NEIGHBORHOOD PERCEPTION AND OBESITY IN AGED MEXICAN AMERICANS

Abstract

Background

Hypotheses on the relationship between neighborhood perception and obesity (as measured by body mass index) seem to generally posit that a positive neighborhood perception may be related with behaviors that positively moderate body weight.

Objective

To determine if and how there is an association between positive neighborhood perception and obesity—while accounting for frailty- and disability-related factors.

Design

Cross-sectional study from Wave-5 of the Hispanic Established Population for the Epidemiological Study of the Elderly (HEPESE).

Setting

Data files housed by the Sociomedical Division in the department of Community Health and Preventive Medicine at the University of Texas Medical Branch in Galveston, Texas.

Participants

A total of 889, aged 75–90 community-dwelling Mexican Americans in the Southwest United States.

Measurements

Body mass index (BMI=Kg/m²), neighborhood perception, grip strength, gait speed, depression symptomatology, chronic conditions, presence of limitations with basic and instrumental basic activities of daily living (ADLs), and other health and demographic variables are used in logistic regressions predicting the likelihood of being obese (BMI > 30 Kg/m²) versus being of normal weight (BMI 18.5–25.4 Kg/m²).

Results

The odds of being obese increase: as the level of positive neighborhood perception increases; grip strength increases; and with having any limitations with basic-ADLs.

Conclusions

These findings provide evidence that a positive neighborhood perception need not always be accompanied with a reduced risk of being obese. Because functional limitations in older ages may influence how positive neighborhood perception affects BMI, more research is needed.

Introduction

Investigations on the relationship between neighborhood perception (NP) and body mass index (BMI=Kg/m²) continue to grow as interest on the topic increases and data become available. In a study by Fish and colleagues (1), the association between perceived neighborhood safety and BMI was investigated using over two-thousand adults with a mean age of 40 who resided in Los Angeles, California between the years of 2000 and 2001. About one-third of their sample was “White,” about half of it (55%) was made up of Latinos/as, and 54% self-identified as being first-generation immigrants. Our study differs in that we only have Mexican Americans (MA) aged 75 and above in our sample.

Fish and colleagues (1) begin their well-written report by pointing out that unhealthy weight levels are a major public health concern. Although a recent publication, using National Health and Nutrition Examination Survey (NHANES) data, on the prevalence of obesity in United States (US) adults age 20 and over shows a leveling off of the increasing obesity trend, they report that in 2009–2010 the obesity prevalence for all US adults was at 36% (2). Others continue to argue that obesity in the US is an epidemic (3). Tables in the Flegal et al (2) report show that 37% of males and 45% of females amongst MAs age > 20 have BMIs > 30 Kg/m²—with 4% of males and 7% of females having BMIs > 40 Kg/m². It is important to note that only 1.1% of their weighted sample was made up of MAs age 60 and above—equaling an unweighted total of only 292 cases. Our study fills this gap by making use of an analytic sample with almost nine-hundred MAs.

Fish and colleagues (1) point out that investigating obesity is important because it has been linked to many health outcomes. For example, investigations have found that excess weight increases the risk of experiencing many chronic diseases, including cardiovascular diseases (4) and diabetes (5). Obesity has also been found to influence life expectancy (6) and to directly affect the psycho-emotional health of the individual (7). In general, it could be said that being obese increases the risk for developing undesirable health conditions (e.g., diabetes). Our investigation contributes to this line of research by investigating the relationship between several health and demographic factors with obesity.

Fish and colleagues (1) correctly highlight the fact that obesity is directly connected with the US’ financial well being and policy issues because it impacts health care costs. For example, some have calculated that about 9% of the total annual US medical expenditures in 1998 ($92.6 billion in 2002 dollars) were due to overweight- and obesity-attributable medical spending (8). Others have used microsimulations to estimate lifetime costs for seventy-year-olds and found that the obese will spend $39,000 more on health care than those of normal weight and that Medicare will spend approximately 34% more on an obese person than on a normal-weight person (9). More recently, research found that elderly men who are overweight or obese at age 65 had 6–13% more lifetime health care expenditure than their normal-weight male-counterparts and females who were overweight or obese at age 65 had 11–17% more lifetime health care expenditure than their normal-weight female-counterparts (10). As is clear, obesity has important health and financial implications.

Despite the many efforts to highlight the obesity epidemic in the US, our understandings of what factors are related to being overweight remain inconclusive. While there is much research on the effects of NP on outdoor physical activity (e.g., walking), research on the relationship between NP and BMI remains limited. Our project fills this gap in the literature. Although investigations on how NP is related to body weight abound, no theoretical premises on the mechanisms between context and obesity have been established. This is why Fish and colleagues (1) explain that the mechanisms through which neighborhood attributes lead to weight gain are not well characterized. We hope our investigation and closing “theory discussion” contribute towards a better characterization of how NP is related to BMI.

Researchers have convincingly argued that increasing our understanding of how NP is related to BMIs may help advance our ability to delineate the mechanisms through which macro-level phenomenon affect micro-level BMI. Because ethnicity is important when understanding micro-level health outcomes (11), insight on if/how MAs have been disproportionally affected by the social structures involved in the US obesity epidemic may reveal what factors significantly impact how they regulate their body weight. For example, factors such as low-access to healthy food or facilities for physical activity, residing in lower socioeconomic areas, and economic disparities associated with poor food choices may be important determinants of obesity (3). While previous research has advanced our knowledge of how micro-level outcomes may be influenced by macro-level variables, it is unfortunate that standard methodological approaches often fail to address the role which previous contexts may play in shaping these relationships. Our project offers in closing a detailed discussion on how inter-context variability may play a role in body weight.

Fish and colleagues (1) succinctly argue that “perceived neighborhood safety” is an important mechanism through which the contextual characteristics may influence obesity. They explain that perceiving the neighborhood of residence as unsafe may increase the risk of obesity through a number of mechanisms. For example, reduced outdoor physical activity as a result of not feeling save outside the home (12, 13), and the potential for increases in stress-eating behaviors (14)—where eating may reduce distress presumably via central opiods (15), may all increase the odds of being obese. The implicit argument is that NP, a self-reported subjective (16) measure of context, has the ability to capture important characteristics in the person's environment that are related to their body weight.

Before moving on, it is important to note that we are only interested in the individual-level subjective measure of neighborhood. In the paper by Fish and colleagues (1), they include Census tract level measures and refer to them as “neighborhood” measures. We use the term neighborhood in a different way and under different circumstances. The term neighborhood has as of yet to be scientifically defined. We use the term because it was used during the administration of the questions from which we derived our measures of NP. We do not advance that calculations explicitly aggregated by Census geographical polygons be referred to as “neighborhood measures.”

Fish and colleagues (1) explain that the limited research on NP of safety and body weight is inconclusive. For example, one study found that perceiving the neighborhood as unsafe was related with greater odds of being obese (17), while another found no significant statistical association (18). Fish and colleagues find that perceiving the neighborhood as unsafe is related with greater odds of being obese (1). A related article explains that perceived neighborhood safety and walking outdoors are related (19). Others have noted that the perception of the built environment is also an important factor related to walking and thus body weight (20).

Our study approaches the same topic with a different NP measure. In contrast to Fish and colleagues (1) where perceived safety is the NP-factor of interest, our investigation measures an aged adult’s “positive” NP. Our study also differs in that it focuses on contrasting those of “normal” weight to those consider to be “obese” as measured by BMI. Positivity in a person’s NP is measured by asking them about their: (1) satisfaction with their neighborhood, (2) their view on neighbors willingness to help, (3) if they feel their area can be described as being a close-knit community, and (4) if they think people in their neighborhood can be trusted. Thus, our study captures a positive and subjective measure of neighborhood perception in contrast to the “negative” and subjective measure used by Fish and colleagues (1). In a sense, we are covering the same topic from the other side of the coin.

Research with positive NP scales (measuring respondents evaluations of the “aesthetic” quality of their environment) abound where the outcome of interest is physical activity—not BMI. For example, a study investigating neighborhood aesthetics and convenience levels by asking questions such as “How would you rate the general friendliness of the people?” found that those who report positive changes in NP were more likely to increase their walking (21). Fortunately, the authors point out that NP with acceptable psychometric properties is lacking (21).

Similar studies have found an association between positive- NP and physical activity in a sample from Japan (22). Another semi-related study on neighborhood attachment where NP was assessed with statements such as my place “is the ideal neighborhood to live in,” found that NP was related to higher levels of fruit and vegetable intake (23). A study with a sample from Nigeria found that a poor evaluation in the aesthetic quality of the neighborhood was related to higher levels of overweight in adults who reside in low SES neighborhoods (24). Others have even used multilevel models with a sample of Canadian adults to show that living in areas with high “material deprivation” is associated with higher BMIs (25). We were unable to locate a single article where a positive NP scale was used to model body weight.

In terms of theory, the argument is that neighborhood and body weight in older ages are linked (26). Some have explained that perceptions of neighborhood are believed to be intertwined with the concept of social cohesion (27)—a concept found to be related to outdoor physical activity (28). Investigators have asked that researchers contribute towards the development of a comprehensive list of measurable indicators for studying the relationship between context and body weight, where NP is given as a candidate for understanding the mechanisms that influence body weight (29). Our exploratory study pays heed to their call and investigates if a positive NP scale offers value towards the development of more scientific and measurable indicators. We compliment this main effort by offering a relevant and important discussion on the theory of how a person’s “history of contexts” can play a role in how their current habitat influences their current BMI.

Because demographic factors may influence NP—and such perceptions can be linked to body weight related behaviors (30)—we explore how positive NP is related to the likelihood of being obese in a sample of MAs. Effects on BMI can come from person-composition, group-characteristic or location-attribute factors. Our models only include person-level measures. Our positive NP measure varies by and is subjective to each individual. As such, it is treated as a person-level compositional characteristic. We complement and extend the competent efforts of Fish and colleagues (1) and others by exploring how a positive NP scale is related to the odds of being obese in a group of aged MAs.

Our general research question in this exploratory analysis is: Are positive NP and the odds of being obese related in a group of aged MAs? Our specific aim is to investigate the statistical relationship between NP and the odds of being obese versus of normal weight—while controlling for other demographic and health factors. We give a hypothesis to help orient the reader. While our hypothesis is not born out of existing research, in theory one would expect the odds of being obese to decrease as positive NP increases, thus: We hypothesized that as the level of positive NP increases, the odds of being obese will decrease. We posit such a hypothesis because we could argue that a positive perception of the habitat would be related with greater levels of outdoor participation—where high body weight may be reduced.

Our hypothesis is labeled as “tentative” due to uncertainty. Previous research has focused on non-aged adults as opposed to our very special analytic sample in which all have at least one chronic health condition and are overweight in general. Thus, we explore the obesity-NP relationship while accounting for our study subject’s level of “weakness” (as measured by grip strength) and “slowness” (as measured by their gait speed)— frailty related measures (31). These two are frailty-related components and have been linked to various health outcomes (32). From our measures, an individual’s physical performance (33) is only measured by their grip strength and gait speed. Thus, when we say an individual is “more frail” than another, we simply mean that on their physical performance, they have weaker grip strength or slower gait speed.

Because a person’s stage along the disablement process may also affect the relationship between NP and BMI (34) and because we distinguish frailty from disability (35), we include measures of disability. In general, the disablement process describes how a person’s health conditions affect his/her ability to function in daily living and how environmental factors influence these person-level mechanisms. We assess our subjects' “stage” in the disablement process by measuring their psycho-emotional status, quantity of chronic health conditions, and the presence of any limitations with activities of daily living.

Because all of the subjects in our analytic sample show evidence of some frailty and some degree of disability, we could expect that even though they may have a positive view of their neighborhood, their personal conditions would limit their ability to be physically engaged with outdoors activities. We first discuss our data source, measures, and statistical methods. Then, we review findings and conclude by outlining some limitations with our project, suggestions for future research, and a detailed theoretical discussion on how context(s) can affect the impact of future contexts on individuals.

Subjects and Methods

Subjects

Observations of aged MAs were obtained from the Hispanic Established Population for the Epidemiological Study of the Elderly (HEPESE), an ongoing study of a sample living in the Southwestern (Arizona, California, Colorado, New Mexico, and Texas) United States (36). Over the last two decades, HEPESE data has been used by various investigators to study a wide range of health outcomes in aged Mexican Americans (37). We make use of Wave-5 data which was collected during 2004 – 2005.

Dependent Variable

We initially ran regressions comparing: (1) underweight (BMI 14.0 – 18.4 Kg/m2) versus normal (BMI 18.5 – 25.4 Kg/m²), (2) overweight (BMI 18.5 – 25.4 Kg/m²) versus normal, and (3) obese (BMI 25.5 – 30.4 Kg/m2) versus normal. Because our NP scale was only significant in the obese versus normal comparison, our report only includes these findings. Among the full sample (n=1,547) at Wave 5 between age 75 and 90 with BMIs between 14 and 45, there were 19 who were underweight, 481 who were normal weight, 612 who were overweight, and 435 who were obese. Because of potential data issues with few cases at the extreme lower end of BMIs and extreme high end of age, we omit them from our investigation. Our analytic sample is made up of 889 aged Mexican Americans—all tables below reflect only statistics on this sample.

Positive Neighborhood Perception Scale

Although a series of other questions are available from the dataset (see Appendix 1), only four Wave-5 HEPESE items were used in the creation of our neighborhood perception (NP) scale. The questions used in the HEPESE Wave-5 neighborhood survey questions were inspired by those used in other surveys (38–40). In Wave-5, HEPESE respondents were asked the following questions regarding their perception of the neighborhood they lived in: All things considered, would you say you are very satisfied, satisfied, neither satisfied nor dissatisfied, dissatisfied, or very dissatisfied with your neighborhood as a place to live? They were allowed to answer using 5 “satisfied” categories in a Likert scale (see Appendix 1). Those who respond “very satisfied” get a “1” on the first items and those from “satisfied” to “very dissatisfied” get a zero.

Respondents were then instructed that surveyors were “going to read [the respondent] some statements, which may or may not be true about [their] neighborhood.” Respondents were then given the following sub-question sections: this is a close-knit neighborhood; people around here are willing to help their neighbors; people in this neighborhood can be trusted. They were allowed to answers using 5 “agree” categories in a Likert scale (see Appendix 1). Those who responded “strongly agree” get a “1” on these three items and those from “agree” to “strongly disagree” get a zero. All non-valid answers (i.g., don’t know, refused, or missing) got no score (i.e., are recorded as missing). Because values of “1” capture very satisfied and strongly agree, our NP scale consequently represents extremely-positive responses from the addition of all these four items. We abstain from labeling the scale with the “extreme” term so as to improve the flow of the discussion. However, readers should note that those responding with satisfied are grouped with unsatisfied individuals and those who agree are grouped with those who disagree. This was primarily done because more subjects responded at the extreme ends of Likert scales—something that may be related to the fact that research has shown a potential problem with Spanish translated Likert-type response scales (41, 42).

High scores on our NP scale indicate a high level of neighborhood satisfaction, where neighbors are perceived as being willing to help, where the person feels their area of residence is a close-knit neighborhood, and where people can be trusted. From the coding scheme above, we get the following distribution on the NP scale for the full (n=1,547) sample: 0=604, 1=546, 2=136, 3=206, 4=28. The NP distribution from our analytic sample (n=889) is as follows: 0=349, 1=328, 2=88, 3=121, 4=11. Limitations arising from the ambiguity of the items in the HEPESE questionnaire are addressed in closing.

Frailty Related Covariates

We calculate a respondent’s grip-strength by using a dynamometer. During data collections, respondents were asked to use their strongest hand for their grip strength test. They placed their strong hand on a table with the palm facing up, grabbed the handles of the dynamometer using an underhand grip and were instructed to squeeze as hard as possible. Two grip strength measures, recorded to the nearest half kilogram, are used to create the average grip strength performance measure. Respondents who had recently had surgery on their arm or hand did not participate in the exercise and were not included.

Gait speed was measured by timing the number of seconds required to walk eight feet of uniform walking surface. Study participants were instructed to walk the “eight foot course” at their usual speed, “just as if [they] were walking down the street to go to the store.” Participants were instructed to walk past the end of the course and not slow down near the end. Respondents were allowed to use assistive devices to complete the exercise while those who could not walk even with assistive devices did not participate in this exercise.

Disability Related Covariates

We do include other health related covariates other than activities of daily living (ADLs). The Center for Epidemiologic Studies-Depression (CES-D) scale is frequently used to determine if depressive symptoms are present (43). We use four CES-D questions from this scale to create a “positive affect” score. To facilitate interpretation, the responses are reverse coded and added such that higher positive affect scores reflect higher positive affect—a lower presence of depression symptomatology.

To create a scale of chronic conditions we constructed a count of the following conditions: (1) pain or discomfort while walking or standing; (2) the presence of doctor diagnosed diabetes, sugar in urine or high blood sugar; (3) ever having cancer diagnosed by a doctor; (4) ever having or suspected having a stroke, blood clot in the brain or brain hemorrhage; (5) ever having a heart attack; (6) ever having a broken or fractured hip; and (7) ever having broken or fractured any other bone. Thus, the chronic health condition scale ranges from 0 – 7. Please note that during HEPESE Wave-5 data collection, respondents from the original cohort (present in Wave-1) reported on their hip (or other) fractures since “the last time we talked”, while new recruits during Wave-5 were asked to report if they have “ever” experienced a hip (or other) fractures. Limitations arising from these and the fact that all our aged MAs have at least one chronic health condition are addressed in closing.

Because we believe the effects of NP can be moderated by the functional ability, we account for the presence of limitations with both basic and instrumental ADLs (BADL and IADL). The ability to perform ADLs is the result of complex interactions between physical, social, environmental, and cognitive factors (44) and has in general been found to diminish in older ages (45). ADLs “constituted the most consistently measured aspects of functional status in” older people [46] and have been used over many decades (47). Our discussion includes interpretations of ADLs on aged MA’s likelihood of being obese versus of normal weight.

Health and Demographic Controls

We include other health and demographic controls. Respondents who reported being in either excellent or good health were split from those who reported fair or poor to form a binary self-reported health variable. To control for the effects of tobacco usage we operationalize this construct as whether or not the respondent reported being a current smoker. We also include demographic covariates. Respondents' self-reported income was categorized as less than $15,000 or $15,000 or more for the year 2005. Self-reported education was dichotomized as 8 years or less or greater than 8 years of schooling. We also include an exploratory variable of whether or not the respondent reported having no friends or family living in the neighborhood—to account for the fact that this may influence their positive NP score.

Statistical Modeling

As is customary, we provide descriptive statistics for analytic sample. We use three logistic models where the likelihood of being obese is contrasted to that of being of normal weight. In the first model (i.e., Model 1), NP and frailty-related factors are regressed on the likelihood of being obese versus of normal weight. After including disability related factors in Model 2, we add all other health and demographic variables in our final model (i.e., Model 3). Our discussion focuses on the outputs from Model 3. All data management and modeling is done in SAS 9.2 (Copyright, SAS Institute Inc. SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc., Cary, NC, USA).

Findings

We begin with descriptive statistics of the analytic sample— this excludes all those who are not of normal weight or obese. From Table 1, we see that 48% of those in our sample of aged MAs are obese and that the mean NP score is 1. On frailty related performance measure covariates, the average grip strength score is 20 and the average gait speed is 5 seconds. For our disability items, we see that: the average CES-D positive affect score is 2; the average chronic condition score is 5; about 25% of the sample has at least one BADL limitation; and about 70% have at least one IADL limitation. About 39% of the sample reports good or excellent health and only 7% currently smoke. Demographically, the analytic sample is: 63% female; has a mean age of 81; about 40% are married; 56% are born in the US; 79% have less than a high school education; and 76% report a household income lower than $15,000 per-year. More than have (52%) report having no family living in their neighborhood, while 22% report having no friends living in neighborhood.

Table 1.

Mean, standard deviation, minimum and maximum values for variables in analytic sample (n=889)

Variable	Mean, or %	SD	Min	Max
Dependent Variable
Obese	48%
Independent Variable
Neighborhood Perception	1.01	1.06	0	4
Frailty Covariates
Grip Strength	19.73	8.39	0	50
Gait Speed in Seconds	5.37	3.60	0	40
Disability Covariates
CESD: Positive affect	2.12	2.35	0	12
Chronic conditions	4.75	1.17	0	7
Any basic ADLs	25%
Any instrumental ADLs	69%
Health Controls
Self-rated health good/excellent	39%
Current smoker	7%
Demographics
Female	63%
Age	81	3.83	75	90
Married	40%
US-Born	56%
Less than some High School	79%
Household income < $ 15,000	76%
No family living in neighborhood	52%
No friends living in neighborhood	22%

Table 2 makes use of six non-standard BMI cut-points to display in greater detail how NP scores are distributed by BMI groupings. This table compliments our discussion by providing more details on the distribution between dependent and independent variables. In our logistic regressions below, we are predicting the odds of being in the obese categories. Thus, 424 obese study subjects are being compared to 465 subjects of normal weight (as per their BMI). Please note that two BMI groups under the obese category have a mean NP score greater than “1”—indicating that some members within the BMI-group reported very positive views of their neighborhood.

Table 2.

Neighborhood Perception (NP) score according to body mass index (BMI)

BMI	Count	Mean NP	SD	Min	Max
Normal Weight
18.5–21.4 Kg/m2	100	0.87	0.97	0	4
21.5–25.4 Kg/m2	365	0.95	1.05	0	4
Overweight
25.5–29.4 Kg/m2	N/A	N/A	N/A	N/A	N/A
Obese
29.5–33.4 Kg/m2	257	1.16	1.10	0	4
33.5–37.4 Kg/m2	110	0.95	1.04	0	4
37.5–41.4 Kg/m2	42	1.10	1.25	0	4
41.5–45.0 Kg/m2	15	0.93	0.96	0	3

Overweight study subjects are not included in the analysis but are displayed in the table to mark the BMI gap between those of normal weight and the obese (as per their BMI)

We now turn our attention to the logistic regressions. From Table 2, we see that throughout all three models, NP remains statistically significant. As the level of positivity in NP increases, the odds of being obese increase as well. For example, the odds of being obese increase by 16.1% with every increase on the positive NP scale (net of all frailty-, disability-related, health, and demographic factors in the model). This finding falsifies the hypothesis under investigation. Although NP is related to the odds of being obese, the relationship is opposite of what we had predicted.

Please note that while gait speed was insignificant in our frailty related performance measures, with each kilogram increase in grip strength, there are 3.2% greater odds of being obese than of normal weight. This finding is likely due to the higher amount of muscle mass in persons with high BMIs. Please note that since BMI is a measure of body mass, having a high BMI does not necessarily imply high adiposity. With disability related items, with every increase in the number of chronic conditions, there is a 47% increase in the odds of being obese versus of normal weight—ceteris paribus. The presence of functional limitation with basic activities of daily living is also related to having greater odds of being obese versus or normal weight. On the demographic variables, greater odds of being obese are found for females and for those with less than some High School education, while age is inversely related with the odds of being obese.

Conclusion

NP may be associated with obesity in intricate ways because of the mechanisms that affect their assumed bidirectional relationship. In answer to our research question, we find that NP is significantly related to the odds of being obese for aged MAs. However, our hypothesis that our positive NP scale would be indirectly related with the odds of being obese finds no support. It may be the case that having a positive perception of the neighborhood does not reduce the odds of being obese if the person is: amongst the oldest old (i.e., over the age of 74), has a high level of frailty and disability related conditions, and has experienced a lifelong struggle at the lower rungs of the socioeconomic hierarchies in the US—where gender egalitarianism is low and moderate levels of education may be related to poorer diets and limited physical leisure-activity. Our hypothesis was posited with the assumption that a positive perception of the habitat would be related with greater levels of outdoor participation. In light of our functional limitation discussion, we advance that although aged MAs hold a positive perception of their neighborhood; their desire and ability for outdoor participation (or physical activity in general) are mitigated by their physical limitations in old age.

On a more theoretical note, while many difficulties exist in the analysis of neighborhood effects on body weight, one area that may further complicate interpretation of macro- on micro-relationships is how both macro- and micro-units change over time. If context only affects people as a function of time and depth of exposure, then we must first seek to account for the simpler of the two: time of exposure. We believe that, too frequently, place effect studies disregard the idea that the amount of time in the current social-environment dictates the degree to which said context influences an individual’s behavior. Such an approach treats micro-level units as being equally exposed, with regards to time and depth, to the macro-level measurement of interest. We now discuss two elements within time of exposure.

A person’s history of neighborhoods is not readily captured in most data and measurements of a neighborhood’s demographic and physical characteristics shifts over time are equally illusive. Although theoretical premises are usually implicit rather than explicit in place effect research on obesity, we would argue that in general, environment-obesity causal mechanisms assume the degree and duration of exposure to measured macro-factor matters. For example, current residence may be causally related with current BMI. However, current BMI may be influenced by: (1) within-person (e.g., degree of frailty), and (2) within-neighborhood shifts over time (e.g., accessibility to sidewalks).

Shifts in these elements may influence the causal mechanism between obesity and NP. For example, using a hypothetical case of the within-person fluctuation, we could say that subject-a (sa) changes marital status from time-1 (t1) to time-2 (t2) and becomes ambulatory-challenged between the same time period. Measuring the shift between sat1 and sat2 may be necessary to outline causal mechanisms between BMI and NP. Even more complex, assume sa moves to different neighborhoods (e.g., across different states) every 20 years. If sa is 80 years of age, then he has four different neighborhoods (n1-4) that may be causally related with his current BMI. Consequently, measurements of his current neighborhood perception would only capture n4t1 and ignore n1-3t1-3 (which may be causally related to his current BMI). Measuring the shift between san1 through san4 may be necessary to decipher the causal mechanisms between BMI and NP.

On the within-neighborhood fluctuation element, assume sa initially moved into his current residence because there was an 80% co-ethnic concentration (c1). Years later, his neighborhood’s co-ethnic concentration drops to 20% (c2) where now his area is largely occupied by non-co-ethnics. Since not all people have the same ability to re-locate or to select their ideal place of residence, within-neighborhood fluctuations between sac1 and sac2 may influence the causal mechanisms between his BMI and NP. Consequently, accounting for within-subject and –neighborhood shifts may be critical if we are to understand the BMI-NP causal mechanism. Future research should seek to establish if these shifts matter.

If our proposition on the importance of within-person and -neighborhood shifts seem logical and important, then a series of challenging questions arise which should be addressed in future work. (1) How does an individual’s history of contexts influence how he/she interacts with their current social and physical environment? (2) Are individuals equally equipped to benefit from beneficial environments and/or resist harsh social habitats? (3) What biological, child-age factors, family attributes, and other elements influence an individual’s disposition as she/he interacts with their multiple environments?

Assessing depth of exposure may prove an extremely challenging task—for its scientific operationalization must first make sense of the various elements involved in its construction. At the forefront of the challenge stands the need to first delineate what a neighborhood is. However, if such geographical boundarization is possible, new challenges will arise. For example, how should we treat micro-level units that reside near neighborhood borders? If we decide that a group of 10 city-blocks making up a semi-even rectangular geographical-polygon constitutes a neighborhood, should we assume that individuals at the center of the “neighborhood” are related in the same way to their neighborhood’s measure as those in the edges of the polygon? More research on these important topics is needed.

There are technical limitations with our study, the NP questions in HEPESE may not provide a complete count of how an individual feels about their residential environment and in particular how these views affect BMI related behaviors like exercise and food consumption. Another limitation is that we do not account for an objective measure of context (e.g., availability of sidewalks and parks)—which may erase the NP-obesity statistically significant relationship we found. As noted earlier, study participants received different hip-fracture related questions. Our chronic condition scale ignores the fact that 20 cases reported having hip fractures two or more years ago and not in the last two years. This is only present with 2% of our analytic sample. Although we found no major issues with the fracture questions, longitudinal surveys should strive towards maintaining similar questions in all waves. A more complex limitation is the fact that we do not have study participants with zero chronic health conditions. This means we are unable to observe how NP is related to obesity in aged MAs who are not chronically ill.

Notwithstanding the limitations, we believe our investigation substantively contributes to the literature by showing that within aged MAs, NP is significantly related to obesity. Our exploratory project provides some validity to the argument that at the very least; discussions of place effects on obesity should be age-specific. Discourse on the effects of place on obesity should also explicitly account for how frailty and disability may play a role in the relationship between NP and BMI. Future research should aim at exploring the same topic with different populations, measurements, and modeling techniques.

Table 3.

Logistic model predicting the likelihood of being obese (BMI > 30 Kg/m²) versus normal (BMI 18.5–25.4 Kg/m²) weight

	Model 1				Model 2				Model 3
	Coeff1	OR2	PC3	p	Coeff	OR	PC	p	Coeff	OR	PC	p
Independent Variable
Neighborhood Perception	0.15	1.17	16.6%	0.02	−2.66	1.17	17.4%	0.02	0.15	1.16	16.1%	0.03
Frailty Covariates
Grip Strength	0.01	1.00	0.4%	0.60	0.16	1.02	1.6%	0.08	0.03	1.03	3.2%	0.01
Gait Speed in Seconds	0.01	1.01	0.7%	0.71	0.02	1.02	1.5%	0.45	0.03	1.03	2.6%	0.20
Disability Covariates
CESD: Positive affect					0.02	1.02	2.3%	0.46	0.03	1.03	3.0%	0.36
Chronic conditions					0.41	1.50	50.0%	<0.01	0.39	1.47	47.0%	<0.01
Any basic ADLs					0.62	1.86	85.8%	<0.01	0.74	2.11	110.6%	<0.01
Any instrumental ADLs					−0.18	0.84	−16.1%	0.29	−0.05	0.95	−5.2%	0.77
Other Health Covariates
Self-rated health									0.18	1.19	19.3%	0.28
Current smoker									−0.58	0.56	−43.8%	0.07
Demographics
Female									0.59	1.81	81.1%	<0.01
Age									−0.11	0.90	−10.1%	<0.01
Married									0.08	1.09	8.8%	0.61
US-Born									0.08	1.09	8.5%	0.59
Less than some High School									0.46	1.58	57.8%	0.02
Household income < $ 15,000									−0.25	0.78	−22.1%	0.17
No family living in neighborhood									−0.08	0.92	−8.0%	0.58
No friends living in neighborhood									0.30	1.35	34.6%	0.10
Intercept	−0.37			0.09	−2.66			<0.01	4.81			0.01
Observations	n = 897				n = 889				n = 889

¹Coeff=Coefficient;

²OR=Odds ratio;

³PC=Percent change in odds ratio

Appendix 1

HEPESE Neighborhood Perception Related Questions

All things considered, would you say you are very satisfied, satisfied, neither satisfied nor dissatisfied, dissatisfied, or very dissatisfied with your neighborhood as a place to live?

• (1)Very Satisfied
• (2)Satisfied
• (3)Neither Sat or Dissatisfied
• (4)Dissatisfied
• (5)Very Dissatisfied
• (8)Don’t know
• (9)Refused
• (.)Missing

About how many adults do you recognize or know by sight in this neighborhood? – Would you say you recognize no adults, a few, many or most?

• (1)No adults
• (2)A few adults
• (3)Many adults
• (4)Most or all adults
• (8)Don’t know
• (9)Refuse
• (.)Missing

Now I am going to read you some statements, which may or may not be true about your neighborhood. Please look at this card. For each statement tell me whether you strongly agree, agree, disagree, or strongly disagree. (If interviewee is unsure, mark neutral)

1. This is a close-knit neighborhood

2. People around here are willing to help their neighbors

3. People in this neighborhood generally don’t get along with each other

4. People in this neighborhood do not share the same values

5. People in this neighborhood can be trusted

   • (1)
     Strongly agree
   • (2)
     Agree
   • (3)
     Neutral
   • (4)
     Disagree
   • (5)
     Strongly disagree
   • (8)
     Don’t know
   • (9)
     Refused
   • (.)
     Missing