burt VL. National Health and Nutrition Examination Survey: estimation procedures, 2011‒2014 Hyattsville, MD: National Center for Health Statistics (US); 2018. [PubMed] [Google Scholar]
- 23.National Cancer Institute. Overview and Background of the Healthy Eating Index https://epi.grants.cancer.gov/hei/. Accessed November 21, 2021.
- 24.Shan Z, Li Y, Baden MY, et al. Association between healthy eating patterns and risk of cardiovascular disease. JAMA Intern Med 2020;180(8):1090–1100. 10.1001/jamainternmed.2020.2176. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Subar AF, Potischman N, Dodd KW, et al. Performance and feasibility of recalls completed using the automated self-administered 24-hour dietary assessment tool in relation to other self-report tools and biomarkers in the Interactive Diet and Activity Tracking in AARP (IDATA) Study. J Acad Nutr Diet 2020;120(11):1805–1820. 10.1016/j.jand.2020.06.015. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Hu EA, Steffen LM, Coresh J, Appel LJ, Rebholz CM. Adherence to the Healthy Eating Index-2015 and other dietary patterns may reduce risk of cardiovascular disease, cardiovascular mortality, and all-cause mortality. J Nutr 2020;150(2):312–321. 10.1093/jn/nxz218. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Xu Z, Steffen LM, Selvin E, Rebholz CM. Diet quality, change in diet quality, and risk of incident cardiovascular disease and diabetes. Public Health Nutr 2020;23(2):329. 10.1017/S136898001900212X. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Bosire C, Stampfer MJ, Subar AF, et al. Index-based dietary patterns and the risk of prostate cancer in the NIH-AARP diet and health study. Am J Epidemiol 2013;177(6):504–513. 10.1093/aje/kws261. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Park S-Y, Boushey CJ, Wilkens LR, Haiman CA, Le Marchand L. High-quality diets associate with reduced risk of colorectal cancer: analyses of diet quality indexes in the multiethnic cohort. Gastroenterol 2017;153(2):386–394.e2. 10.1053/j.gastro.2017.04.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Qiao Y, Tinker L, Olendzki BC, et al. Racial/ethnic disparities in association between dietary quality and incident diabetes in postmenopausal women in the United States: the Women’s Health Initiative 1993‒2005. Ethn Health 2014;19(3):328–347. 10.1080/13557858.2013.797322. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Vargas AJ, Neuhouser ML, George SM, et al. Diet quality and colorectal cancer risk in the Women’s Health Initiative Observational Study. Am J Epidemiol 2016;184(1):23–32. 10.1093/aje/kwv304. [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.