Understanding Obesity and Program Participation in the Context of Poverty and Food Insecurity

Abstract

Health officials, scientists and the public are increasingly concerned about the obesity epidemic and accompanying morbidity, mortality, health care costs, and loss of quality of life. This concern motivated Gibson’s article in this issue (1). Rapid shift in the distribution of relative weight to higher values has resulted in a substantially greater number of people being classified as obese. There is considerable interest in what policy and programmatic actions can and should be taken to reverse this shift. Besides constructing new policies and programs, we might also question whether existing policies and programs are inadvertently contributing to the problem. Nutrition programs designed in the past were primarily intended to help avert nutritional inadequacies. Concern about possible unintended effects of nutrition programs is not new (2), but apparently little evidence of this has existed until the recent linking of program participation with obesity. In 2001 Townsend and colleagues, using nationally representative cross-sectional data, reported that Food Stamp Program (FSP) participation and overweight are associated for women, but not for men (3). Gibson (1) studied FSP participation and obesity, attempting to overcome limitations in the Townsend study. Data used were from the 1985 to 1996 waves of the National Longitudinal Survey of Youth 1979; earlier data measured long-term family resources. Both concurrent and long-term FSP participation were associated with increased absolute risk of obesity of about 2 and 4.5 percentage points, respectively. The study used longitudinal data, but had no data on food insecurity. This commentary discusses Gibson’s use of data, conceptual frameworks, food insecurity and the role of program participation. Longitudinal data have advantages in reducing effects of unmeasured factors, having a clear temporal sequence, directly measuring change, and estimating dynamic relations of change with other factors (4). They can provide the best means, absent a randomized design, to establish that observed associations are not due to confounding and are causal (5). Gibson’s analysis partially utilized these advantages. Indicator variables representing individuals accounted for unmeasured time-invariant factors that may have otherwise confounded the relation between FSP participation and obesity, but cannot account for unmeasured time-varying factors. The temporal sequence was partially used when obesity at one year was modeled as a function of resources and FSP participation in the previous five years. The analytic approach, however, did not directly measure change in weight or obesity and did not estimate fully dynamic relations. Future research with these or other longitudinal data should directly examine how change is associated dynamically with other factors using multiple time points. For example, how does transition from not participating to participating in the FSP relate to subsequent weight gain or to transition from a nonobese to an obese state? Using multiple years of data and dynamic analytic approaches (4,6), these transitions can be examined in detail, using the temporal sequence to strengthen the plausibility that associations are causal. Several studies investigated whether poverty and food insecurity of families is linked with obesity (3,7–14). The only consistent finding is that white women in families that are resource-constrained–as measured by poverty, food insecurity, or both–are more likely to be obese. Food insecurity in children is not associated with overweight except in older (8–16 y), non-Hispanic white girls (11–12). One study found that, although mild or moderate food insecurity is associated with a higher risk of obesity, severe food insecurity is associated with a lower risk (9–10); three other studies showed similar but nonsignificant trends (3,13–14). Since resource-constrained families are more likely to participate in nutrition programs, any association of program participation and obesity must be understood in the context of poverty and food insecurity. Gibson’s conceptual framework depicted personal characteristics (including poverty and FSP participation) as causally related to obesity, either directly or by being mediated (i.e., indirectly) through food insecurity and health behaviors. This framework is generally consistent with others depicting the conceptual relations of poverty and other personal characteristics (but not program participation), food insecurity, and outcomes (11,15–16). However, Gibson’s framework and analysis lead to two questions. First, are placement of FSP participation among personal characteristics in the framework and assumed causal paths reasonable? Second, what are the possible consequences of not including a measure of food insecurity in analysis? The framework assumed that FSP participation is a personal characteristic that differs over time, and causal flow is from FSP participation to food insecurity and health behaviors to obesity. An alternate view is that FSP participation is a behavior that responds to personal circumstances, including food insecurity. These are two fundamentally different perspectives. Does FSP participation represent something that a person is or that a person does? Especially if FSP participation is a behavior, then the plau1 To whom correspondence should be addressed. E-mail: eaf1@cornell.edu.