Gender-Specific Disparities in Obesity
Tyrone F. Borders, PhD; James E. Rohrer, PhD; Kathryn M. Cardarelli, PhD
Abstract and Introduction
Abstract
Little prior research has investigated whether the correlates of obesity differ between men and women. The objective of this study was to examine gender-specific disparities in obesity by rurality of residence, race/ethnicity, and socioeconomic status. Particular emphasis was devoted to examining potential differences between residents of urban, suburban, and rural areas. Data from the adult version of the 2003 Behavioral Risk Factor Surveillance System (BRFSS) for the state of Texas were used to model the crude and adjusted odds of being obese as compared to normal weight. The findings showed that males of other race/ethnicity had lower adjusted odds of obesity than non-Hispanic whites, but other race/ethnicity was insignificant for females. Females who were Hispanic or black/African American had higher adjusted odds of obesity than non-Hispanic whites, but Hispanic ethnicity and black/ African American race were insignificant for males. Men and women residing in non-metropolitan areas had higher adjusted odds of obesity than their counterparts in metropolitan areas. No economic disparities were revealed among men, but females with high household income had lower odds of obesity than those with low income. Educational status was insignificant for men and women. The findings suggest that programs and policies aimed at curbing obesity should target males and females residing in non-metropolitan localities. Other initiatives should focus on particular groups of women, including those who are Hispanic or black/ African American and have low household income.
Introduction
Obesity is frequently cited as one of America's more pressing public health problems. Although its incidence appears to be steadying,[1] a substantial proportion of adult Americans remain obese. According to estimates from the 2001-2002 National Health and Nutrition Examination
Survey (NHANES), 27.3 percent of adult Americans are moderately or severely obese.[1] Obesity is of national public health concern not only because of its exorbitant prevalence, but also because of its deleterious effects on health status. Several studies have linked obesity to an increased risk of chronic disease,[2] poor health-related quality of life,[3,4] and functional disability.[5] In fact, the public health impact of obesity has been shown to exceed that of two other behavioral problems, smoking and heavy alcohol use.[4]
Because obesity can be considered a personal behavioral choice, one could argue that public interventions are unwarranted. An important argument in favor of governmental involvement concerns the substantial costs attributable to obesity, which have been estimated to range between 5.3 and 5.7 percent of overall Medical care expenditures.[6,7] The Medical care spending associated with obesity is notably higher among members of the government-funded Medicaid (6.7 percent) and Medicare (6.5 percent) insurance programs.[7]
As public health policy makers and practitioners expand programs aimed at combating obesity, additional research is needed to identify high risk population subgroups. Prior research has indicated that residents of rural areas, ethnic and racial minorities, and persons with low socioeconomic status are particularly vulnerable to health problems and the federal government has placed a great deal of emphasis on research and programs aimed at reducing such disparities.[8,9] A report by the U.S. Surgeon General described disparities in the prevalence of obesity,[10] but with the exception of a few studies conducted in the United States[11] and other countries,[12-15] a dearth of research has investigated disparities after adjusting for potential confounders. The present study examined differences in obesity according to rurality of residence, ethnicity or race, and socioeconomic status after controlling for extraneous demographic, social, behavioral, and psychological variables. Because prior research has revealed sex differences in the predictors of obesity,[12] separate models were tested for males and females.
Methods
Data
The data source was the 2003 Behavioral Risk Factor Surveillance System (BRFSS) for the state of Texas. The BRFSS is conducted on an annual basis by the Texas Department of Health in coordination with the Centers for Disease Control and Prevention. A unique aspect of the Texas BRFSS data set used in this study is its inclusion of a variable reflecting the rurality of residence. Over a 12-month period, 6,035 non-institutionalized, civilian adult Texans were sampled via random-digit dialing. The participation rate, calculated as the proportion of individuals who were contacted via the phone and completed the survey, was 70 percent. If fewer than 5 percent of the cases for a given independent variable had missing values, that variable was excluded from the analyses. Cases that had missing BMI values (n = 379) were also excluded, resulting in final unweighted and weighted sample sizes of 5,078 and 13,387,565.
Measures
Body mass index (BMI) was calculated according to self-reports of height and weight and then categorized as underweight (BMI ≤18.5), normal weight (18.5 > BMI < 25), overweight (≥25 BMI < 30), obese (BMI ≥30), and BMI missing. Obese and normal weight persons were compared under the assumption that the achievement of normal weight is a public health goal.
Disparities variables were race/ethnicity (non-Hispanic white vs. Hispanic, black/African American, and other race/ethnicity), rurality of residence (metropolitan central city, metropolitan suburban, and non-metropolitan), household income (<$25,000 vs. $25,000 to $74,999, $75,000 or more, and income missing), and educational status (less than a high school degree vs. A high school degree, some college, and college graduate or more education). Most health-related studies only distinguish metropolitan and non-metropolitan status, which could fail to accurately differentiate residents of urban, suburban, and rural areas. We therefore used a trichotomous classification of metropolitan-central city (an urban core county containing 50,000 or more person), metropolitan-suburban (a metropolitan county adjacent to a core metropolitan-central city area), and non-metropolitan (all other counties). The income missing dummy variable was created to retain those cases that had missing values, an approach that has been used in several prior studies.[16-18]
Control variables included demographic, social, behavioral, and psychological characteristics. Demographic variables were sex (Male vs. Female) and age category (18-24 vs. 25-34, 35-44, 45-54, 55-64, and ≥65). Social variables included marital status (married vs. Single), the number of children in the household (none vs. 1-2 and ≥3), and employment/ working status (employed/working vs unemployed not working). Prior research has demonstrated that the number of children is related to women's weight status.[19-21] Behavioral characteristics included smoking and exercise. In the BRFSS, respondents were asked whether they had ever smoked and, if so, whether they smoked cigarettes every day, some days, or not at all. Responses were recategorized to distinguish current (smoke every day or some days) and past smokers. Respondents were also asked several questions about their physical activity during a usual week: if they engage in moderate physical activity (PA) at least ten minutes at a time, if they engage in vigorous PA for at least ten minutes, the number of times per week the engage in PA, and the total time per day they engage in PA. Federal guidelines recommend at least 30 minutes of moderate exercise for 5 days per week or 20 minutes of vigorous exercise for 3 days per week.[22] Three exercise groups were computed: none, some, and meets recommendations. Self-rated mental health was assessed with the following question: ''Now, thinking about your mental health, which includes stress, depression and problems with emotions, for how many days during the past 30 days was your mental health not good?'' Poor self-rated mental health was defined as having 14 or more days of ''not good'' mental health.
Statistical Analyses
STATA 8.0[23] was used to conduct descriptive and multivariate analyses and account for the complex sampling scheme and population weights. The frequency of each disparities variable and covariate (demographic, social, behavioral, and psychological factors) was calculated for normal weight and obese individuals. Univariate and multivariate logistic models were conducted by gender. Univariate logistic regression analyses were conducted to model the association between each disparities variable and obesity, respectively. Multivariate logistic regression analyses were conducted to model the association between each disparities variable and obesity while adjusting for demographic, social, behavioral, and psychological covariates.
Results
Of the 5,078 respondents included in the analyses, 36.48 percent were normal weight and 25.03 percent were obese. While not reported in the table, 2.13 percent were underweight and 36.36 percent were over-weight. Disparities variables and demographic and social covariates are shown for the overall sample and for normal weight and obese individuals in Table 1 . Behavioral and psychological covariates are shown for the overall sample and by BMI status in Table 2 .
Although the findings are not reported in the tables presented in this paper, we did test for differences in the odds of obesity between men and women. Males were found to have increased crude (OR = 1.27, 95% CI = 1.07, 1.50) and adjusted odds (OR = 1.63, CI = 1.36, 1.96) of obesity as compared to females.
Table 3 shows crude and adjusted odds of obesity for the disparities variables by gender. Among males, Hispanic ethnicity and black/African American race were not significantly associated with obesity. However, other race/ethnicity was significantly associated with lower crude and adjusted odds of moderate obesity as compared to non-Hispanic whites. The findings for females were noticeably different. Compared to non-His-panic white females, Hispanic and black/African American females had higher crude and adjusted odds of obesity. Other race/ethnicity was insignificant for females.
Males living in non-metropolitan areas had higher crude and adjusted odds of moderate obesity than males living in metropolitan-central city areas. Similarly, females residing in non-metropolitan areas had higher adjusted, but not crude, odds of obesity than females residing in metropolitan-central city areas. There were no significant differences between males or females residing in metropolitan-central city and suburban areas.
The associations between income and obesity differed between males and females. Among males, those who had household incomes of $25,000 to $74,999 had higher crude, but not adjusted, odds of obesity as compared to those with incomes of $25,000 or less. Females with household incomes of $75,000 or more had lower crude and adjusted moderate odds of obesity.
Educational status was insignificant at the 0.05 alpha level in the models for females and males. However, having a college degree or more education was associated with lower adjusted odds of obesity at the 0.10 level for males and females.
Discussion
Despite much recent attention devoted to health disparities in the United States,[10] little research has investigated differences in obesity according to rurality of residence, race or ethnicity, and socioeconomic status (SES). A limited number of studies foreign to the United States have examined the social and economic correlates of obesity,[12-15] but whether their findings are relevant to Americans is questionable, especially because of cross-country differences in ethnic and racial composition. Analyses of the National Health and Nutrition Examination Survey (NHANES) have demonstrated disparities in the prevalence of obesity by race, ethnicity, and economic status,[24] but have not controlled for the full range of factors known to influence obesity, including exercise and mental health. A recent study using data from the National Health Interview Survey (NHIS) examined differences in obesity by the rurality of residence while controlling for selected confounders, but did not stratify analyses by gender.[11] Thus, the present study lends further insight into gender-specific disparities in obesity.
Regardless of gender, residents of non-metropolitan counties appear to have a somewhat higher risk of obesity than persons residing in metropolitan areas. This finding is supported by at least one national[11] and a limited number of state or regionally specific studies.[25-27] Higher rates of obesity among rural residents may be attributable to poor health behaviors which extend from childhood to adulthood, including poor dietary intake[28] and sedentary lifestyles.[11,29] Others have speculated that a general shift from farming to manufacturing-based economies in rural areas explains the higher prevalence of obesity among rural residents.[30]
According to our data, race and income are stronger risk factors for obesity among women than non-metropolitan residence. As mentioned earlier, Hispanic and black/African American females were more likely to be obese than non-Hispanic whites, even when controlling for other potential factors. Similarly, descriptive findings from the NHANES indicate that Mexican-American and black/African American women have a lower prevalence of overweight and obesity than their non-Hispanic white counterparts.[10,24] Yet, the findings for men were noticeably different. Possible explanations include racial or ethnic differences in body image expectations. For example, in Texas, perhaps the thin idea is more common among non-Hispanic white women. On the other hand, the Male-beauty ideal of a large torso may apply to all races and ethnic groups.
Higher household income protected females against obesity, but not males. This finding is coherent with other studies that consistently find an inverse association between high SES and obesity among women, but not men.[31,32] One possible explanation for such a difference may be that family and social pressures to conform to a certain body image exert a stronger influence in women of higher SES.[32] Another explanation is that women of higher SES are more likely to engage in preventive health behavior, such as regular exercise and better diet, compared to their Male counterparts.[33] Finally, women of higher SES may be less likely to have high parity,[34] which is associated with overweight and obesity.[20,21] In a population-based study of Swedish women, investigators found that of the factors that accounted for the SES-obesity association, reproductive history (chiefly, early menarche and high parity) was the strongest explanatory factor, accounting for 44 percent of the variance.[35]
Because we focused on the state of Texas, the generalizability of the findings may be limited to areas with similar geographic, demographic, and social characteristics. However, because the CDC mandates that each state conduct the BRFSS on an annual basis, health researchers in states other than Texas could replicate our analyses to identify subgroups of their own populations which have a high risk of obesity. To facilitate similar comparisons in other states or for the nation as a whole, the CDC and affiliated state agencies will need to provide indicators of rural, suburban, and urban residence (most states do not make such indicators publicly available). Another drawback to the BRFSS is that it relies on self-reports of height and weight, which are used to calculate BMI. Prior research based on the national sample of BRFSS respondents suggests that self-reports are less valid than objective height and weight measures,[36] although the reliability of the BRFSS measures for height, weight, and BMI were found to have correlation coefficients of 0.84 to 0.94.[37] Other national telephone surveys have yielded prevalence rates similar to those from the NHANES,[4] suggesting that valid estimates of BMI can be obtained through self reports. Because our primary objective was not to estimate the prevalence of obesity in Texas, whether the Texas BRFSS accurately represents adult Texans is of less concern than whether the associations found in the present study are valid. Despite the aforementioned limitations of the BRFSS, it remains the most feasible means of evaluating health problems at the state level.
In summary, our findings support that some programs should be directed toward residents of non-metropolitan areas. The main thrust of the obesity campaign should selectively target specific groups of men and women, particularly Hispanic, black/African-American, and poor females. Targeting these risk groups is not only more efficient, but allows for the implementation of tailored interventions.
Reprint Address
Tyrone F.Borders, PhD Department of Health Policy and Management, Fay W. Boozman College of Public Health, 4301 W. Markham St., Slot 820, Little Rock, AR 72205-7199; e-mail: tfborders@uams.edu
Enviado por Dr. José Manuel Ferrer Guerra