| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Department of Health and Human Performance, University of Houston, Houston, Texas (R.E.L.)
Departments of Family Medicine and Preventive Medicine and Public Health, University of Kansas Medical Center, Kansas City, Kansas (K.A.G.)
Department of Preventive Medicine and Public Health, University of Kansas Medical Center, Kansas City, Kansas (S.H.)
Department of Psychology, Baker University, Baldwin City, Kansas (W.B.)
Department of Family Medicine, University of Kansas Medical Center, Kansas City, Kansas (K.S.K.)
Kansas Health Policy Authority, Topeka, Kansas (A.A.)
Office of Clinical Research, University of Minnesota Academic Health Center, Minneapolis, Minnesota (J.S.A.)
Address reprint requests to: Rebecca E. Lee, Ph.D., Department of Health and Human Performance, University of Houston, Giarrison Gymnasium 104 E, 3855 Holman Rd., Houston, TX 77004. E-mail: releephd{at}yahoo.com
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSMeasuresRESULTSACKNOWLEDGMENTSREFERENCES |
|---|
Methods: The joint contributions of individual and environmental factors on obesity status (obese vs. morbidly obese) and trying to lose weight (yes vs. no) were evaluated using generalized estimating equations. Patients at 29 clinics in rural areas (N = 414, M age 55.0 years (SD = 15.4), 66.3% female) completed anthropometric assessments of weight and height along with survey assessments of individual sociodemographics and trying to lose weight. Rural environments were assessed on aggregated physician access, and sociodemographic context.
Results: Most participants (70%, M BMI = 38.3) were obese and 30% morbidly obese. A majority (73%, n = 302) of the sample was trying to lose weight. Compared to obese, morbidly obese participants were more likely to be younger, disproportionately female, not have private insurance, have more comorbid conditions, and rate themselves in worse health in comparison to their obese peers. Compared to not trying to lose weight, trying to lose weight participants were more likely to be younger, disproportionately female, have fewer comorbid conditions, and have attempted to lose weight more times through exercise. Few relationships were seen between environmental variables and obesity or trying to lose weight.
Conclusions: There was no consistent pattern of relationships between environment factors and obesity or trying to lose weight was seen. Unique aspects of rural living may not be captured by traditionally available neighborhood measures.
Key words: rural health, obesity, environment, diet therapy, physical activity
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSMeasuresRESULTSACKNOWLEDGMENTSREFERENCES |
|---|
30) has increased from 23.0% to 30.5% over the last decade, while the prevalence of morbid obesity (BMI 40) has increased at an even faster rate, escalating from 2.9% to 4.7% over the same period [7]. Obesity is associated with numerous health compromising conditions, and the escalation to morbid obesity is associated with imminent disease, disability and even death [8–9]. Obesity and morbid obesity are chronic diseases that develop over the course of many years. These annually measurable increases in obesity and morbid obesity, despite widespread weight loss attempts are alarming, and suggest a common underlying factor, such as an obesogenic environment, rather than an individual biological or behavioral mechanism [10–12]. Epidemiological studies provide evidence for an obesogenic environment. These studies suggest that obesogenic environments promote obesity above and beyond the individual characteristics of the residents themselves. Obesogenic environments are characterized by lower socioeconomic status (SES; e.g., median household income) [13–17], fewer or lower quality material resources (e.g., goods and services) [18–24], higher crime rates [25] and unappealing surroundings [24]. Other studies have suggested that the concept of obesogenic environments is misguided; these neighborhoods are simply populated with individuals who have poorer dietary habits, lower rates of physical activity, and higher rates of obesity. Cause and effect can be difficult to disentangle, but there are a number of studies that suggest these population characteristics are, at least in part, a function of the limited supply, higher cost, and reduced access access to healthful food [21–23] in these areas, as well as the limited supply, higher cost, reduced access [18,19,24], and lower quality of physical activity resources [20].
The work investigating obesogenic environments has been conducted almost exclusively in metropolitan areas, despite the fact that obesity prevalence is higher in rural areas compared to metropolitan areas [25–29]. Rural areas may differ from metropolitan areas on a number of dimensions, but little is known about how rural areas might differentially promote obesity and the escalation to morbid obesity. Methods used to characterize metropolitan, obesogenic environments may not be appropriate for rural areas. Rural areas are defined by lower population density [30], higher rates of poverty [31], and fewer widely available goods and services. Goods and services, such as food outlets and health care facilities, may be clustered together in townships, rather than more widely geographically distributed as is seen in more traditional, suburban metropolitan areas. Cluster development has become increasingly common in rural areas as agricultural patterns shift and farming operations are consolidated [32]. In metropolitan areas, theorists have suggested that goods and services availability may stimulate lifestyles that contribute to weight gain [33,34]. For example, high-fat, low nutrient dense, "fast" food is widely advertised and accessible in metropolitan areas. In rural areas, it may be many miles from the home to the township cluster of goods and services; nevertheless, residents may be routinely exposed to the township cluster influences in the course of daily errands (to the school, store, pharmacy, health care provider, church). This type of daily exposure to an area may influence obesity and efforts to lose weight among rural residents in much the same way as metropolitan residents.
Evidence suggests that obesogenic environmental influences such as neighborhood SES and available goods and services are important in metropolitan studies of obesity and weight loss attempts. These relationships have not been tested in rural areas, and different land use strategies may render traditional neighborhood definitions less appropriate. Studies are needed to investigate these relationships and hypothesize alternative strategies for defining environmental influences that are appropriate for rural areas. This study investigated the relationship between obesity in rural communities with local environmental characteristics using a sample of obese rural primary care clinic patients. Patient data were geographically associated with environmental data from the US Census and physician accessibility data from state medical board records. These data were then used to investigate their joint influences on obesity status (obese vs. morbid obese) and trying to lose weight.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSMeasuresRESULTSACKNOWLEDGMENTSREFERENCES |
|---|
Primary care practices were selected for inclusion in this study based on availability of voluntary physician faculty for applicant medical students. No systematic methods were used to assure a random distribution of these practices. All physicians were familiar with patient recruitment and survey administration by students within their practices and approved all program research components prior to being matched with students. All procedures were approved by the Internal Review Board at the University of Kansas Medical Center.
Participants
Participants were 414 obese rural primary care patients aged 19–79 seen in twenty-nine clinics. The number of patients seen at each practice that entered the study ranged from one to thirty-six. There were seven practices that had fewer then 10 patients participating, thirteen practices that had between 10 and 19 patients, eight practices that had between 20 and 29 and one practice that had more then 30 patients participating.
|
Measures |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSMeasuresRESULTSACKNOWLEDGMENTSREFERENCES |
|---|
40).
Neighborhood Definition
Most studies investigating multilevel relationships between environmental factors and health have defined neighborhood as the area within a boundary such as a census tract, block group or zip code where the participant resides. Rural areas are geographically quite different from metropolitan areas where the bulk of these studies have been done, thus this strategy may not be the best for studies of rural areas. Therefore, in order to capture the township cluster, we selected the zip code of the physician practice to represent the "neighborhood" of the patient, based on the assumption that rural patients utilize physician practices that are located near where the patient resides. The physician office address is likely within the same township cluster near goods and services where these patients also engage in many economic and social transactions. This assumption is based on research evaluating small area variation in health care services utilization [36–38]. This body of health services research has consistently shown links between local population characteristics and health care use at the most geographically proximal facilities. Based on this strategy, twenty-six zip code areas were included as "neighborhoods" in the analyses. Several of the zip codes had more than one physician practice in them, so analyses were designed to account for this additional nesting effect.
Neighborhood-Level Data
United States Census data from the year 2000 [39] were used to measure neighborhood sociodemographic status. These variables were aggregated at the zip code level within which each physician practice was located. Data were used to assess educational attainment (proportion of residents with at least a high school education), median family income, proportion of residents living in poverty (below 175% of the poverty level), median housing value, and proportion living in crowded housing. Variables were chosen based on previous investigations that have associated obesity and obesity related behaviors with area socioeconomic status [13–17]. All variables were entered into analyses using empirical quartiles.
Physician density was defined as the number of physicians per 1000 residents in the zip code of the physician practice. Based on the commonly used rural township cluster design [32] and previous utilization research [36–38], this was used as an indicator of the density of available goods and services. Data were drawn from State of Kansas Medical Board Licensing data.
Procedure
Medical students who had successfully completed their first year of medical school applied to participate in a competitive, six-week summer elective. Thirty-two students were chosen based on their academic qualifications. Prior to field placement, students received 40 hours of training during the first week of the elective in human participant protection, study design, participant recruitment, survey administration, data management, and clinical and administrative skills. Data collection instruments were reviewed item by item and students conducted mock interviews with each other and under the supervision of two of the authors (a psychologist and a physician).
During the next six weeks, students spent the majority of their time in clinical activities. Using a systematic strategy, students asked the first patient with a BMI of 30 or greater during each morning and afternoon clinic session to participate in a survey following their appointment with the physician. An intake nurse completed and recorded anthropometric assessments of weight and height for each patient participant. Body mass index (BMI) was computed using the standard BMI formula (kilograms/meters2). Students sequentially approached potentially eligible patients until they enrolled one participant for each half-day session.
All patients were informed of the nature of the study and provided verbal consent to participate. All data were collected anonymously and in private exam rooms. Students read each survey question aloud and recorded responses directly onto survey forms.
Students sequentially approached additional patients only if initial patients meeting the criteria in a half-day session declined participation. Patients were excluded if the office visit was for pregnancy, early post-partum, a critical or acute illness, the patient appeared to be in immediate emotional distress, cognitive impairment or dementia, or if language difficulties precluded survey administration in English.
Students mailed in reports of participants recruited and all surveys administered every two weeks. A study coordinator maintained contact with all students via phone and e-mail throughout the course of the study to assist in problem solving and answer questions on patient eligibility, the study protocol, and practice issues or problems that arose.
Analysis
Data from completed interviews and surveys were returned by mail every two weeks and manually entered into an AccessTM database designed for this project. All surveys were double-entered by different individuals and cross-checked for data entry accuracy. We calculated descriptive statistics appropriate to the form of each variable. Statistical analyses were conducted in SASTM V9.1 [40].
We used generalized estimating equation (GEE) models in the analysis. Our two dependent variables were (1) whether the person was obese or morbidly obese (coded as 0 = obese and 1 = morbidly obese), and (2) whether the person was trying to lose weight (coded as 0 = no and 1 = yes) at the time of the survey. For each of the two dependent variables, three models were tested. First, individual-level, patient variables, including age, gender, and health status, and weight loss attempts were entered into the models and tested for significant contribution. Next, the physician density data were included in the analysis. Last, each census variable was tested in a separate model, due to the high levels of multicollinearity between these variables, to determine patterns in the relationship of area sociodemographic status to individual dependent measures. GEE modeling was chosen as it allows a logistic regression model to be used while also controlling for the appropriate levels of clustering of the neighborhood data (clustering of physician practice within zipcode). The PROC GENMOD procedure was used with the repeated statement and an unstructured correlation matrix was assumed.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSMeasuresRESULTSACKNOWLEDGMENTSREFERENCES |
|---|
|
|
|
Neighborhood Sociodemographic Factors
Odds ratios, confidence intervals and significance values from the models describing relationships of individual factors, physician density, and neighborhood sociodemographic factors to obesity and trying to lose weight are presented in Tables 4 and 5. Across the full models, individual factors and neighborhood-level factors remain consistent with the previously described models.
|
|
|
DISCUSSION |
|---|
In this obese sample, those with lower BMIs, who were "less" obese appeared to be in generally better health, this finding supports previous data that suggest even a modest amount of weight loss among obese individuals can have clinically important implications. Further, those individuals in the sample who reported that they were currently trying to lose weight also appeared to be in better health than those who were not trying to lose weight. Similar to other health behaviors (e.g., quitting smoking), those who were trying to lose weight had typically engaged in some type of weight loss strategy in the past, suggesting that repeated efforts may be important for long term success with complex health challenges such as obesity. Another interesting finding was that those who were trying to lose weight at the time of the study were more likely to have used exercise as a weight loss strategy in the past, but not diet modification. Regularly performed physical activity is consistently cited as a key factor in maintaining a weight loss [41–45]. Increasing attention to the benefits of physical activity, whether by health care providers or public media coverage, may be influencing those actively trying to lose weight or maintain health to engage in more activity-focused weight-loss strategies.
Sampling patients who attend a regular physician or clinic is convenient and efficient for the researcher; however, clinic patients may be different from other area residents. Future research is needed to determine whether other sampling strategies might produce strong relationships between neighborhood-level factors and individual-level behaviors. These relationships might also be influenced by the neighborhood selection strategy. Our analyses relied on zip code level data, which, in the context of our research question was appropriate. It may be that smaller area definitions might be more appropriate. However, the analyses were also run using aggregated neighborhood SES data (not presented) at the census tract level, and similar results were found, suggesting that this is probably not the case for these data.
The lack of expected associations between neighborhood factors and individual outcomes may reflect a true lack of association; however, it is plausible that data or methodologic limitations may also be at fault. Aggregated neighborhood level factors have demonstrated good reliability in previous studies [13, 15]; however, it is not always clear how the constructs represented by neighborhood level factors are operationally defined. Future research may require more detailed sampling of the ecologic characteristics of both rural township clusters and the area immediately surrounding the addresses of rural residents.
Cross-sectional associations must not be interpreted causally; therefore, implications of correlational research must be considered with care. More work needs to be done in rural areas to investigate the growing burden of obesity. Rural residents appear particularly vulnerable to obesity, and this geographic differentiation alone suggests that there is something unique to rural living that is obesogenic. Further studies are needed to clarify the current findings, which suggest that for those who are already obese and living in rural areas, area characteristics may have little influence over their degree of obesity or efforts to lose weight.
This work addresses an important issue in a most vulnerable subsection of the US population, obese residents of rural areas. This study boasts a sizeable sample, anthropometric assessments of BMI, and innovative analyses strategies drawn from a multidisciplinary research team to investigate multilevel determinants of obesity and weight loss attempts. Despite this strength, only obese clinic patients participated in this study. As a result of the restriction of range of weight status, neighborhood characteristics that distinguish between normal weight and overweight or obese individuals may not be useful for distinguishing between different degrees of obesity (obese vs. morbid obese). This may suggest that individual level factors may be more important for preventing and controlling the escalation from obesity to morbid obesity, or that those who are morbidly obese are unique along dimensions unrelated to neighborhood.
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSMeasuresRESULTSACKNOWLEDGMENTSREFERENCES |
|---|
Received March 22, 2005. Accepted November 17, 2006.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSMeasuresRESULTSACKNOWLEDGMENTSREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
