Yuka Asada, Ph.D.
Institute for Health Research & Policy, University of Illinois at Chicago

Background

The Healthy, Hunger-Free Kids Act—a federal policy response to childhood obesity—advanced local wellness policy requirements that impact population health for K-12 students. Measuring wellness policy implementation is complex, given dynamic and multi-level school environments. Yet these intermediary outcomes are critical to influencing ‘obesogenic’ school environments. Rigorous qualitative methods allow for complex implementation processes to be identified by key stakeholders; however, few studies apply qualitative approaches in a comprehensive and theory-driven manner. This poster presentation will outline the evidence-informed application of qualitative methods to examine wellness policies as an approach to obesity prevention.

Methods

The National Wellness Policy Study (NWPS) is a mixed methods study, employing a series of qualitative focus group, key informant interviews, and qualitative secondary analysis (QSA) to examine policy implementation at district and school levels. The study intentionally included key informants from a range of stakeholders, including food service directors, superintendents, high school students, and parents. Focus groups and interviews were transcribed and uploaded into Atlas.ti Qualitative Data Analysis Software v8. Transcripts were team coded and analyzed using constant comparative analysis. The study design and execution included activities to enhance ‘trustworthiness,’ also known as validity and reliability.

Results

The methods aligned with best practices and principles of ‘trustworthiness’ identified by Consolidated Criteria for Reporting Qualitative Research (COREQ) and O’Brien et al. (2014). The study utilized an iterative design to allow findings to inform subsequent data collection. Instruments were developed from existing literature, and informed by practitioner and academic expert feedback. Coding guides and analysis were also iteratively developed through team coding activities and several rounds of feedback as themes emerged. As a result, the methods generated policy- and practice-relevant findings for school professionals, as well as for the academic literature.

Conclusions

Examining the impact of policies on ‘obesogenic’ environments requires a method that aptly measures the dynamic, process-related factors and experiences that elucidate how policies are translated into practices on the ground. This poster presentation advocates for the application of comprehensive and evidence-informed qualitative methods as they apply to natural experiments to more effectively address obesity prevention at a population level.

Rahmatollah Beheshti, Ph.D.
Johns Hopkins University

Background

Computational models have gained popularity as a predictive tool for assessing proposed policy changes affecting dietary choice. Specifically, they have been used for modeling dietary changes in response to economic interventions, such as price and income changes.

Methods

Herein, we present a novel addition to this type of modeling by incorporating habitual behaviors that drive individuals to maintain or conform to prior eating patterns. Specifically, we show how to develop an agent-based model (ABM) for simulating the actual food choices of individuals. Using the information such as the mean diet of the population and price elasticity of demand, a set of realistic behaviors are defined for each simulated individual. We demonstrate how our method can capture some of the aspects of food choice that cannot be modeled using existing techniques. We examine our method in a simulated case study of food choice behaviors of low-income adults in the US. We use data from several national datasets, including the National Health and Nutrition Examination Survey (NHANES), the US Bureau of Labor Statistics and the USDA, to parameterize our model and develop predictive capabilities in (1) quantifying the influence of prior diet preferences when food budgets are increased and (2) simulating the income elasticities of demand for four food categories. Food budgets can increase because of greater affordability (due to food aid and other nutritional assistance programs), or because of higher income.

Results

Our model predictions indicate that low-income adults consume unhealthy diets when they have highly constrained budgets, but that even after budget constraints are relaxed, these unhealthy eating behaviors are maintained. Specifically, diets in this population, before and after changes in food budgets, are characterized by relatively low consumption of fruits and vegetables and high consumption of fat. The model results for income elasticities also show almost no change in consumption of fruit and fat in response to changes in income, which is in agreement with data from the World Bank’s International Comparison Program (ICP).

Conclusions

The proposed method can be used in assessing the influences of habitual dietary patterns on the effectiveness of food policies.

Jessica Breland, Ph.D.
VA Palo Alto Health Care System

Background

Electronic medical records (EMRs) are increasingly used to evaluate natural experiments in obesity. We assessed whether two EMR-based algorithms for obesity case finding—Body Mass Index (BMI; from weight and height data) and ICD-9 diagnosis codes—yielded different estimates of: (1) obesity prevalence and (2) comorbidity prevalence among individuals with obesity.

Methods

Based on EMR data from the largest health system in the United States, the Veterans Health Administration (VHA), we created a cohort of the roughly 5 million veteran VHA primary care patients in fiscal year 2014. Algorithm A identified obesity based on a BMI ≥30 kg/m2, calculated from weight most proximate to the individual’s first primary care visit that year, combined with “best” height. Algorithm B used obesity ICD-9 codes. We also used ICD-9 codes to identify physical and mental health comorbidities. We compared obesity prevalence by algorithm, and then comorbidity prevalence among individuals with obesity, by algorithm. Given the large sample size, we used a ≥5% cut-off (rather than p-values) to indicate clinically meaningful differences.

Results

Algorithm A identified almost three times as many individuals with obesity compared to Algorithm B (2,037,810 vs. 741,082 individuals), resulting in a large difference in obesity prevalence depending on the algorithm used (41% vs 15%). The proportion of women versus men was similar between methods as was the prevalence of most comorbidities. Only four comorbidities had clinically meaningful differences in prevalence between algorithms: sleep apnea (9% difference), diabetes (6% difference), hypertension (5% difference), and lipid disorders (6% difference). In all four cases, comorbidity prevalence was higher with Algorithm B.

Conclusions

When using EMR data to assess overall obesity prevalence, an algorithm based on clinical BMI data was superior to one based on ICD-9 codes for obesity. However, when looking at comorbidity prevalence among individuals with obesity, the algorithm drawing on ICD-9 codes identified higher prevalence of some comorbidities, possibly related to obesity severity or frequency of primary care visits among those with ICD-9 obesity diagnoses. Results also demonstrate how VHA’s size, obesity burden, and EMR data make it an important place for future investigations of natural experiments in obesity.

Katie Crist, M.P.H.
University of California, San Diego

Background

Physical inactivity is associated with increased risk of obesity. Research suggests that new transit infrastructure may increase physical activity (PA) through active travel to transit. Accelerometers are considered the gold standard in objective PA measurement; however, transport behaviors are misclassified by accelerometer processing methods. Novel machine learning (ML) algorithms to classify transport mode from Global Positioning System (GPS) and accelerometer data are presented in comparison to traditional accelerometer cut points.

Methods

GPS and accelerometer data were collected in two free living and two prescribed activity cohorts. Annotated images from person-worn cameras provided ground truth of transport behaviors. The cohorts included: free-living adult cyclists (N=40) and overweight and obese women (N=36), and prescribed activities in child cyclists (N=36) and adults (N=2). Random forest classifiers, using multiple features from each device and trained separately on each data set, predicted behaviors for each minute of data. Hidden Markov Modeling improved predictions by modeling the probability of transitioning between different activities. Predictions were evaluated using leave-one-out cross-validation.

Results

Standard accelerometer cut points accurately detected 8.7% of cycling, 36.5% of walking, and 77.2% of sedentary behaviors, including vehicle time. The ML algorithms, trained and tested on adult prescribed activities, had accuracies of 99% – cycling, 93% – vehicle time, 97% – walking. Accuracy in the adult cyclist dataset was 97% – cycling, 94% – vehicle time, and 88% – walking. In overweight/obese women, accuracy was 91% for vehicle time and 85% for walking. No cycling behavior was observed. In child cyclists, 100% of cycling and walking and 99% of vehicle time was correctly classified. Accuracy was higher in algorithms using features from both data types compared to either GPS or accelerometer data alone.

Conclusions

Traditional accelerometer cut points misclassify transport-related behaviors, underestimating PA from cycling and walking and incorrectly classifying vehicle time as light-intensity activity. Algorithms are most accurate when applied to populations with similar characteristics as the training data set. Machine learning techniques, using accelerometer and GPS data sources, greatly improve our ability to accurately identify transport mode and estimate associated PA.

Sara Grajeda, Ph.D.
University of Delaware

Background

In 2012, Supplemental Nutrition Assistance Program (SNAP) participation increased at U.S. farmers markets from 750 markets in 2008 to 3,200. Despite such change in the food landscape, there is very limited information known about purchasing and consumption, and the relationship between the two, of participants using SNAP at farmers markets. The 2014 Farm Bill established the $31.5 million USDA Food Insecurity Nutrition Incentive (FINI) grant program to increase the amount of fruits and vegetables purchased and consumed among participants in SNAP. FINI funds support this natural experiment, in which we sought to discover purchasing and consumption data patterns across seven non-consecutive months (Sept. – Nov. 2015 and Jan. – Apr. 2016) from 649 SNAP consumers at 77 farmers markets across 14 states.

Methods

Using an online survey, including all fruit and vegetable (FV) questions from a modified version of the National Health and Nutrition Examination Survey(NHANES) Dietary Screener Questionnaire (DSQ) and a two-item food insecurity screener on health status and perception from the Behavioral Risk Factor Surveillance Questionnaire (BRFSS). A DSQ algorithm was applied to estimate participants’ daily FV consumption. Survey responses were received from 2,520 adult SNAP participants, primarily white (77.7%), female (82.6%), and overweight (61.5% with a body mass index (BMI) of 25 or above).

Results

On average, participants spent $394.86 monthly per household on food and drink, 45.5% of which was reported spending on FV (35% more than national SNAP participants), which resulted in the consumption of 3.04 cups (around 0.5 cups more than the national average for SNAP shoppers) of FV daily. Additionally, SNAP consumers at farmers markets have different health behaviors.

Conclusion

Although findings are generally better than the national SNAP populace, dietary patterns still fail to meet national recommendations for a healthy diet. This warrants further investigation and potential intervention targeting SNAP shoppers at farmers markets.

Laura Hopkins, M.S.P.H.
The Ohio State University

Background

Racial/ethnic minority children from low-income households are at risk for unhealthy weight gain during the summer months. Routine participation in structured programming (e.g., USDA Summer Food Service Program) may have a protective effect. The objective of this study is to determine if and to what extent routine engagement in structured programming during the summer protects against inappropriate weight gain among economically disadvantaged African American school-age children.

Methods

Two elementary schools in low-income neighborhoods of Columbus, OH, were recruited. At the end of the 2016–17 school year, families with pre-K–5th graders were invited to participate. Caregivers completed a demographic survey, including questions: (1) about intent to send their child(ren) to summer programming and (2) to participate in a texting program to track child attendance in summer programming. Biometric data—height(cm), weight(kg), waist circumference(WC)(cm), systolic(s)and diastolic(d)blood pressure (BP)(mmHG)—were obtained from children at the end of the school year. Additional data collection will occur at the beginning of and 3 months into the 2017–18 school year. Each week, caregivers receive the following text: “Hi from Project SWEAT! How many days did your child attend a summer camp/program this week? Respond with a number from 0 to 5 with “0” meaning no days, “1” meaning 1 day, etc.” Non-responders receive a follow-up text and phone call.

Results

113 children representing 79 families enrolled. Mean age was 7.10±0.21yr, 79.65% were African American, and 72.73% low-income. At baseline, child mean zBMI, zWaist Circumference, zSBP, and zDBP were 0.75±0.10, 0.51±0.08, 1.54±0.11, and 0.97±0.13, respectively. At baseline, it was reported by caregivers that 33.63% of children would attend summer structured programming with 19.51%, 12.20%, and 68.29% intending to attend 1–2 weeks, 1–2 days a week, and almost every day, respectively. Data collection is ongoing and analysis is forthcoming.

Conclusions

This study can be expected to have a significant impact by providing information on the factors that protect underserved children from unhealthy summer weight gain, which may be used by stakeholders in policy reform.

Linghui Jiang, M.P.H.
University of California, Los Angeles

Background

Childhood obesity is a serious public health problem in the US. Although the prevalence of childhood obesity has plateaued in recent years, obesity still disproportionately affects young children from low-income families. There is some evidence suggesting that breastfeeding is associated with many health benefits, including decreased risk of childhood obesity. Given such evidence, the World Health Organization and the United Nations Children’s Fund launched the Baby-Friendly Hospital Initiative (BFHI) to promote the health and well-being of children. Although BFHI has shown promising results in promoting breastfeeding in the short-term, its long-term impact on children’s weight trajectory is less clear. This study aimed to estimate the population impact of BFHI via increased breastfeeding on reducing obesity risk and project the mean weight-for-height Z (WHZ) score under various levels of BFHI reach and fidelity among low-income preschool-age children in Los Angeles County (LAC).

Methods

Using data from published literature and from the Special Supplemental Program for Women, Infants and Children (WIC), a hypothetical realistic population of WIC beneficiaries living in LAC was simulated. We then estimated the path-specific effect of BFHI through breastfeeding on WHZ using the causal modeling method known as the parametric g-formula algorithm. The simulations varied the reach and fidelity of BFHI intervention. Reach refers to the proportion of the WIC population reached by BFHI, while fidelity refers to how well BFHI was implemented.

Results

Comparing 100% to 0% implementation of BFHI intervention, the population impact of BFHI on WHZ score, in terms of WHZ population mean difference, was -0.015 (95%CI: -0.024 to -0.006). Correspondingly, the mean projected WHZ score dropped from 0.816 to 0.791, and the projected mean prevalence of obesity dropped from 15.0% to 14.8% when the fidelity and reach of BFHI intervention varied from 0% to 100%.

Conclusions

Intervening to implement BFHI more widely appears to lower the risk of childhood obesity ultimately by improving breastfeeding among low-income preschool-age children. Interestingly, the projected mean WHZ and prevalence of obesity varied little with intervention fidelity and reach. This may underscore the need to examine complex roles of relevant community and individual risk factors for childhood obesity.

Micaela Karlsen, M.S.P.H.
Tufts Friedman School of Nutrition Science and Policy

Background

Adhering to Dietary Approaches for Personal Taste (ADAPT) Feasibility Survey assessed the practicality of web-based survey methods to capture demographic data and self-reported body weight from individuals following a wide range of popular diets. The objective was to demonstrate feasibility of recruiting a cohort to examine factors associated with long-term adherence.

Methods

A convenience sample was recruited using social media to participate in a short web-based survey on dietary preferences. Self-reported data were collected on demographics, height, weight, and current and past diets followed. Current reported diets were collapsed for analysis as follows based on reported frequency: whole food plant-based (WBPB, n=2344; 26%); vegan & raw vegan (n=1763; 20%); Paleo (n=1326; 15%); try to eat healthy (n=1048; 12%); vegetarian & pescatarian (n=883; 10%); whole food (n=754; 8%); Weston A. Price (n=493; 5%); and low-carb (n=408; 5%).

Results

Of 9,019 individuals with complete data, 82% were female, 93% white, 96% non-Hispanic, and 84% completed the survey in the US. Among a subset with complete physical activity data (n=4,679) the majority reported being moderately (35%) or highly (63%) physically active. A total of 49% (n=4,420) reported following their diet for 1 to 5 yrs. Among this subset, “try to eat healthy” geometric mean BMI (26.6 kg/m2) was significantly higher (p<0.05) than all other diet groups’ after adjusting for age, sex, location, and time on diet. Mean BMI ranged from 23.2 kg/m2 for WFBP followers to 25.3 kg/ m2 for vegetarians. Among all 9,019 participants, 87% gave permission to be contacted to participate in a larger study; data collection for this study is currently underway.

Conclusions

Self-reported BMI appears to be lower among adults who made an active decision to adhere to a specific diet compared to those who simply “try to eat healthy.” Future research will prospectively examine aspects of behavior change that influence successful adherence to popular diets and how these changes translate to changes in health outcomes.

Gregory Knell, Ph.D.
The University of Texas Health Science Center (UTHealth) at Houston

Background

It is thought that increasing and improving sidewalk infrastructure may lead to more physical activity by providing safe, defined, and connected walking spaces. However, evidence linking sidewalk infrastructure with physical activity remains inconclusive. This study examined the association of living near recently installed or improved sidewalks with changes in self-reported and accelerometer-derived physical activity among a diverse, community-based sample of adults from the Houston Travel Related Activity in Neighborhoods (TRAIN) Study, a natural experiment.

Methods

Baseline (2014–2016) and one-year follow-up (2015–2017) TRAIN data (median interval = 14.3 months) were used for this analysis. The exposure of interest was an ordinal count of the number of recent sidewalk improvements, categorized as 0, 1, 2, ≥3 improvements, within 250-meters of a participant’s residential address. The outcomes of interest were continuous estimates of one-year changes in physical activity measured using self-reports (n=430) and accelerometry (n=228). Two-step regression models, adjusted for covariates, were built to test the hypothesis that living near a recently improved sidewalk was associated with higher levels of physical activity at follow-up relative to those not living near an improved sidewalk.

Results

The majority of participants were female, non-Hispanic black, low income, and low education. Forty-six percent of participants lived near at least one sidewalk improvement. After adjustment, among participants reporting some physical activity at baseline, living near 2 recent sidewalk improvements was associated with 1.58 times more self-reported minutes per week of walking, and 1.60 times more MET-minutes per week of leisure-time physical activity at follow-up relative to those not living near any recent sidewalk improvements (p<0.05). There were no statistically significant associations based on accelerometry.

Conclusions

Although these mixed findings warrant further research, the results suggest that sidewalks may contribute to increases in walking and leisure-time physical activity among those with some baseline physical activity. Future work is warranted to elucidate these findings, and should also examine the effect of behavioral interventions alongside changes to the built environment.

Stephanie Pike Moore, M.P.H.
Case Western Reserve University

Background

Food retail-based initiatives are a focal point for addressing disparities in obesity and other diet-related chronic conditions. Methods for evaluating food retail-based interventions often focus on measuring changes within one targeted food retailer (e.g., changes in a corner store). However, food retailers exist within a dynamic system that responds to fluctuations related to both supply and demand. To better evaluate effects of food retail-based interventions, methods are needed to rigorously assess granular, areal changes in the food environment. As part of an NIDDK-funded, rapid-response, quasi-experimental, natural experiment focused on evaluating the impact of a new food hub and related social food access programming, we describe methods for examining changes in the food environment, including within-store variability and areal-level changes in the food retail choice context (FRCC). Measures take into account availability, price, quality, and advertising of healthy foods in stores in the intervention (IC) and comparison (CC) communities over two years.

Methods

In 2015 and 2016, we evaluated every food store in matched, limited-healthy-food-access communities located in two cities. In the IC, 2015 data collection occurred before food hub implementation, and by 2016, food hub social access programming included implementation of two additional food retail projects in corner stores. In the CC, by 2016 one food retail-based corner store intervention was being implemented through efforts unrelated to this study. Trained researchers evaluated all stores using an adapted Nutrition Environment Measures Survey in Corner Stores (NEMS-CS) tool to assign point values for availability, price, and quality of food options in stores, and an adapted Bridging the Gap: Community Obesity Measures Project (BTG-COMP) tool to assign point values for advertising. Stores were rated low, medium, and high based on NEMS-CS/BTG-COMP scores. Stores were mapped with a ½-mile network buffer. FRCCs were assigned to census blocks to reflect the spatial overlap of store types with the highest category representing areas with the highest availability, lowest pricing, and best quality of healthy food items and lowest proportion of unhealthy advertising.

Results

Fifty-five stores were evaluated each year: 34 and 36 in the IC and 21 and 19 in the CC for each year, respectively. In the IC, 25% of stores evaluated in both years had a substantial improvement in NEMS-CS/BTG-COMP scores, whereas only 6.3% of stores in the CC saw this same increase, and 6.3% experienced a decrease. None of the stores receiving a food retail-based intervention saw significant changes in NEMS-CS/BTG-COMP scores. In terms of FRCC, 14.9% of the blocks in the IC saw an improvement in FRCC compared to only 2.3% of blocks in the CC, where an additional 8.5% of the CC blocks saw a decrease in FRCC.

Conclusions

This measurement approach may capture system-level changes associated with food retail-based interventions because they are sensitive to small changes in availability, price, quality, and advertising of healthy food items.

Drew Sayer, Ph.D.
University of Colorado Anschutz Medical Campus

Background

Increasing physical activity (PA) is a cornerstone of behavioral weight loss therapies, but maintaining high levels of PA is difficult for most individuals. Convenient access to an exercise facility may support greater PA. The purpose of this study was to determine if gym usage (Anschutz Health and Wellness Center, AHWC) differed between participants who worked on the University of Colorado Anschutz Medical Campus (ON) vs. subjects who worked off-campus (OFF) during their participation in a 6-month randomized weight loss trial.

Methods

Participants were provided free memberships to the AHWC during the participation in the trial. Gym usage was tracked via the participants’ membership key cards. Leisure-time PA was self-reported by participants using a 7-day PA recall at baseline and at months 4 (end of group-based weight-loss intervention) and 6 (follow-up assessment) of the trial.

Results

Data were available for 117 adults (ON: n=62, OFF: n=55) with obesity. At baseline, ON had lower body weight (94.5 ± 19.8kg vs. 108.3 ± 24.5 kg, p=0.001) and BMI (33.5 ± 6.4 kg/m2 vs. 38.0 ± 7.0 kg/m2, p=0.0004) than OFF. When controlling for age, baseline BMI, and home distance to the AHWC, gym usage was greater for ON (0.93 ± 1.2 times/week) than OFF (0.55 ± 2.4 times/week, p=0.04). Total leisure PA was not different between ON and OFF at baseline (646±1002 vs. 521±819 MET-minutes/week, p=0.5). ON increased leisure PA more than OFF at Month 4 (2559±1652 vs. 1867±1602 MET-minutes/week, p=0.04) but not at Month 6 (1825±1745 vs. 1894±1754 MET-minutes/week, p=0.9).

Conclusions

More convenient access to the AHWC gym was associated with greater usage of the exercise facility and leisure-time PA during weight loss, but not in the follow-up period. These results suggest fitness centers should consider both the convenience of access to the center and the availability of exercise and/or weight loss programming for increasing center usage and PA among members.

Tasia Smith, Ph.D. and Elizabeth Budd, Ph.D.
University of Oregon

Background

Childhood obesity is associated with adverse outcomes that can persist into adulthood, including continued obesity, type 2 diabetes, and hypertension. Understanding the underlying mechanisms that contribute to childhood obesity is critical for prevention efforts. Research has identified both genetic and environmental determinants that influence children’s risk for obesity. However, parsing the relative effects of these determinants is more challenging. The present study uses a novel design, which included biological and adoptive families, to examine how the rearing environment and genetics differentially influence the body mass indices (BMI) of children.

Methods

In this quasi-experimental feasibility study, the sample included 115 linked sibling pairs, sometimes both residing in the same home (either adoptive or birth parent), and sometimes residing in different homes (one sibling in the adoptive home and one sibling in the birth parent home). Parent-reported height, weight, age, and gender were used to calculate BMI scores for all children. Self-reported height and weight were used to calculate parents’ BMI scores. Pearson correlations were conducted to examine familial BMI similarity as a function of genetic relatedness and rearing environment.

Results

There were no significant relationships between the adopted children’s BMI and their biological parents’ BMI (r = .16, ns) or their genetically unrelated siblings’ BMI (r = -.11, ns). Trending toward significance were the relationships between the adopted children’s BMI and their adoptive parents’ BMI (r =.25, p =.08)–suggesting environmental influences on child weight, and their reared-apart biological siblings’ BMI (r = .32, p = .07)–suggesting genetic influences on weight.

Conclusions

Most studies on childhood obesity are limited to information about biological family members, and therefore do not distinguish whether intrafamilial associations are due to heritable or to environmental causes. The results of this study begin to confirm the dual influence of genetics and the rearing environment on childhood weight status. A next step is to identify modifiable characteristics of rearing environments (potential intervention targets) that are not confounded by genetics, which influence children’s risk for obesity.

Calvin Tribby, Ph.D.
National Cancer Institute, National Institutes of Health

Background

An emerging area of public health research is using natural experiments to estimate the effect of built environment modifications on neighborhood residents’ physical activity levels. The potential for built environment modification to support more active lifestyles is an important intervention to address the obesity epidemic. The residents’ self-defined neighborhood provides a novel exposure measure to assess change in physical activity associated with an intervention, compared to researcher-defined exposure measures, such as a fixed distance.

Objectives: The main research question is: Is the inclusion of a built environment intervention within study participants’ self-defined neighborhoods associated with an increase in the neighborhood physical activity levels?

Methods

This research uses Global Positioning System (GPS), accelerometer, and digitized, participant-drawn, self-defined neighborhood data from the MAPS research project in Salt Lake City, Utah for 2012 and 2013. We use a natural experiment study design with Geographic Information System (GIS) and Python scripting methods to analyze the data. First, we test the change in GPS-PA between years within and outside of the self-defined neighborhood for participants whose neighborhood intersects the intervention and those whose neighborhood does not, using the 2012 self-defined neighborhood as the constant spatial unit. We compare this to a 1 km residential buffer, a common neighborhood definition. Second, we model the change in neighborhood GPS-PA as a function of whether the intervention intersects the neighborhood, controlling for sociodemographic variables and GPS wear time.

Results

Paired t-tests for the difference between years for individuals’ unadjusted GPS-PA are insignificant for the group whose self-defined neighborhood intersects the intervention (t=0.09, df=159, p=0.93) and the no intersect group (t=-1.62, df=105, p=0.11). Similarly, the change in GPS-PA is insignificant outside the neighborhood for the intersect group (t=-0.25, df=159, p=0.80) and no intersect group (t=-1.13, df=105, p=0.26). For comparison, the difference between years for the individuals’ GPS-PA within 1 km residential buffer that intersects is insignificant (t=0.14, df=167, p=0.89), whereas the group of individuals’ whose neighborhood does not intersect is significant (t=-1.98, df=119, p=0.05).

Conclusions

Self-defined neighborhoods provide a novel spatial area that offers insight into classifying participants as exposed or unexposed to built environment modifications in natural experiments, compared to residential buffers. Exploratory results suggest that the relationship between self-defined neighborhoods and a built environment intervention is stable between years. However, it is important to consider the resident-defined neighborhoods, not only researcher-defined neighborhoods, to assess the built environment and physical activity.