Obesity-Related Health Lifestyles of Late Middle-Age Black Americans: The Jackson Heart Study

Abstract

Introduction:

This paper examines the obesity-related health lifestyle practices of a late-middle age cohort of socioeconomically diverse Black Americans. Black people have the highest prevalence of obesity of any racial group in the U.S. Consequently, the obesity-related health lifestyles of this population is an important topic of investigation, including those in late-middle age for whom there is little data.

Methods:

This study employs latent class analysis (LCA) and multinomial logit models to investigate dietary habits, levels of exercise, alcohol use, and smoking. The analysis sample is from the first exam of the Jackson Heart Study (2000–2004) analyzed in 2021 using LCA. The sample consists of 739 Black men and 1,351 women between the ages of 50 and 64.

Results:

Three classes of lifestyles were found for both genders: healthy diet, unhealthy diet, and unhealthy smokers. For women only, a most healthy lifestyle was added. Major findings are the low levels of physical activity, a clear socioeconomic pattern in health lifestyles among Black men and women, and the association of diagnoses of diabetes and CVD with healthier lifestyle practices among Black men but not women.

Conclusions:

Obesity-related health lifestyles among late-middle age Black Americans generally do not converge toward a healthier norm with impending old age. An exception is men who have been diagnosed as having diabetes or heart disease. Otherwise, healthy and unhealthy lifestyle practices remain aligned by social class during this period of the life course.

INTRODUCTION

Nationally, non-Hispanic Black adults have the highest prevalence of obesity (nearly 50%) of any racial group in the U.S.¹ Black women have the highest prevalence of obesity (nearly 57%) by sex and race.¹ Obesity was also a significant comorbid variable in the high mortality of Black adults from COVID-19.^2–4 Consequently, the obesity-related health lifestyles of Black people are a significant topic of investigation, including those in late-middle age for whom there is a paucity of data.^5,6

Late-middle age is a period in the life course when an individual’s weight is usually well-established and chronic diseases linked to the cumulative effects of disadvantaged social conditions, stress, and constraints on health-promoting behaviors often surface, regardless of race.⁵ Obesity, heart disease, diabetes, high blood pressure, breathing difficulties, and various other afflictions connected to smoking, lack of a nutritious diet, excessive alcohol consumption, and lack of exercise become more common during this stage of life.^5,6 Given that Black Americans have the highest mortality rates in the U.S. from chronic diseases,^7–9 it is important to understand the combinations of obesity-related health behaviors that coalesce into health lifestyles for this group.

Past studies have largely focused on comparing individual health behaviors across racial/ethnic groups. Research shows that Black people are not as likely as White people to indulge in heavy episodic drinking,¹⁰ but are more likely to have a less healthy diet^11–15 and not exercise.^16–18 Proportionately more Black men smoke than either White men or White women, but Black women smoke less than Black men and White people generally.^19–23

However, in many studies, Black Americans and members of other racial groups have low representation. Instead, they are embedded, sometimes in small numbers, within much larger and predominantly White samples, serving as “token” minorities in some analyses.²⁴ In such instances, general conclusions about Black American health practices are based on relatively small samples. Moreover, a related criticism of past samples of Black Americans is that low-income individuals are frequently overrepresented and the more affluent underrepresented, possibly skewing results to be more characteristic of disadvantaged Black people than the Black population as a whole.^25–27 Samples failing to capture the diversity of socioeconomic backgrounds among Black people do not provide the most representative results,^24,25 which this study addresses by analyzing an all-Black sample of varying SES.

The purpose of this study is to examine the obesity-related health lifestyles of a late-middle age cohort of socioeconomically diverse Black Americans. Latent class analysis (LCA) and health lifestyle theory is used to investigate health lifestyle practices associated with obesity among Black adults in the Jackson Heart Study (JHS). These data were collected from the tri-county (Hinds, Madison, and Rankin) area of Jackson, Mississippi, which is noteworthy because Mississippi has the highest mortality rates for Black people of any state in the U.S. (801.0 deaths per 100,000 in 2017), especially from coronary heart disease (CHD) and Type 2 diabetes.^9,28 Mississippi also has the second-highest prevalence (46.0%) of obesity for non-Hispanic Black people in the nation after West Virginia.²⁹

The theoretical perspective guiding this analysis is health lifestyle theory.^30–32 This theory defines health lifestyles as collective patterns of health-related behavior based on choices from options available to people according to their life chances. The basic orientation of health lifestyle theory is that the behavioral choices people make that affect their health either positively or negatively—namely smoking, consuming alcohol, eating particular foods, exercising, and the like—cluster into distinct lifestyle configurations characteristic of particular groups and social classes. Although these lifestyle practices are enacted by the individual, they are affected (enabled or constrained) by what is available with respect to the norms, values, and material resources associated with the individual’s SES and other variables consistent with the individual’s living situation or chances in life. The choices typically align the chooser with others sharing the same or similar social backgrounds.

The practices that result can either impair, maintain, or promote health over time. However, even though these health behaviors have a general binary character (positive or negative), they may not be exclusively one way or the other in differentiating between lifestyles. Instead, health lifestyles often include a mixture of health-promoting and health-harming behaviors that are nevertheless aligned along social gradients with the most healthy practices invariably associated with higher SES and the least healthy with lower SES.^32–36 Health lifestyle theory is described more fully elsewhere.^30–32

This study moves beyond past research in 3 ways. First, rather than analyzing individual health behaviors related to obesity, this research investigates health lifestyles that capture the relatively durable combinations and mutually reinforcing nature of various health behaviors. Second, the focus is on predictors of health lifestyles among Black men and women rather than across races. Because of the unique life course processes involving stress related to discrimination, systemic inequities, and structural racism as well as experiences with the healthcare system among Black people,^37–40 this study seeks to determine whether SES, a key predictor of health lifestyles, has a different result in an all-Black sample than found in Black-White comparisons. Third, attention is given to late middle-age as an important period of the life course when chronic illnesses begin to emerge.

METHODS

The JHS is a single-site prospective epidemiological cohort study focusing on the risk factors of coronary heart disease among Black men and women. The original sample consisted of 5,306 persons between the ages of 20 and 95. Details of the study’s design are published elsewhere.⁴¹ Data were collected in 3 exams—(1) 2000–2004, (2) 2005–2008, and (3) 2009–2013—and the study continues. Overall, the sample is relatively well-educated, with 56% having a college degree or higher and 6% having some college compared to 38% with a high school diploma or less. In addition, over a quarter of the sample have family incomes of $50,000 or more annually at baseline (2000–2004), with over 8% earning $75,000 or more, compared to about 31% who have an income of less than $16,000 per household annually. Consequently, the sample is not overrepresented by low SES individuals, but includes a majority of middle class participants.

As the focus is on late-middle age, the analysis was restricted to 739 Black men and 1,351 Black women between the ages of 50 and 64 at baseline. The mean age of participants in the study sample is around 57 years, with nearly 60% of both sexes having a 4-year college degree. A small number of participants were excluded (N=22, <1%) due to having unclassifiable occupations. With the exception of income (missing for 15%), covariates were missing for less than 5% of participants. Multiple imputation was used to address missing data. Stata’s suite of mi commands was used to construct 25 complete datasets via multiple imputation with chained equations.⁴² Diagnostics of the imputations indicated minimal departures from the distributions of the observed data.⁴³

Measures

Five indicators of obesity-related health lifestyles collected at baseline (2000–2004)^a were utilized: (1) eating the American Heart Association (AHA) recommended amount (4.5 cups) of fruit/vegetable servings per day and (2) drinking no more than the AHA recommended amount of (less than 36 ounces or 450 kcals) of sugary beverages per week, along with (3) meeting the AHA’s Life’s Simple 7 metric for ideal physical activity, (4), alcohol consumption, and (5) smoking. These 5 health behaviors have strong relationships with obesity and later life health outcomes.^30–34

The AHA metric for ideal physical activity is based on the number of minutes per week that participants report engaging in sport/exercise activities. Participants were ranked as having a poor (0 minutes in moderate or vigorous physical activity per week), intermediate (between 0 and 150 minutes in moderate or between 0 and 75 minutes in vigorous physical activity per week), or ideal (more than 150 minutes in moderate or more than 75 minutes in vigorous physical activity per week) exercise regimen. Following this metric, poor or intermediate physical activity was coded as failing to meet the criteria and ideal physical activity as meeting the criteria.

For drinking, a measure was constructed distinguishing between abstainers (0 drinks daily), light to moderate drinkers (1–3 drinks daily for men and 1–2 drinks daily for women), and heavy drinkers (more than 3 drinks daily for men and more than 2 drinks daily for women). For smoking, a measure was developed differentiating between participants who never smoked, smoked in the past but were not current smokers, and current smokers. For both drinking and smoking the 3 categories are treated as nominal rather than ordinal in the measurement of health lifestyles.

The analysis draws on 4 measures of SES collected at baseline—education (years of schooling), logged income, occupation, and self-reported subjective social class. Logged income is based on an 11-category measure of household income. The categories were recoded to midpoints with the top code set to $125,000 and then logged. As for occupation, the vast majority of participants (99%) fell into 1 of 3 occupation categories: (1) management/professional, (2) sales/service, and (3) production/construction. Subjective social status is based on a question that asked participants to rate their standing in their community on a 1 to 10 scale.

Additional correlates include age and whether participants have been told by a physician they are in a pre-diabetic stage or have Type 2 diabetes, and whether they have a history of coronary heart disease (CHD). The medical histories are self-reported.

Statistical Analysis

The analysis proceeds in 2 steps. The first step involves identifying health lifestyles based on the 5 indicators of diet, exercise, smoking, and drinking. This step is accomplished with latent class analysis (LCA).⁴⁴ LCA treats the individual health behaviors as indicators of an underlying categorical latent variable. In this context, the categories of the latent variable represent distinct health lifestyles composed of different combinations of health behaviors. The number of health lifestyles is determined by balancing the latent class model that has the best fit with the data based on several model fit statistics and considerations of parsimony and interpretation. This analysis relies on the Akaike Information Criterion (AIC), the Bayesian Information Criterion (BIC), and the entropy.⁴⁴

The second step of the analysis involves examining the correlates of the health lifestyles that emerge in the first step. This component of the analysis relies on multinomial logit models predicting membership in a specific health lifestyle. Participants were assigned to the health lifestyle for which they had the highest estimated probability of membership.^b

All analyses are stratified by gender as women are generally found to have healthier lifestyles compared to men.^32–34,37 All analyses were conducted in Stata 16, use 25 imputed datasets and report the average parameter estimates across all of the complete datasets along with corrected SEs.

RESULTS

Table 1 provides descriptive statistics for all of the analysis variables. It shows late middle-aged Black men generally engaging in worse health behaviors than Black women, in that a higher proportion of men smoke currently or in the past, are heavy or moderate drinkers, and a lower proportion meet the recommended fruit/vegetable intake than women. A somewhat higher proportion of men than women, however, meet the recommended guidelines for exercise.

Table 1.

Means or Proportions for Indicators of Health Lifestyles, Sociodemographic Measures, and Medical History by Sex

Variable	Overall (N=2,090)	Men (N=739)	Women (N=1,351)	p-value for difference
Health lifestyle indicators
Smoke: never	0.64	0.50	0.71	<0.001
Smoke: past	0.21	0.30	0.16	<0.001
Smoke: current	0.15	0.19	0.12	<0.001
Drink: none	0.59	0.42	0.68	<0.001
Drink: moderate	0.38	0.53	0.30	<0.001
Drink: heavy	0.03	0.05	0.02	<0.001
Exercise: yes	0.18	0.21	0.16	0.007
Fruit/Vegetable: yes	0.72	0.63	0.77	<0.001
Sugary beverage: yes	0.46	0.45	0.46	0.744
Sociodemographic measures
Age	57.80	57.60	57.91	0.143
Education: ≥16 years of schooling	0.58	0.59	0.58	0.906
Log income	10.44	10.67	10.31	<0.001
Occupation: management/professional	0.38	0.32	0.41	<0.001
Occupation: service/sales	0.40	0.27	0.47	<0.001
Occupation: construction/production	0.22	0.41	0.12	<0.001
Subjective social status	7.86	7.89	7.85	0.629
Diabetes: no	0.36	0.36	0.36	0.661
Diabetes: pre	0.37	0.40	0.35	0.034
Diabetes: yes	0.26	0.24	0.27	0.156
History of CHD: yes	0.12	0.18	0.09	<0.001

Notes: Boldface indicates statistical significance (p<0.05). Sugary beverages, fruit/vegetables, and exercise coded such that “yes” indicates participants met healthy guidelines. P-value for difference between sexes based on 2-tailed test for difference in means or proportions as appropriate.

CHD, coronary heart disease.

The model fit statistics for latent class models allow from 2 to 5 health lifestyles (Table 2). For men, the minimum AIC is the model allowing for 3 health lifestyles, and the minimum BIC is the model allowing for 2 health lifestyles. For women, there is a similar pattern with models allowing for 4 and 2 health lifestyles, respectively. The model allowing for 3 health lifestyles for men and 4 health lifestyles for women were selected as offering the best fit with the data and a parsimonious, interpretable set of health lifestyles. In addition, for both men and women, the entropies are high (0.82) and indicate that the model performs well in classifying respondents into one of the health lifestyles.

Table 2.

Model Fit Statistics for Latent Class Models

Variable	AIC	BIC	Entropy
Men (N=739)
2-class	5,268	5,337	0.40
3-class	5,259	5,356	0.82
4-class	5,263	5,397	0.49
5-class	5,259	5,415	0.48
Women (N=1,351)
2-class	8,159	8,237	0.81
3-class	8,147	8,262	0.69
4-class	8,130	8,287	0.82
5-class	8,130	8,297	0.68

AIC, Akaike Information Criterion; BIC, Bayesian Information Criterion.

Figures 1 and 2 illustrate the conditional item response probabilities for the health lifestyle indicators estimated for each lifestyle among men and women, respectively. This information depicts the general characteristics of each health lifestyle class that allow labels to be developed. For the men, 3 labels capture the salient characteristics of each lifestyle: (1) healthy diet (HD), (2) unhealthy diet (UHD), and (3) unhealthy smokers (UHS). For women, there is a similar set of labels with the addition of a (4) most healthy (MH) lifestyle. The modal lifestyle of healthy diet or HD for men and women (characteristic of a majority, some 55% of men and 77% of women) in this sample has a relatively healthy profile—low rates of smoking and heavy drinking coupled with reasonably positive dietary practices, though low rates of exercise.

Figure 1.

Illustration of item response probabilities and prevalence for each of the latent classes for men.

Notes: For each of these indicators, Y=met healthy guidelines. For smoking, N=never smoked, P=smoked in the past, and C=current smoker. For drinking, A=abstains, M=light to moderate drinking, and H=heavy drinking. See discussion in text for details about coding for each of these indicators.

Ex, exercise; FV, fruits and vegetables; SB, sugary beverages.

Figure 2.

Illustration of item response probabilities and prevalence for each of the latent classes for women.

Notes: For each of these indicators, Y=met healthy guidelines. For smoking, N=never smoked, P=smoked in the past, and C=current smoker. For drinking, A=abstains, M=light to moderate drinking, and H=heavy drinking. See discussion in text for details about coding for each of these indicators.

Ex, exercise; FV, fruits and vegetables; SB, sugary beverages.

The unhealthy diet (UHD) lifestyle for both men and women is characterized by relatively poor diets in that it has the lowest probabilities of healthy fruit/vegetable consumption (pr<0.01 for both men and women) and highest consumption of sugary beverages (pr=0.35 for men and pr=0.16 for women). This lifestyle, however, has negligible probabilities of current smoking and heavy drinking.

The chief distinguishing feature of the unhealthy smokers or UHS lifestyle for both men and women is the very high probability of current smoking (pr=0.86 for men and pr=0.85 for women) and the highest but still low probability of heavy drinking (pr=0.13 for men and pr=0.12 for women) as compared with the other lifestyles. This lifestyle also comes with low probabilities of exercise (pr=0.12 for men and pr=0.02 for women), but moderate probabilities of healthy fruit/vegetable consumption (pr=0.53 for men and pr=0.38 for women) and sugary beverages (pr=0.61 for men and pr=0.35 for women).

Finally, it is notable that a broadly healthy lifestyle emerged among a small percentage of women (only 2%) and not at all among men. This most healthy lifestyle or MH among women has a high probability of exercise (pr=0.94) and healthy fruit/vegetable consumption (pr=0.84) along with limited consumption of sugary beverages (pr=0.93). This lifestyle is also characterized by high rates of moderate drinking (pr=0.89). None of the health lifestyle configurations are exceedingly healthy, but the MH lifestyle for a few women stands out as the healthiest by far.

Table 3 reports estimates of average marginal effects (ame) from multinomial logit models predicting membership in each health lifestyle (Appendix Table 1 provides descriptive statistics by health lifestyle). Beginning with men, subjective social status is found to be unrelated to health lifestyle membership, but objective measures of SES were associated with such memberships. In particular, income is associated with a higher probability of being a member of the healthy diet (HD) lifestyle (ame=0.07) and a lower probability of being a member of the unhealthy smoker (UHS) lifestyle (ame= −0.04). Similarly, working in production/construction jobs is associated with a lower probability of being a member of the healthy diet (HD) lifestyle (ame= −0.10). This pattern is consistent with the association of lower SES with less healthy lifestyles. With respect to medical histories, being pre-diabetic is associated with a lower probability of the UHS lifestyle (ame= −0.07), being diabetic is associated with a higher probability of the HD lifestyle (ame=0.10), and a history of coronary heart disease is associated with a lower probability of the UHD lifestyle (ame= −0.11).

Table 3.

Average Marginal Effects From Multinomial Logistic Regressions Predicting Health Lifestyles

Variable	Men (N=739)			Women (N=1,351)
	UHS 19%	UHD 27%	HD 55%	UHS 10%	UHD 11%	HD 77%	MH 2%
Age	−0.006 (0.003)	−0.002 (0.004)	−0.008 (0.004)	−0.004 (0.002)	−0.003 (0.002)	0.006 (0.003)	0.001 (0.001)
≥16 years schooling	0.007 (0.034)	−0.028 (0.040)	0.021 (0.045)	0.001 (0.020)	0.020 (0.022)	−0.009 (0.030)	0.030 (0.015)
Log income	−0.044 (0.019)	−0.026 (0.024)	0.070 (0.027)	−0.035 (0.010)	0.003 (0.011)	0.025 (0.015)	0.008 (0.006)
Service/sales	0.047 (0.036)	−0.051 (0.044)	0.004 (0.049)	0.041 (0.019)	−0.021 (0.021)	−0.017 (0.028)	−0.003 (0.008)
Construction/production	0.090 (0.038)	0.010 (0.047)	−0.100 (0.051)	0.044 (0.030)	0.068 (0.038)	−0.122 (0.046)	0.009 (0.020)
Subjective social status	−0.012 (0.008)	0.003 (0.009)	0.008 (0.010)	0.001 (0.004)	−0.001 (0.004)	0.001 (0.006)	−0.001 (0.002)
Diabetes: pre	−0.074 (0.034)	−0.007 (0.038)	0.081 (0.042)	0.004 (0.020)	0.043 (0.021)	−0.042 (0.021)	−0.004 (0.009)
Diabetes: yes	−0.065 (0.038)	−0.036 (0.043)	0.101 (0.049)	−0.016 (0.020)	−0.001 (0.021)	0.029 (0.028)	−0.013 (0.008)
CHD: yes	0.054 (0.036)	−0.108 (0.048)	0.054 (0.050)	0.085 (0.022)	0.015 (0.028)	−0.106 (0.036)	0.007 (0.013)

Notes: Boldface indicates statistical significance (p<0.05). Unstandardized estimates of average marginal effects with SEs in parentheses. Estimates based on 25 multiple imputation datasets.

HD, healthy diet; UHD, unhealthy diet; UHS, unhealthy smokers; MH, most healthy lifestyles; CHD, coronary heart disease.

Turning to women, age has a positive association with the HD lifestyle (ame=0.01). However, there is a different pattern of associations involving SES than observed among men. The results favor higher SES individuals as education rather than income is associated with a higher probability of the MH lifestyle (ame=0.03) and working in construction/production is associated with a lower probability of the HD lifestyle (ame= −0.12). In contrast to men, women reported as pre-diabetic are associated with a higher probability of the UHD lifestyle (ame=0.04), which might explain why they are pre-diabetic. In comparison, women with coronary heart disease are associated with a lower probability of the HD lifestyle (ame= −0.11) and a higher probability of the UHS lifestyle (ame=0.09). Here again, less healthy lifestyle practices (low likelihood of a healthy diet and high likelihood of smoking) are linked to diagnoses of coronary heart disease.

DISCUSSION

Higher SES Black individuals were found to have healthier lifestyle practices than those lower on the socioeconomic scale. For the entire sample, levels of exercise were low, but the highest scores for exercise nevertheless accrued to those of higher SES. Late middle-age college-educated Black women as well as Black men with professional and managerial jobs and higher incomes, were most likely to engage in the healthiest practices involving obesity. Overall, the findings show that the health lifestyles of Black adults in late middle-age are aligned by SES consistent with health lifestyle theory.

Gender was also relevant in that late middle-age Black women were more likely to have healthier lifestyle practices than Black men. They drank and smoked less and had healthier food habits than men. For both genders, obesity-related health lifestyle practices are generally mixed (both good and bad) in late middle-age, thereby indicating they have not coalesced toward a healthier norm at this time of life. Most past research on health lifestyles has focused on earlier stages of the life course and assumed health lifestyles formed in adolescence and young adulthood persist later in life.^33–36 The finding of a general lack of healthier lifestyles in late middle-age in this sample of Black Americans suggests the possibility of less healthy lifestyles at earlier ages, which could in part reflect lifelong exposures to stress and discrimination.

However, the analysis revealed opposite associations between diagnoses of chronic conditions and health lifestyle practices for men and women. Among men, the associations are consistent with the probability that men change their health behaviors towards healthier lifestyles following a diagnosis of diabetes or coronary heart disease. In contrast, among women, there is no evidence of this. These patterns may reflect different levels of severity or different timings of diagnoses (earlier for men than women) in late middle-age. Nevertheless, the diagnoses of diabetes and heart disease among women were not associated with healthier lifestyles in the same way observed among men. Additional research is needed to determine whether the association among men represents a causal effect and the mechanisms that could underlie gender differences.

Limitations

This study has a few limitations. First, latent class analysis is inductive with respect to identifying health lifestyles. As such, one would not necessarily expect to observe these same health lifestyles in all Black populations. Rather, we view this study joining others using LCA to identify health lifestyles in different populations that will ultimately form the basis for generalized deductive analyses. Second, the measures of diet were limited to sugary drinks and fruit/vegetable consumption. More dietary measures would have been preferable. Third, the study design is cross-sectional which limited the ability to draw causal inferences. Fourth, this study was conducted in a single metropolitan area in the southeastern U.S., possibly limiting its generalizability to other black populations.

CONCLUSIONS

Despite these limitations, there are notable strengths. First, the JHS includes a large proportion of well-educated, middle-class Blacks that have often been underrepresented in large surveys. This provided an opportunity to assess the lifestyles of a more sizable and broader range of SES among Blacks than seen in most studies. Second, a novel but increasingly prevalent statistical measure (LCA) was used to examine the association of SES with obesity-related lifestyle behaviors in an all-Black sample. Third, as suggested by health lifestyle theory, SES appears to operate within this sample of Blacks much the same as other studies show it does among Whites in patterning health lifestyles. Fourth, major findings are the low levels of physical activity, a clear socioeconomic pattern in health lifestyles among Black men and women, and the association of diagnoses of Type 2 diabetes and CHD with healthier lifestyle practices among Black men but not women. These findings lay the groundwork for future research on late middle-age health lifestyles that include Black Americans.