Over two-thirds (67%) of US adults are obese or overweight. The immediate cause of obesity is sustained positive energy balance, where energy intake exceeds energy expenditure. However, a variety of individual and social factors also contribute to obesity. At the individual level, gender, race, and socioeconomic status are important predictors of obesity. For example, women are more susceptible to weight gain [
Risks of obesity also vary at the neighborhood level, and poor neighborhoods bear the disproportionate burden. Due to ongoing residential segregation, poor and minority individuals may be clustered in “obesegenic” neighborhoods that promote and sustain obesity. Neighborhood can influence the likelihood of obesity by influencing behavioral norms, access to food, and opportunities for physical activity [
There is a growing body of literature emphasizing the adult health risks of childhood overweight. Children who are overweight are at an increased risk for obesity and its associated conditions such as hypertension, hyperlipidemia, and diabetes [
We used a unique longitudinal cohort of inner city women and their children in order to determine how socioeconomic factors across the lifespan contribute to adult obesity. Our three aims were (1) to quantify the association between individual and neighborhood poverty and obesity in women; (2) to examine whether individual and neighborhood poverty in childhood influenced the likelihood of adult obesity; (3) to characterize whether these relationships varied between African American and White women.
We used a subset of a longitudinal study, the Johns Hopkins Perinatal Collaborative Study (PCS), and the Pathways to Adulthood (PTA) followup. These studies followed three generations of families initially living in inner city Baltimore. The Perinatal Collaborative Study enrolled 2307 inner-city women (referred to throughout as first-generation mothers [G-1s]) who were selected at the time of their first prenatal visit to a public obstetric clinic at Johns Hopkins Hospital between 1959 and 1965 [
From 1992 to 1994, the Pathways to Adulthood Study (PTA) collected additional information from 1758 G-2s (then aged 27 to 33) about their lives from age 9 to present. Followup data included information on education, employment, family composition, health, health care usage, and income [
Our final sample included the 986 female G-2s (75%) who provided information (in person or via telephone) for the Pathways to Adulthood Study, had information on self-reported height and weight, and could be linked to the Perinatal Collaborative Study for data on their childhood clinical and sociodemographic characteristics. We excluded women who were pregnant at the time of the interview (
Obesity was our outcome variable which we defined as a body mass index (BMI) greater than or equal to 30. Information on height (feet and inches) and weight (lbs) was obtained by self-report. First, we calculated BMI as a continuous linear outcome variable according to the formula: (weight (lbs)/height (inches)2) × 703. Then we defined our categorical outcome variable—nonobese versus obese—using cutoffs consistent with those used by the Centers for Disease Control, the World Health Organization, and the National Institutes of Health [
We adjusted for individual characteristics including demographic (age, race, number of children, age at birth of first child, and marital status) and health related (current smoking status and self-reported health) characteristics.
As an indicator of socioeconomic status (SES), we used years of education and homeownership. Since little consensus exists in the literature regarding how best to measure SES, we also examined education, income, and assets as potential measures of SES. We treated years of education as a continuous variable and also coded as a binary variable for college graduate or not. Although we had two measures for income, self-reported total household income and personal income, the high rate of missing values precluded their use. We had six measures of assets. We treated assets as a continuous variable ranging from 0 to 6, with one point given for a “yes” response when asked about each of six assets (current personal checking account; current IRA or pension; own house or condo; car, truck, or motorcycle ownership; credit or charge account; current savings account). Homeownership was the most robust asset measure with the largest effect in both magnitude and statistical significance. Therefore, we used homeownership as our additional SES measure.
We derived G-2s adult neighborhood characteristics from the 1990 census data. The PTA data link each respondent’s address at the time of the interview to the appropriate census tract. For measures of neighborhood SES, we examined median household income for census tract as a continuous variable and also categorized neighborhoods based on percent of respondents below federal poverty level (nonpoor [less than 20%] and poor [>20%]). We based our categories of percent poverty on previous literature and the Census Bureau definition of poverty areas as those in which at least 20% of the population lives below the federal poverty line [
We used childhood BMI at age 7 as a covariate. BMI was categorized as underweight, normal weight, at-risk for overweight and overweight according to CDC BMI-for-age percentile formulas [
Individual childhood poverty status was characterized by a binary variable to indicate whether the individual was ever poor in childhood. In addition, we used parental homeownership as a proxy for assets.
Childhood neighborhood characteristics were based on G-2 report at age 8 and were derived from 1970 census data by the PTA investigators. Categories for neighborhood racial composition and neighborhood poverty were consistent with those mentioned above. We based neighborhood racial composition on percentage of African American residents. There was a binary variable for African American neighborhood [≥90%] or not [<90%]. We also categorized neighborhoods based on percent of respondents below federal poverty level (nonpoor [less than 20%] and poor [>20%]).
We conducted univariate and multivariate analysis for our outcome variables and our covariates of interest. For univariate analyses, we compared baseline characteristics for African American and White women in childhood and adulthood. Then we conducted a bivariate analysis to compare characteristics of African American and White women by obesity status. We used cross tabulations to compare all categorical variables by race and obesity status. We used chi-square statistics as the corresponding measure of heterogeneity. For continuous variables, we determined summary measures (mean and standard deviation) for each subgroup. We used analysis of variance (ANOVA) to compare mean values across subgroups.
We examined the multivariate associations between obesity status and the covariates of interest using logistic regression. During this process, we considered those in which we had substantive a priori interest based on prior literature. In the final model, we included the covariates that were at least of borderline statistical significance during forward stepwise selection (
A summary of baseline sociodemographic, health, and neighborhood characteristics of the cohort is presented in Table
Comparison of participant characteristics by race.
African American females |
White females | ||||
Mean | SD | Mean | SD | ||
Demographic characteristics | |||||
Age, mean (SD) | 30.12 | 1.5 | 29.93 | 1.44 | 0.134 |
Years of education | 12.75 | 2.07 | 10.92 | 2 | <0.001 |
Percentage with college degree or above | 14.61 | 35.34 | 2.89 | 16.8 | <0.001 |
Number of assets | 2.49 | 2.16 | 3.2 | 2.15 | <0.001 |
Average family income ($) | $31,825 | $22,384 | $35,164 | $24,818 | 0.188 |
Average personal income ($) | $16,077 | $12,194 | $12,235 | $11,993 | 0.001 |
Number of children | 1.49 | 1.33 | 1.59 | 1.21 | 0.388 |
Percentage married | 27.18 | 44.52 | 59.41 | 49.25 | <0.001 |
Age at first child’s birth | 23.61 | 5.22 | 23.32 | 5.07 | 0.504 |
Percent below poverty level | 32.94 | 47.04 | 27.74 | 44.91 | 0.216 |
Health characteristics | |||||
BMI | 26.56 | 6.13 | 26.13 | 6.29 | 0.422 |
Percent overweight | 36.05 | 48.06 | 30.95 | 46.41 | 0.278 |
Percent obese | 24.58 | 43.08 | 25.00 | 43.43 | 0.908 |
Percent current smoker | 42.62 | 49.48 | 53.18 | 50.04 | 0.011 |
Percent with very good/excellent self-reported health | 56.03 | 49.67 | 54.12 | 49.98 | 0.650 |
Adult neighborhood characteristics | |||||
Median household income ($) | $25,323 | $10,616 | $33,352 | $12,641 | <0.001 |
Percent African American | 72.17 | 31.29 | 8.06 | 11.73 | <0.001 |
Percent poor neighborhood | 35.42 | 15.9 | 24.77 | 12.48 | <0.001 |
Childhood neighborhood and personal characteristics | |||||
Percent African American | 72.8 | 33.44 | 8.95 | 19 | <0.001 |
Mother married at time of birth (%) | 66.92 | 47.08 | 87.86 | 32.75 | <0.001 |
Mother married to father at age 8 (%) | 60.17 | 48.98 | 70.86 | 45.57 | <0.001 |
Percent not on welfare as child/adult | 41.3 | 49.26 | 29.71 | 45.83 | 0.004 |
Percent poor neighborhood | 22.36 | 15.49 | 13.33 | 11.2 | <0.001 |
Household below poverty level (%) | 31.81 | 46.6 | 32.57 | 46.99 | 0.845 |
Overweight as child (%) | 5.15 | 22.12 | 4.65 | 21.12 | 0.787 |
At risk for overweight as child (%) | 7.79 | 26.82 | 8.72 | 28.3 | 0.683 |
Although household income was not significantly different between African American and White women ($31,825 versus $35,164,
African American and White women also substantially differed by current neighborhood characteristics (Table
Both African American and White women experienced similarly high rates of childhood poverty (32% and 33% resp.,
Respondents who were obese were similar to their nonobese counterparts in racial makeup, neighborhood characteristics, and family composition (Table
Comparison of participant characteristics by race and obesity status.
African American females | White females | |||||||||
Nonobese | Obese |
Nonobese | Obese | |||||||
Mean | SD | Mean | SD | Mean | SD | Mean | SD | |||
Demographic characteristics | ||||||||||
Age, mean (SD) | 30.12 | 1.52 | 30.14 | 1.45 | 0.827 | 29.98 | 1.45 | 29.75 | 1.38 | 0.367 |
Years of education | 12.84 | 2.09 | 12.56 | 2 | 0.114 | 11.07 | 2.12 | 10.52 | 1.61 | 0.127 |
Percentage with college degree or above | 15.94 | 36.64 | 10.64 | 30.92 | 0.074 | 3.97 | 19.6 | 0 | 0 | 0.192 |
Number of assets | 2.63 | 2.17 | 2.16 | 2.05 | 0.011 | 3.48 | 2.2 | 2.38 | 1.78 | 0.004 |
Average family income ($) | $33,085 | $23,203 | $28,672 | $19,923 | 0.069 | $37,537 | $27,481 | $28,220 | $14,764 | 0.104 |
Average personal income ($) | $16,535 | $12,442 | $14,734 | $11,405 | 0.118 | $13,408 | $12,717 | $8,838 | $8,902 | 0.045 |
Number of children | 1.49 | 1.37 | 1.51 | 1.25 | 0.859 | 1.58 | 1.23 | 1.62 | 1.21 | 0.856 |
Percentage married | 26 | 43.9 | 31.38 | 46.53 | 0.150 | 60.32 | 49.12 | 54.76 | 50.38 | 0.529 |
Age at first child’s birth | 23.76 | 5.26 | 23.15 | 5.06 | 0.166 | 23.36 | 5.16 | 23.25 | 4.96 | 0.908 |
Percent below poverty level | 31.94 | 46.68 | 35.48 | 48 | 0.422 | 23.89 | 42.83 | 40 | 49.61 | 0.052 |
Health characteristics | ||||||||||
BMI | 23.75 | 3.12 | 35.18 | 4.94 | <0.001 | 23.2 | 3.22 | 34.94 | 4.85 | 0.000 |
Percent current smoker | 42.81 | 49.52 | 42.02 | 49.49 | 0.850 | 51.59 | 50.17 | 59.52 | 49.68 | 0.375 |
Percent with very good/excellent self-reported health | 61.18 | 48.78 | 42.55 | 49.57 | <0.001 | 58.73 | 49.43 | 42.86 | 50.09 | 0.074 |
Adult neighborhood characteristics | ||||||||||
Median household income ($) | $25,838 | $10,938 | $23,886 | 9566 | 0.036 | $34,481 | $13,947 | $30,303 | $8,291 | 0.095 |
Percent African American | 72.08 | 31.07 | 73 | 31.61 | 0.735 | 8.02 | 12.34 | 7.3 | 8.22 | 0.747 |
Percent poor Neighborhood | 35.16 | 15.97 | 36.13 | 15.44 | 0.485 | 23.4 | 11.23 | 28.86 | 14.73 | 0.026 |
Childhood neighborhood and personal characteristics | ||||||||||
Percent African American | 71.97 | 34.03 | 75.23 | 31.58 | 0.290 | 8.47 | 18.36 | 11.36 | 22.17 | 0.483 |
Mother married at time of birth (%) | 67.44 | 46.9 | 67.03 | 47.14 | 0.919 | 87.2 | 33.54 | 90.24 | 30.04 | 0.606 |
Mother married to father at age 8 (%) | 61.7 | 48.65 | 60.11 | 49.1 | 0.698 | 71.43 | 45.36 | 71.43 | 45.72 | 0.999 |
Percent not on welfare as child/adult | 40.9 | 49.2 | 46.27 | 49.99 | 0.196 | 27.77 | 44.96 | 38.09 | 49.15 | 0.210 |
Percent poor Neighborhood | 21.58 | 15.26 | 24.29 | 15.76 | 0.046 | 13.16 | 10.66 | 14.46 | 13.49 | 0.589 |
Household below poverty level (%) | 29.28 | 45.54 | 36.7 | 48.32 | 0.057 | 30.95 | 46.41 | 33.33 | 47.71 | 0.775 |
Overweight as child (%) | 1.4 | 11.77 | 17.13 | 37.78 | <0.001 | 1.63 | 12.7 | 11.9 | 32.78 | 0.004 |
At risk for overweight as child (%) | 5.44 | 22.7 | 15.47 | 36.26 | <0.001 | 6.5 | 24.76 | 16.67 | 37.72 | 0.048 |
Table
Simple and multiple logistic regression.
Simple Logistic Regression | MLR-Std model + childhood socioeconomic variables* | MLR-Std model + adult socioeconomic variables# | MLR-final modelX | |||||||||
(1) | (2) | (3) | (4) | |||||||||
Obese | OR | 95% C.I. | OR | 95% C.I. | OR | 95% C.I. | OR | 95% C.I. | ||||
Age | 50.974 | 0.040 | 64927 | 99.328 | 0.125 | 79006 | 26.367 | 0.122 | 5678 | 60.326 | 0.050 | 72421 |
Age*Age | 0.939 | 0.835 | 1.056 | 0.928 | 0.831 | 1.036 | 0.948 | 0.867 | 1.035 | 0.936 | 0.833 | 1.052 |
African American | 1.068 | 0.249 | 4.579 | 1.216 | 0.635 | 2.326 | 0.962 | 0.524 | 1.764 | 1.227 | 0.556 | 2.704 |
Married | 2.575 | 0.721 | 9.191 | 1.973 | 1.246 | 3.124 | 2.156 | 1.413 | 3.288 | 2.175 | 1.304 | 3.627 |
Underweight at age 7 | 0.288 | 0.065 | 1.283 | 0.276 | 0.064 | 1.191 | 0.165 | 0.040 | 0.683 | 0.288 | 0.064 | 1.294 |
At risk for overweight at age 7 | 2.832 | 1.524 | 5.262 | 2.515 | 1.363 | 4.639 | 4.055 | 2.420 | 6.794 | 2.841 | 1.527 | 5.286 |
Overweight at age 7 | 12.039 | 4.824 | 30.043 | 13.343 | 5.496 | 32.396 | 12.326 | 5.888 | 25.803 | 12.100 | 4.873 | 30.046 |
Number of children | 0.678 | 0.427 | 1.078 | 0.833 | 0.666 | 1.043 | 0.895 | 0.747 | 1.073 | 0.796 | 0.629 | 1.006 |
Age at first child’s birth | 0.969 | 0.911 | 1.031 | 0.969 | 0.915 | 1.026 | 0.968 | 0.923 | 1.015 | 0.969 | 0.911 | 1.031 |
Years of education | 0.967 | 0.859 | 1.090 | 0.964 | 0.869 | 1.069 | 0.935 | 0.849 | 1.029 | 0.966 | 0.858 | 1.087 |
Home ownership in adulthood | 0.417 | 0.206 | 0.844 | 0.477 | 0.254 | 0.896 | 0.515 | 0.311 | 0.854 | 0.430 | 0.214 | 0.866 |
>90% African American Neighborhood in Childhood | 0.926 | 0.575 | 1.493 | 1.032 | 0.642 | 1.660 | — | — | — | 0.919 | 0.570 | 1.482 |
>90% African American Neighborhood in adulthood | 0.993 | 0.630 | 1.564 | — | — | — | 1.060 | 0.738 | 1.522 | 0.998 | 0.634 | 1.571 |
Poor neighborhood in childhood (>20% poverty) | 0.971 | 0.625 | 1.510 | 0.956 | 0.633 | 1.443 | — | — | — | 0.984 | 0.637 | 1.522 |
Poor neighborhood in adulthood (>20% poverty) | 1.089 | 0.590 | 2.011 | — | — | — | 1.217 | 0.726 | 2.039 | 1.089 | 0.593 | 2.002 |
Home ownership in childhood (by parents) | 1.212 | 0.609 | 2.411 | 0.967 | 0.482 | 1.942 | — | — | — | 1.187 | 0.597 | 2.361 |
Married X African American | 0.825 | 0.210 | 3.242 | — | — | — | — | — | — | — | — | — |
Married X number of children | 1.201 | 0.758 | 1.902 | — | — | — | — | — | — | — | — | – |
574 | 640 | 800 | 574 |
*Controlled for standard model (age, race, marital status, childhood weight status, age at first child’s birth, education, and homeownership in adulthood)+ childhood SES variables (as a child: predominantly African American (>90%) neighborhood, poor neighborhood (>20% poverty), and homeownership by parents); #controlled for standard model (age, race, marital status, childhood weight status, age at first child’s birth, education, homeownership in childhood) + Adult SES (as an adult: African American (>90%) neighborhood, poor neighborhood (>20% poverty)); XControlled for Standard Model (age, race, marital status, childhood weight status, age at first child’s birth, education, and homeownership in childhood); childhood SES variables (as a child: predominantly African American (>90%) neighborhood, poor neighborhood (>20% poverty), and homeownership by parents); adult SES (as an adult: African American (>90%) neighborhood, poor neighborhood (>20% poverty)).
Table
and Whites did not differ in the characteristics associated with weight status. Multiple tests for the interaction terms for race were not significant (analysis not shown).
In this longitudinal cohort of African American and White women from Baltimore, MD, we sought to quantify the association between adult obesity and individual and neighborhood socioeconomic status in both childhood and adulthood, and to determine whether these relationships varied by race.
We found that rates of obesity were high and similar between the African American and White women. Several things could potentially explain this finding. One possibility is that the number of White respondents was too small to detect a difference if it exists. However, an additional reason is that there is little racial difference in characteristics associated with adulthood obesity due to socioeconomic homogeneity in our cohort. Studies that show differences in obesity rates between African Americans and White women may be unable to fully control for differences in socioeconomic status.
We also found a strong and consistent relationship between childhood overweight at age 7 and adult obesity. Our work supports the importance of identifying and managing overweight in childhood to prevent life-long morbidity. Children who are overweight are at an increased risk for obesity and its associated conditions such as hypertension, hyperlipidemia, and diabetes [
Married women were more likely to be obese than their unmarried counterparts even after controlling for assets and parity. The effects of marriage on obesity did not significantly vary by race. The literature on the effect in marital status and BMI has been mixed. Some studies show married women have increased BMI, while others show mixed or no relationship [
One strength of the study is the longitudinal, multilevel nature of the data. Many studies that attempt to measure childhood effects on adulthood health outcomes rely on retrospective data which is subject to recall bias. Our measures were obtained prospectively. Another strength of this study is its use of census data for aggregate neighborhood measures in childhood and adulthood. Studies that aggregate respondents’ characteristics to determine community SES may be subject to atomistic fallacy because the study population may not be a representative sample of the population [
The pathways to adulthood data had a limited geographic focus, children born in the Johns Hopkins Hospital catchment area. This group had a higher proportion of African Americans, higher poverty rates, and higher obesity rates than the US as a whole for that time period [
Our study has important limitations. The primary outcome variable, BMI, was calculated through self-reports of height and weight. Several studies have found underreporting of weight, particularly among women and all those of higher weights [
In this sample of women with high rates of childhood and adulthood poverty, obesity rates were high. Although living in a poor neighborhood was not an independent risk factor for obesity, poor neighborhoods in our sample had higher rates of adult obesity. Childhood at risk for overweight and overweight was strongly associated with adult obesity. Being married was also associated with obesity. Efforts to combat obesity should be focused not only on individual patients, but also within at-risk and affected families and communities.
None of the authors of this paper has any conflict of interests to disclose related to employment, consultancies, honoraria, stock, expert testimony, patents, royalties, or any other relationships related to this project.
Dr. M. R. Saunders gratefully acknowledges funding support from the NIH Health Disparities Loan Repayment Program. Dr. M. R. Saunders had full access to all of the study data and takes responsibility for the integrity of the data and accuracy of the data analysis. The data was obtained through the Inter-university Consortium for Political and Social Research (ICPSR). No sponsor had any role in the design and conduct of the study: collection, management, analysis, and interpretation of the data: or preparation, review, and approval of the paper. An abstract of this paper was presented at the American Public Health Association Annual Conference in Denver, CO, November 2010.