Abstract
Introduction:
Natural experiments can strengthen evidence linking neighborhood food retail presence to dietary intake patterns and cardiometabolic health outcomes, yet sample size and follow-up duration are typically not extensive. To complement natural experiment evidence, longitudinal data were used to estimate effects of neighborhood food retail presence on incident disease.
Methods:
The Cardiovascular Health Study (CHS) recruited adults age 65+ years in 1989–1993. Analyses conducted in 2021–2022 included those in good baseline health, with addresses updated annually through year of death (restricted to 91% who died during >2 decades of cohort follow-up). Baseline and annually updated presence of two combined food retail categories (supermarkets/produce markets and convenience/snack focused) was characterized using establishment-level data for 1-km and 5-km Euclidean buffers. Cox proportional hazards models estimated associations with time to each incident outcome (cardiovascular disease [CVD], diabetes), adjusting for individual and area-based confounders.
Results:
Among 2,939 participants, 36% with baseline supermarket/produce market presence within 1 km had excess incident CVD (HR: 1.12; 95% CI: 1.01 to 1.24); the association was attenuated and no longer statistically significant following adjustment for sociodemographic characteristics. Adjusted associations were robustly null for time-varying supermarket/produce market or convenience/fast food retail presence across analyses with outcomes of CVD or diabetes incidence.
Conclusions:
Food environment changes continue to be studied to provide an evidence base for policy decisions, and null findings in this longitudinal analysis add literature that casts doubt on the sufficiency of strategies targeting food retail presence alone of an elderly cohort for curtailing incident events of clinical importance.
Keywords: Residence Characteristics, Food environment, Access to Healthy Foods, Cardiovascular Disease, Longitudinal Study
Introduction
Modifiable risk factors, including individual dietary patterns, are associated with cardiovascular disease (CVD) incidence.1 CVD is a leading cause of death in the US.2 A lack of healthy food retail within neighborhoods may constrain individuals’ dietary choices.3–8 Built environment characteristics may be incorporated into health-related behavior change strategies to prevent CVD and related conditions such as diabetes.1,9–11 For example, local initiatives have brought new supermarkets or produce markets into areas previously lacking affordably-priced fresh foods.12,13
Natural experiments have examined health effects of changing availability of supermarkets, produce markets, or similar food stores.14–20 For produce markets, a cluster randomized trial noted increased self-efficacy for regularly integrating healthy food into one’s diet.20 However, natural experiments of supermarket presence do not uniformly support hypothesized benefits including increased fruit and vegetable intake15,16 or healthier items purchased.14 In one study, dietary quality improvements in a neighborhood with a new supermarket were experienced regardless of regular new store use, raising the possibility that benefits were due to other concurrent neighborhood change.17 Indeed, expectations of health benefits from new supermarket presence should be tempered by awareness of frequent promotion and purchases of highly-processed items in supermarkets.21–23
Place-based experiments face challenges and limitations, which may be shared between food retail presence studies and those connecting place and health through psychosocial or physical activity pathways.24 Comparative prospective longitudinal data collection is costly, limiting sample size and follow-up duration. Limited statistical power is a challenge as effects on health may be small and slow to emerge, especially for cardiometabolic risk factors or clinical events. Associations may not generalize across stores due to differences in cultural acceptability, perceived quality, and pricing. Associations of food retail change with incident disease may depend on individual characteristics such as age or process of food environment change including the level of community engagement or concurrent community development. Thus, questions remain about average effects of food retail change on cardiometabolic health.
To complement evidence from natural experiments, this study estimated associations of baseline and annually updated food retail presence near participants’ homes with incident CVD and diabetes among older adults using Cardiovascular Health Study (CHS) cohort data. Food retail sources were classified as supermarkets/produce markets or convenience/fast food. The focus on presence of each type, versus alternatives such as proximity or density, was informed by a systematic review of food environment and obesity studies25 which noted the highest proportion of significant findings in the expected direction among studies examining presence of food retail. In stratified analyses, time-varying information on food retail presence of each type was examined among those lacking such retail at baseline (for whom newly available food retail could be associated with diabetes or CVD incidence); analyses likewise considered time-varying presence in contexts that did have food retail of a given type present at baseline (informative about health effects of sustained presence versus closure).
Methods
Study Population
Longitudinal personal and health data were from CHS.26,27 Eligible participants age 65+ at baseline were from the original CHS cohort (n=5,201) recruited 1989–1990 and the primarily African-American cohort (n=687) recruited 1992–1993. Participants provided informed consent. Study protocols were approved by Institutional Review Boards at participating institutions (Columbia University, Drexel University, and University of Washington). Geographic characterization was restricted to 5,384 decedents (91% known to be deceased in August 2016). This restriction was done while preparing for linkage to LexisNexis personal profile data (for residential history28) to limit risks of sharing personally identifying data.
Inclusion was restricted to participants with both self-reported excellent, very good, or good self-rated health status29 and self-reported ability to complete activities of daily living30 at baseline (N=3,792); those with impaired health or functional status may be less sensitive to nearby food retail presence. Those with prevalence of either outcome at enrollment were excluded from corresponding analyses, as were those for whom baseline address was not successfully geocoded or with missing covariate data, resulting in 2,939 and 2,497 participants for analyses of CVD and diabetes, respectively.
Measures
Incident CVD included CHS events committee adjudicated events: myocardial infarction, congestive heart failure (not procedure-related), stroke, or cardiovascular death.26,27 Criteria, identification, and adjudication process have been described.26,27
Diabetes was defined for the first decade of follow-up, in which annual visits were conducted, as either elevated serum glucose (fasting of ≥126 mg/dL or random [non-fasting] ≥200 mg/dL) or use of diabetes treatment medication; subsequently, diabetes medication alone was used. As a secondary analysis, electronic record linkage to Center for Medicare and Medicaid Services through 2010 was used; for this secondary analysis, exclusion of prevalent diabetes was based on diagnosis by year 5, the origin for corresponding time-to-event analyses.
The main geographic area was a 1-km radial buffer, chosen to represent a walk 10–20 minutes from home for healthy older adults, an area likely to be frequently traversed via walking or vehicle.31,32 Pre-specified sensitivity analyses used 5-km radial buffers,33 representing a short driving distance to include broader food environments relevant to adult food acquisition trips.34
Food retail presence by type was updated annually, based on commercially sourced spatial data and home address updates. Categorization of National Establishment Time Series (NETS) (1990–2014) data has been described.33 The NETS data were licensed from Walls & Associates, which used Dun & Bradstreet data that include a comprehensive census of US establishments each January.
Establishment categorization relied on 8-digit Standard Industrial Classification (SIC) codes, informed by systematic spot checks within 81 ambiguously-labeled SIC codes across 30 locations of varying urbanicity spanning 10 US census regions.
Food retail establishment SIC codes were classified based on prior literature.31,33,35,36 Supermarkets and produce markets may be health-supportive based on availability and affordability of fresh foods and usual intent of customers, while recognizing that highly processed items are also available and often promoted and purchased at supermarkets.21–23 In addition to SIC codes, supermarket definition used large size (≥25 employees or annual sales ≥$2 million) or chain name searches based on licensed industry reports (Retail Trade Channel Database compiled by TDLinx, a Nielsen Holdings subsidiary; subchannels: “Supermarket-Conventional,” “Supermarket-Limited Assortment,” “Supermarket-Natural/Gourmet Foods,” and “Warehouse Grocery”). Produce markets were defined using SIC codes for stand-alone establishments primarily selling fruits and/or vegetables.
Convenience/fast food retail included fast food, quick service, and pizza restaurants; bakery, ice cream, coffee, and candy shops; and convenience and small grocery stores. While establishments offer items with varying nutritional value, categorization as convenience/fast food retail was informed by affordability and salience of highly processed and calorie dense items and emphasis on convenience of consumption. Definitions used specific 8-digit SIC codes or a search within establishments with food service SIC codes for key words such as “pizza” or national chain names (revenue-ranked lists of 400 restaurant chains in 2001 and 2003–2008 from Restaurant & Institutions and of 250 restaurants in 2014–2015 from Technomics).
Baseline presence of supermarkets/produce markets and convenience/fast food retail was dichotomized as present versus absent. Time-varying analyses were conducted using annually updated presence for each unit of person-time. To emulate a natural experiment, analyses considered time-varying food environment measure within strata defined by baseline presence. Among those without any supermarket/produce market present at baseline, time-varying supermarket/produce market presence is relevant to estimating effects of gaining new food retail of this type. Among those with any supermarket/produce market at baseline, time-varying supermarket/produce market presence is relevant to effects of sustaining (versus losing) at least some food retail of this type. Analyses considered time-varying convenience/fast food retail presence overall and within strata based on convenience/fast food retail presence at baseline. Some areas had one transition over follow-up (e.g., from absent to present, or from present to absent); other areas experienced fluctuations.
Potential individual-level confounders included clinic site (recruitment location) and baseline sociodemographic characteristics (age, gender, race/ethnicity, marital status, and education level), smoking status, and driving status (based on a driving-related vision question37). Analyses did not adjust for characteristics along the causal pathway (e.g., body mass index), but did examine effects of adjustment for baseline self-reported walking ability (categorized as did not walk, walked slower than a normal pace, walked normal pace or faster) which may have affected neighborhood exposures through decisions about where to live during follow-up.
Area-based socioeconomic context (median household income, percent of older residents in poverty, and percent of adult residents with ≤high school education) and built environment context (population density and walkable destination density38) were estimated at baseline. Area-based characteristics other than food retail presence by type and walkable destinations were estimated using data from population census and American Community Survey data, harmonized in Brown University’s Longitudinal Tract Database.39 To estimate values at year of enrollment, decennial data were linearly interpolated for the years 1990–1993 and carried backward from 1990 to 1989.
An area-weighted spatial overlay was used to estimate characteristics of buffer areas (1-km, or 5-km buffers in sensitivity analysis) from census tract polygons. Area-weighting relies on the simplifying assumption that populations are uniformly distributed within census tracts. Total population count and population count within sociodemographic strata were estimated based on proportion of each intersecting census tract with the buffer; if a buffer included 100% of tract A and 25% of tract B, buffer population would be estimated as 100% × (Population of tract A) + 25% × (Population of tract B). Weighted averages for estimation of non-count characteristics such as median household income depended instead on proportion of the buffer that each tract represents. If tract A represents 80% of the buffer land area and tract B represents 20% of the buffer land area, median household income of the buffer would be estimated as 80% (Median household income of tract A) + 20% (Median household income of tract B).
Statistical Analysis
Cox proportional hazards regression models examined associations with time to incident CVD or incident diabetes. The origin was study enrollment. Censoring occurred for participants relocated into a nursing home, from which food acquisition is unlikely to rely on nearby establishments; 30 such censoring events were noted for incident CVD. For analyses of incident diabetes, a censoring date corresponding to end of follow-up for this outcome was assigned immediately after the participant’s final opportunity for ascertainment (day after the last attended clinic visit or medication inventory). Failure date was either CVD event date or date of diabetes ascertainment. Death was counted as failure for fatal incident CVD in models of CVD, and otherwise treated as censoring. In all time-varying analyses, person-time in which a given food retail type (healthy or unhealthy) is present was compared to when that food retail type was absent.
Tests for proportional hazards assumption violations found no statistically significant violation for the exposure of interest. However, as there was a proportional hazards violation for sex, sex-stratified models were also fitted, and other information about proportional hazards violations (which suggests caution in interpreting covariate hazard ratios) is provided in the online appendix (Appendix Tables 1–2 and Appendix Figure 1).
Hazard ratios (HRs) and 95% confidence intervals (CIs) presented used a traditional level of statistical significance (p<0.05); an alternative critical value (p<0.01) was identified to address multiple testing and no adjusted associations for exposures of interest reached this level of statistical significance.
Results
Of 2,939 participants included in analyses of incident CVD, 36% had any supermarket or produce market within 1 km of their baseline home address. Supermarket/produce market presence at baseline was associated with being recruited in Pittsburgh, PA or Davis, CA and with greater walkable destination and population density (Table 1).
Table 1.
Individual- and Community-Level Characteristics by Baseline Supermarket/Produce Market Presence
| Characteristics | All (N = 2,939) |
No Supermarket/Produce Marketa at Baseline (N = 1,875) |
Any Supermarket/Produce Market at Baseline (N = 1,064) |
|---|---|---|---|
| Individual-Level Characteristics, % | |||
| Age, years | |||
| 65–69 | 33.0 | 33.4 | 32.1 |
| 70–74 | 34.3 | 35.2 | 32.7 |
| >=75 | 32.8 | 31.4 | 35.2 |
| Male Gender | 46.5 | 46.1 | 47.3 |
| Black Race | 3.6 | 2.8 | 4.0 |
| Clinic b | |||
| Bowman Gray | 22.7 | 27.5 | 14.3 |
| Davis | 29.4 | 25.1 | 37.0 |
| Hopkins | 24.3 | 28.9 | 16.0 |
| Pittsburgh | 23.6 | 18.5 | 32.7 |
| Status as a Current Driver | 95.0 | 95.4 | 94.2 |
| Smoking Status | |||
| Never Smoked | 44.9 | 46.8 | 41.5 |
| Former Smoker | 43.8 | 42.2 | 46.7 |
| Current Smoker | 11.3 | 11.0 | 11.8 |
| Marital Status | |||
| Married | 70.6 | 72.5 | 67.2 |
| Widowed | 21.9 | 21.3 | 23.1 |
| Other | 7.5 | 6.2 | 9.8 |
| Educational Attainment | |||
| Less than HS | 23.7 | 24.9 | 21.6 |
| HS or More | 52.9 | 52.9 | 52.9 |
| College or More | 23.4 | 22.2 | 25.5 |
| Walkable Destination Countc Above Median | 50.3 | 31.6 | 83.2 |
| Community-Level Characteristics Defined for 1-km Buffers, Mean (SD) | |||
| Population Density (per km2) | 1347 (1329) | 898 (940) | 2139 (1530) |
| Median Household Income, in Thousands per Year | $33.2 ($10.6) | $34.9 ($10.7) | $30.3 ($9.6) |
| Percent of Persons 65+ in Poverty | 1.4% (1.1%) | 1.3% (0.9%) | 1.6% (1.3%) |
| Percent of Persons 25+ with HS Degree or Less | 48.2% (17.6%) | 50.7% (17.4%) | 45.6% (17.5%) |
Values shown are column percent or mean (SD) for included Cardiovascular Health Study (CHS) participants’ individual and 1-km buffer characteristics categorized at baseline.
Supermarket/produce market presence at baseline was defined using the count within 1-km radial buffer centered on a participant’s home address.
Clinic represents CHS recruitment site for this population-based study, though does not necessarily reflect continued location of residence throughout the duration of the study.
Walkable destinations were defined based on the count of establishments with Standard Industrial Codes (SIC) identified as potentially representing either regular walking destinations or pedestrian amenities, dichotomized based on the sample median of 29.
As shown in Table 2, baseline supermarket/produce market presence was associated with excess incident CVD prior to adjustment (HR: 1.12; 95% CI: 1.01 to 1.24). The association was attenuated and no longer statistically significant following adjustment (HR: 1.02; 95% CI: 0.90 to 1.15). Higher population density and lower area-level education were associated with incident CVD (HR: 1.09 per SD of population density; 95% CI: 1.00 to 1.19 and HR: 1.10 per SD for proportion ≤high school degree; 95% CI: 1.00 to 1.20). Time-varying supermarket/produce market presence (person-time with a given food retail type present compared to person-time with that food retail type absent) did not show a trend in the hypothesized protective direction (Table 2).
Table 2.
Associations of Baseline and Time-Varying Supermarket/Produce Market Presence with Incident Cardiovascular Disease
| Independent Variables | Baseline Presence Unadjusted (N=2,939) |
Baseline Presence Adjusteda (N=2,939) |
Adjusteda Time-Varying Presence (N=2,939) |
Adjusteda Time-Varying Presence, Supermarket/Produce Market Absent at Baseline (N = 1,875) |
Adjusteda Time-Varying Presence, Supermarket/Produce Market Present at Baseline (N=1,064) |
|---|---|---|---|---|---|
| Supermarket/Produce Market at Baseline b | |||||
| Absent | Reference | Reference | |||
| Present | 1.12 (1.01–1.24) | 1.02 (0.90–1.15) | |||
| Time-Varying Supermarket/Produce Market Presence b | 1.04 (0.93–1.17) | 0.98 (0.81–1.20) | 1.15 (0.87–1.53) | ||
| Age, years | |||||
| 65–69 | Reference | Reference | Reference | Reference | |
| 70–74 | 1.20 (1.06–1.35) | 1.20 (1.06–1.36) | 1.07 (0.92– 1.26) | 1.46 (1.18–1.80) | |
| >=75 | 1.86 (1.64–2.13) | 1.88 (1.64–2.14) | 1.83 (1.55– 2.17) | 1.99 (1.60–2.48) | |
| Gender | |||||
| Female | Reference | Reference | Reference | Reference | |
| Male | 1.40 (1.25–1.56) | 1.39 (1.24–1.56) | 1.31 (1.13–1.50) | 1.52 (1.26–1.85) | |
| Race | |||||
| Not Black | Reference | Reference | Reference | Reference | |
| Black | 0.79 (0.58–1.08) | 0.81 (0.59–1.12) | 0.70 (0.45–1.11) | 0.95 (0.62–1.47) | |
| Clinic c | |||||
| Bowman Gray | Reference | Reference | Reference | Reference | |
| Davis | 1.00 (0.86–1.18) | 1.01 (0.86–1.19) | 1.15 (0.94–1.40) | 0.79 (0.58–1.06) | |
| Hopkins | 1.00 (0.84–1.18) | 0.99 (0.83–1.17) | 1.05 (0.86–1.23) | 0.77 (0.54–1.09) | |
| Pittsburgh | 0.84 (0.70–1.01) | 0.84 (0.70–1.01) | 0.83 (0.66–1.03) | 0.74 (0.52–1.06) | |
| Status as a Current Driver | |||||
| Yes | 1.12 (0.88–1.42) | 1.13 (0.89–1.43) | 1.23 (0.90–1.69) | 1.00 (0.68–1.46) | |
| No | Reference | Reference | Reference | Reference | |
| Smoking Status | |||||
| Never Smoked | Reference | Reference | Reference | Reference | |
| Former Smoker | 1.09 (0.98–1.22) | 1.10 (0.98–1.23) | 1.15 (1.00–1.32) | 1.07 (0.89–1.29) | |
| Current Smoker | 1.27 (1.06–1.51) | 1.23 (1.03–1.47) | 1.29 (1.03–1.61) | 1.16 (0.86–1.57) | |
| Marital Status | |||||
| Married | Reference | Reference | Reference | Reference | |
| Widowed | 1.21 (1.06–1.39) | 1.22 (1.07–1.40) | 1.30 (1.10–1.54) | 1.11 (0.89–1.39) | |
| Other | 1.02 (0.84–1.24) | 1.05 (0.86–1.28) | 1.08 (0.83–1.42) | 1.02 (0.75–1.38) | |
| Educational Attainment | |||||
| Less than HS | Reference | Reference | Reference | Reference | |
| HS or More | 0.94 (0.83–1.07) | 0.96 (0.84–1.09) | 0.93 (0.79–1.10) | 0.99 (0.79–1.23) | |
| College or More | 0.87 (0.75–1.02) | 0.90 (0.77–1.06) | 0.94 (0.77–1.15) | 0.83 (0.63–1.08) | |
| Walkable Destinations d | |||||
| At or Below Median | Reference | Reference | Reference | Reference | |
| Above Median | 1.14 (0.99–1.31) | 1.13 (0.98–1.29) | 1.16 (0.97–1.39) | 1.08 (0.83–1.40) | |
| Population Density e | 1.09 (1.00–1.18) | 1.09 (1.00–1.19) | 1.08 (0.94–1.23) | 1.08 (0.95–1.23) | |
| Median Household Income e | 1.01 (0.93–1.09) | 1.01 (0.94–1.10) | 1.03 (0.94–1.14) | 0.98 (0.85–1.23) | |
| Persons 65+ in Poverty e | 0.94 (0.88–1.01) | 0.94 (0.88–1.02) | 0.95 (0.85–1.05) | 0.95 (0.85–1.06) | |
| Persons 25+ with HS Degree or Less e | 1.10 (1.00–1.20) | 1.11 (1.01–1.21) | 1.12 (1.00–1.26) | 1.09 (0.94–1.26) |
Values shown are hazard ratios and 95% confidence intervals for Cox proportional hazards models with failure defined as incident cardiovascular disease (myocardial infarction, congestive heart failure [not procedure-related], stroke, or cardiovascular death); there were 1,546 such events as adjudicated by the Cardiovascular Health Study (CHS) events committee during 32,816 person-years of follow-up; bold font indicates statistical significance.
Covariates included age, gender, race, clinic, driving status, smoking status, marital status, educational attainment, and contextual factors including walkable destinations, population density, median household income, poverty among older adults, and adults with a high school education or less.
Supermarket/produce market presence was defined using the count within 1-km radial buffer centered on a participant’s home address, both at baseline and updated annually to reflect whether such retail was present or not at the start of each year throughout follow-up; in all time-varying analyses, person-time with any supermarket/produce market present was compared to person-time with that food retail type absent.
Clinic represents CHS recruitment site for this population-based study, though does not necessarily reflect continued place of residence throughout the duration of the study.
Walkable destinations were defined based on the count of establishments within 1 km of a participant’s baseline home address with Standard Industrial Codes (SIC) identified as potentially representing either regular walking destinations or pedestrian amenities.
Continuous context covariates have been converted to z-scores, such that the hazard ratios represent a contrast of 1 standard deviation.
Time-varying presence of convenience/fast food retail within 1-km buffers was not significantly associated with incident CVD (Tables 3). Neither supermarket/produce market nor convenience/fast food retail were associated with incident diabetes (Table 4). Conclusions were similar across modeling approaches and sensitivity analyses (Appendix Tables 2, 3, 4, 5). Results using 5-km buffers likewise did not significantly support the hypothesized direction of association (Appendix Table 3); there was, however, a statistically significant association of gaining supermarket/produce market presence with higher risk of incident cardiovascular disease among those with no supermarket/produce market present within a 5-km buffer at baseline (HR: 1.68; 95% CI: 1.01 to 2.78) and no statistically significant associations for incident diabetes (data not shown).
Table 3.
Baseline and Time-Varying Convenience/Fast Food Retail Associations with Incident Cardiovascular Disease (N=2,939)
| Independent Variables | Adjusteda Baseline Presence (N = 2,939) |
Adjusteda Time-Varying Presence (N = 2,939) |
Adjusteda Time-Varying Presence, Convenience/Fast Food Retail Absent at Baseline (N = 1,875) |
Adjusteda Time-Varying Presence, Any Convenience/Fast Food Retail Present at Baseline (N = 1,064) |
|---|---|---|---|---|
| Convenience/Fast Food Retail at Baseline b | ||||
| Absent | Reference | |||
| Present | 0.95 (0.82–1.10) | |||
| Time-Varying Convenience/Fast Food Retail Presence b | 1.03 (0.89–1.19) | 1.15 (0.90–1.47) | 1.29 (0.88–1.89) |
Values shown are hazard ratios and 95% confidence intervals for Cox proportional hazards models with failure defined as incident cardiovascular disease (myocardial infarction, congestive heart failure [not procedure-related], stroke, or cardiovascular death); there were 1,546 such events as adjudicated by the Cardiovascular Health Study (CHS) events committee during 32,816 person-years of follow-up; bold font indicates statistical significance.
Covariates included age, gender, race, clinic, driving status, smoking status, marital status, educational attainment, and contextual factors including walkable destinations, population density, median household income, poverty among older adults, and adults with a high school education or less.
Convenience/fast retail presence was defined using establishments including fast food, quick service, and pizza restaurants; bakery, ice cream, coffee, and candy shops; and convenience and small grocery stores within 1-km radial buffer centered on a participant’s home address, both at baseline and updated annually to reflect whether such retail was present or not at the start of each year throughout follow-up; in time-varying analyses, person-time with convenience/fast food retail present was compared to person-time with that food retail type absent.
Table 4.
Associations of Food Retail Presence with Incident Diabetes (N=2,497)
| Independent Variables | Adjusteda Baseline Presence | Adjusteda Time-Varying Presence | Adjusteda Time-Varying Presence, Absent at Baseline | Adjusteda Time-Varying Presence, Present at Baseline |
|---|---|---|---|---|
| Supermarket/Produce Market at Baseline b | ||||
| Absent | Reference | |||
| Present | 1.18 (0.85–1.64) | |||
| Time-Varying Supermarket/Produce Market Presence | 1.06 (0.78–1.45) | 1.05 (0.60–1.84) | 0.87 (0.44–1.71) | |
| Convenience/Fast Food Retail at Baseline c | ||||
| Absent | Reference | |||
| Present | 1.20 (0.85–1.70) | |||
| Time-Varying Convenience/Fast Food Retail Presence | 1.10 (0.78–1.56) | 0.87 (0.43–1.76) | 1.25 (0.50–3.12) |
Values shown are hazard ratios and 95% confidence intervals from Cox proportional hazards models with failure defined as incident diabetes defined as elevated serum glucose or both diabetes treatment medication plus a self-reported diagnosis of diabetes; there were 227 observations of incident diabetes during 33,072 person-years of follow-up.
Covariates included age, gender, race, clinic, driving status, smoking status, marital status, educational attainment, and contextual factors including walkable destinations, population density, median household income, poverty among older adults, and adults with a high school education or less.
Supermarket/produce market presence was defined using the count within 1 km of a participant’s home address, both at baseline and updated annually throughout follow-up; in time-varying analyses, person-time with any supermarket/produce market present was compared to person-time with that food retail type absent.
Convenience/fast food retail presence was defined using the count of such retail within 1 km of a participant’s home address, updated annually through follow-up; in time-varying analyses, person-time with any convenience/fast food retail present was compared to person-time with that food retail type absent.
Discussion
In these longitudinal analyses, hypothesized associations of food retail presence with CVD or diabetes incidence were not supported. Null results were noted in adjusted models for both outcomes, whether models included supermarket/produce market or convenience/fast food retail presence, operationalized using baseline or annually updated time-varying measures.
Prior studies are relevant to understanding whether 1-km and 5-km radial buffers are an appropriate scale for food environment measurement. Distance traveled for food shopping varies by store type.40 Farther distances traveled were reported for superstore shopping as compared with grocery stores (mean was 7.5 km). Thus, some older adults classified as lacking a nearby supermarket/produce market may nonetheless have access to such establishments within a typical travel distance. Additionally, food acquisition is often part of trips with several destinations chained together, suggesting broader activity spaces could relevant.40
Supermarket/produce market presence alone may not be sufficient to improve health. A study conducted in Worcester County, Massachusetts found healthy grocery store proximity supported nutrition improvements.41 However, individuals were encouraged by their physician to seek out healthy food items, potentially amplifying or generating effects. Further, within-store change to promote healthy choices are warranted,42 potentially increasing healthy food sales and consumer knowledge.43 Policies on labeling food and menu items may also influence consumer intake and impact industry practices.44 Future work may benefit from attention food retail site selection45 and closures.46
Determinants of food choice are not limited to the food environment. Self-efficacy and identity, including viewing oneself as a healthy-eater, impact behavior.47 Perceived stress impacts dietary patterns.48 Specifically among the elderly, health-related priorities such as disease management, maintaining independence, and ease of preparation and physical transport may determine food choices.49 Affordability, affected by price and individual assets, is frequently cited as a constraint on healthy food consumption; lack of affordability may be a more influential barrier than distance.13 The effect of poverty itself may outweigh effects of food retail availability.50 Economic incentives including subsidies for fresh food items and taxation on highly processed food and beverage items has been shown to positively impact dietary behaviors.51
Although potential health consequences of food environment changes continue to be studied to provide an evidence base for policy decisions, null findings in this longitudinal analysis add to a growing body of literature14,52,53 that casts doubt on short-term sufficiency of food retail presence in neighborhoods of older adults for curtailing incident events of clinical importance. Null results may have been due to a lack of nutrition behavior change in response to nearby food retail presence, or to a change that was insufficient in magnitude or duration to result in a detectable difference in clinical outcomes. Future work could be designed to distinguish these possibilities.
Findings for population density within this predominately urban sample (99% lived in a metropolitan area at baseline) add to an evolving understanding of whether populations experience an urban health advantage54 or urban health penalty.55 The observed association of elevated risk with greater walkable destination or population density complements a recent study showing cardiovascular risk factor elevation in urban versus rural environments among a large sample of middle aged and older adults.56
Noted strengths of the CHS cohort include well-characterized outcomes. In addition, time-varying analyses stratified on baseline food retail presence were planned to emulate a natural experiment, leveraging cleaned and coded establishment-level food environment data.
Limitations
Several limitations were noted. Despite 32,816 person-years of follow-up for incident CVD, confidence intervals for some null results allowed for a small but meaningful association. Statistical and graphical evidence suggest proportional hazards assumptions did not hold for all covariates (see Appendix Table 1 and Appendix Figure 1); hazard ratios mask variation over time in magnitude of covariate-outcome associations and should be interpreted with caution. Further, inference is limited due to reliance on the assumption of non-informative censoring, which may not hold for events such as death from other causes or residence in a nursing home. Reliance on national data to characterize supermarket/produce market or convenience/fast food retail presence across an extended period precluded detailed attention to characteristics that vary among establishments, such as item availability, signage, and pricing. The supermarket/produce market category included stores which offer a wide variety of foods with varying nutritional quality and may have missed some healthy food sources such as mobile produce vendors due to limitations in historically available establishment-level data. In this observational study, selection bias or unmeasured confounding may have masked true associations. Findings in this older adult cohort are not generalizable to other settings or age groups. Finally, food environment data focused on recent retail presence may not reflect cumulative influences across the life course,57 may differ from ratio-based food environment measures in socioeconomic and health associations,58 and data were not available to examine more proximal outcomes of interest such as household food item purchases, travel time for food shopping, or meals consumed out of home.
Conclusions
Among eligible members of a population-based elderly cohort, no statistically significant association was noted for neighborhood presence of supermarkets/produce markets or convenience/fast food retail with diabetes or CVD incidence. Neither hypothesized risk reduction resulting from time-varying presence of supermarkets/produce markets nor risk elevation for presence of convenient/fast food retail were supported. Alongside prior evidence including natural experiments, this suggests complementary policy strategies are needed to maximize health benefits of initiatives such as those to influence new food retail presence or closures.
Supplementary Material
Acknowledgements
The authors would like to thank Jeffrey Moore of Drexel University and the CHS internal reviewers for their guidance during proposal and analysis plan development, and James Quinn of Columbia University for his leading role in geographic characterization and data documentation.
Cardiovascular Health Study research was supported by contracts HHSN268201200036C, HHSN268200800007C, HHSN268201800001C, N01HC55222, N01HC85079, N01HC85080, N01HC85081, N01HC85082, N01HC85083, N01HC85086, 75N92021D00006, and grants U01HL080295 and U01HL130114 from the National Heart, Lung, and Blood Institute (NHLBI), with additional contribution from the National Institute of Neurological Disorders and Stroke (NINDS). Additional support was provided by R01AG023629 from the National Institute on Aging (NIA). A full list of principal CHS investigators and institutions can be found at CHS-NHLBI.org.
Geographic characterization including food retail measures were supported by the National Institute of Aging (grants 1R01AG049970, 3R01AG049970-04S1), Commonwealth Universal Research Enhancement (C.U.R.E) program funded by the Pennsylvania Department of Health - 2015 Formula award - SAP #4100072543, the Urban Health Collaborative at Drexel University, and the Built Environment and Health Research Group at Columbia University.
This work was supported by grants and contracts as specified in acknowledgements. Funders had no role in the study design, data collection and analysis, interpretation of findings, writing, or decision to publish.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
This project was approved by the Columbia University, Drexel University, and University of Washington Institutional Review Boards (protocols #IRB-AAAI8002, #1612004989R005, and #STUDY00000109).
The research presented in this paper is that of the authors and does not reflect the official policy of the National Institutes of Health or other funders.
Conflict of interest statement: The authors declare that there are no conflicts of interest.
Financial disclosures: No financial disclosures have been reported by the authors of this paper.
DATA ACCESS STATEMENT
RECVD/CHS linked data underlying the findings cannot be shared publicly due to the terms and conditions of licensed RECVD data. Linked data are available for researchers who meet the criteria to work with the licensed data. For more information regarding access to linked data, contact gslovasiresearch@gmail.com for more information. Researchers can request access to unlinked parent CHS data through the Cardiovascular Health Study, at https://chs-nhlbi.org/. All other questions contact Gina S Lovasi PhD, MPH at gsl45@drexel.edu.
References
- 1.Malambo P, Kengne AP, De Villiers A, Lambert EV, Puoane T. Built Environment, Selected Risk Factors and Major Cardiovascular Disease Outcomes: A Systematic Review. PLoS One. 2016;11(11):e0166846. 10.1371/journal.pone.0166846 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Murphy SL, Xu J, Kochanek KD, Arias E. Mortality in the United States, 2017. NCHS Data Brief. Nov 2018;(328):1–8. https://www.cdc.gov/nchs/data/databriefs/db328-h.pdf [PubMed] [Google Scholar]
- 3.Hosler AS, Rajulu DT, Fredrick BL, Ronsani AE. Assessing retail fruit and vegetable availability in urban and rural underserved communities. Prev Chronic Dis. Oct 2008;5(4):A123. http://www.cdc.gov/pcd/issues/2008/oct/07_0169.htm [PMC free article] [PubMed] [Google Scholar]
- 4.Zenk SN, Powell LM, Rimkus L, et al. Relative and absolute availability of healthier food and beverage alternatives across communities in the United States. Am J Public Health. Nov 2014;104(11):2170–8. 10.2105/AJPH.2014.302113 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Andreyeva T, Blumenthal DM, Schwartz MB, Long MW, Brownell KD. Availability and prices of foods across stores and neighborhoods: the case of New Haven, Connecticut. Health Aff (Millwood). Sep-Oct 2008;27(5):1381–8. 10.1377/hlthaff.27.5.1381 [DOI] [PubMed] [Google Scholar]
- 6.Leone AF, Rigby S, Betterley C, et al. Store type and demographic influence on the availability and price of healthful foods, Leon County, Florida, 2008. Preventing Chronic Disease. November 2011;8(6):A140. www.cdc.gov/pcd/issues/2011/nov/10_0231.htm. [PMC free article] [PubMed] [Google Scholar]
- 7.Jetter KM, Cassady DL. The availability and cost of healthier food alternatives. Am J Prev Med. Jan 2006;30(1):38–44. 10.1016/j.amepre.2005.08.039 [DOI] [PubMed] [Google Scholar]
- 8.Liese AD, Weis KE, Pluto D, Smith E, Lawson A. Food store types, availability, and cost of foods in a rural environment. J Am Diet Assoc. Nov 2007;107(11):1916–23. 10.1016/j.jada.2007.08.012 [DOI] [PubMed] [Google Scholar]
- 9.Li F, Harmer PA, Cardinal BJ, et al. Built environment, adiposity, and physical activity in adults aged 50–75. Am J Prev Med. July 2008;35(1):38–46. 10.1016/j.amepre.2008.03.021 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Sallis JF, Floyd MF, Rodriguez DA, Saelens BE. Role of built environments in physical activity, obesity, and cardiovascular disease. Circulation. Feb 7 2012;125(5):729–37. 10.1161/CIRCULATIONAHA.110.969022 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Ezzati M, Hoorn SV, Rodgers A, et al. Estimates of global and regional potential health gains from reducing multiple major risk factors. Lancet. Jul 26 2003;362(9380):271–80. 10.1016/S0140-6736(03)13968-2 [DOI] [PubMed] [Google Scholar]
- 12.Lovasi GS, Rundle A, Bader M, Neckerman KM. Case Study 1 Healthy and Unhealthy Food Sources in New York City. In: Fund MM, ed. Population Health: The Translation of Research to Policy. 2018. https://www.milbank.org/publications/population-health-the-translation-of-research-to-policy/ [Google Scholar]
- 13.Ghosh-Dastidar B, Cohen D, Hunter G, et al. Distance to store, food prices, and obesity in urban food deserts. Am J Prev Med. 2014;47(5):587–595. 10.1016/j.amepre.2014.07.005 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Allcott H, Diamond R, Dubé J-P, Handbury J, Rahkovsky I, Schnell M. Food deserts and the causes of nutritional inequality. The Quarterly Journal of Economics. 2019;134(4):1793–1844. 10.1093/qje/qjz015 [DOI] [Google Scholar]
- 15.Cummins S Large scale food retailing as an intervention for diet and health: quasi-experimental evaluation of a natural experiment. J Epidemiol Community Health. 2005-12-01 2005;59(12):1035–1040. 10.1136/jech.2004.029843 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Cummins S, Flint E, Matthews SA. New Neighborhood Grocery Store Increased Awareness Of Food Access But Did Not Alter Dietary Habits Or Obesity. Health Aff (Millwood). 2014-02-01 2014;33(2):283–291. 10.1377/hlthaff.2013.0512 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Dubowitz T, Ghosh-Dastidar M, Cohen DA, et al. Diet And Perceptions Change With Supermarket Introduction In A Food Desert, But Not Because Of Supermarket Use. Health Aff (Millwood). Nov 2015;34(11):1858–68. 10.1377/hlthaff.2015.0667 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Evans AE, Jennings R, Smiley AW, et al. Introduction of farm stands in low-income communities increases fruit and vegetable among community residents. Health Place. Sep 2012;18(5):1137–43. 10.1016/j.healthplace.2012.04.007 [DOI] [PubMed] [Google Scholar]
- 19.Ruelas V, Iverson E, Kiekel P, Peters A. The role of farmers’ markets in two low income, urban communities. J Community Health. Jun 2012;37(3):554–62. 10.1007/s10900-011-9479-y [DOI] [PubMed] [Google Scholar]
- 20.Leone LA, Tripicchio GL, Haynes-Maslow L, et al. Cluster randomized controlled trial of a mobile market intervention to increase fruit and vegetable intake among adults in lower-income communities in North Carolina. Int J Behav Nutr Phys Act Jan 5 2018;15(1):2. 10.1186/s12966-017-0637-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.An R, Maurer G. Consumption of sugar-sweetened beverages and discretionary foods among US adults by purchase location. Eur J Clin Nutr. 2016;70(12):1396–1400. 10.1038/ejcn.2016.136 [DOI] [PubMed] [Google Scholar]
- 22.Powell LM, Kumanyika SK, Isgor Z, Rimkus L, Zenk SN, Chaloupka FJ. Price promotions for food and beverage products in a nationwide sample of food stores. Prev Med. 2016;86:106–113. 10.1016/j.ypmed.2016.01.011 [DOI] [PubMed] [Google Scholar]
- 23.Ravensbergen EA, Waterlander WE, Kroeze W, Steenhuis IH. Healthy or unhealthy on sale? A cross-sectional study on the proportion of healthy and unhealthy foods promoted through flyer advertising by supermarkets in the Netherlands. BMC Public Health. 2015;15(1):1–10. 10.1186/s12889-015-1748-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Lovasi GS, Mooney SJ, Muennig P, DiMaggio C. Cause and context: place-based approaches to investigate how environments affect mental health. Soc Psychiatry Psychiatr Epidemiol. 2016;51(12):1571–1579. 10.1007/s00127-016-1300-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Gamba RJ, Schuchter J, Rutt C, Seto EY. Measuring the food environment and its effects on obesity in the United States: a systematic review of methods and results. J Community Health. 2015;40(3):464–475. 10.1007/s10900-014-9958-z [DOI] [PubMed] [Google Scholar]
- 26.Fried LP, Borhani NO, Enright P, et al. The Cardiovascular Health Study: design and rationale. Ann Epidemiol. Feb 1991;1(3):263–76. 10.1007/s10900-004-6091-4 [DOI] [PubMed] [Google Scholar]
- 27.Ives DG, Fitzpatrick AL, Bild DE, et al. Surveillance and ascertainment of cardiovascular events. The Cardiovascular Health Study. Ann Epidemiol. Jul 1995;5(4):278–85. 10.1016/1047-2797(94)00093-9 [DOI] [PubMed] [Google Scholar]
- 28.Jacquez GM, Slotnick MJ, Meliker JR, AvRuskin G, Copeland G, Nriagu J. Accuracy of commercially available residential histories for epidemiologic studies. Am J Epidemiol. Jan 2011;173(2):236–43. 10.1093/aje/kwq350 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Mossey JM, Shapiro E. Self-rated health: a predictor of mortality among the elderly. Am J Public Health. 1982;72(8):800–808. 10.2105/AJPH.72.8.800 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Ettinger WH Jr, Fried LP, Harris T, et al. Self-reported causes of physical disability in older people: The Cardiovascular Health Study. J Am Geriatr Soc. 1994;42(10):1035–1044. 10.1111/j.1532-5415.1994.tb06206.x [DOI] [PubMed] [Google Scholar]
- 31.Hoehner CM, Schootman M. Concordance of commercial data sources for neighborhood-effects studies. J Urban Health. Jul 2010;87(4):713–25. 10.1007/s11524-010-9458-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Hirsch JA, Winters M, Clarke P, McKay H. Generating GPS activity spaces that shed light upon the mobility habits of older adults: a descriptive analysis. Int J Health Geogr. 2014;13:51. 10.1186/1476-072X-13-51 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Hirsch JA, Moore KA, Cahill J, et al. Business Data Categorization and Refinement for Application in Longitudinal Neighborhood Health Research: a Methodology. J Urban Health. Oct 1 2020; 10.1007/s11524-020-00482-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Marwa WL, Radley D, Davis S, McKenna J, Griffiths C. Exploring factors affecting individual GPS-based activity space and how researcher-defined food environments represent activity space, exposure and use of food outlets. Int J Health Geogr. 2021;20(1):1–18. 10.1186/s12942-021-00287-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Moore LV, Diez Roux AV. Associations of neighborhood characteristics with the location and type of food stores. Am J Public Health. Feb 2006;96(2):325–31. 10.2105/AJPH.2004.058040 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Berger N, Kaufman TK, Bader MDM, et al. Disparities in trajectories of changes in the unhealthy food environment in New York City: A latent class growth analysis, 1990–2010. Soc Sci Med. 2019/08 2019;234:112362. 10.1016/j.socscimed.2019.112362 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Sims RV, Mujib M, Mcgwin G, et al. Heart Failure Is a Risk Factor for Incident Driving Cessation Among Community-Dwelling Older Adults: Findings From a Prospective Population Study. J Card Fail. 2011-12-01 2011;17(12):1035–1040. 10.1016/j.cardfail.2011.08.014 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Rundle AG, Chen Y, Quinn JW, et al. Development of a Neighborhood Walkability Index for Studying Neighborhood Physical Activity Contexts in Communities across the U.S. over the Past Three Decades. J Urban Health. Aug 2019;96(4):583–590. 10.1007/s11524-019-00370-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Logan JR, Stults BD, Xu Z. Validating Population Estimates for Harmonized Census Tract Data, 2000–2010. Ann Am Assoc Geogr. 2016;106(5):1013–1029. 10.1080/24694452.2016.1187060 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Kerr J, Frank L, Sallis JF, Saelens B, Glanz K, Chapman J. Predictors of trips to food destinations. Int J Behav Nutr Phys Act. 2012-01-01 2012;9(1):58. 10.1186/1479-5868-9-58 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Wedick NM, Ma Y, Olendzki BC, et al. Access to Healthy Food Stores Modifies Effect of a Dietary Intervention. Am J Prev Med. 2015-03-01 2015;48(3):309–317. 10.1016/j.amepre.2014.08.020 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Karpyn A, Mccallops K, Wolgast H, Glanz K. Improving Consumption and Purchases of Healthier Foods in Retail Environments: A Systematic Review. Int J Environ Res Public Health. 2020-10-16 2020;17(20):7524. 10.3390/ijerph17207524 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Gittelsohn J, Rowan M, Gadhoke P. Interventions in small food stores to change the food environment, improve diet, and reduce risk of chronic disease. Preventing Chronic Disease. 2012;9:E59. 10.5888/pcd9.110015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Shangguan S, Afshin A, Shulkin M, et al. A Meta-Analysis of Food Labeling Effects on Consumer Diet Behaviors and Industry Practices. Am J Prev Med. 2019-02-01 2019;56(2):300–314. 10.1016/j.amepre.2018.09.024 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Rummo PE, Guilkey DK, Ng SW, Popkin BM, Evenson KR, Gordon-Larsen P. Beyond supermarkets: Food outlet location selection in four US cities over time. Am J Prev Med. 2017;52(3):300–310. 10.1016/j.amepre.2016.08.042 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Warren JL, Gordon-Larsen P. Factors associated with supermarket and convenience store closure: a discrete time spatial survival modelling approach. J R Stat Soc Ser A Stat Soc. 2018;181(3):783–802. 10.1111/rssa.12330 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Strachan SM, Brawley LR. Healthy-eater Identity and Self-efficacy Predict Healthy Eating Behavior: A Prospective View. J Health Psychol. 2009-07-01 2009;14(5):684–695. 10.1177/1359105309104915 [DOI] [PubMed] [Google Scholar]
- 48.Barrington WE, Beresford SAA, Mcgregor BA, White E. Perceived Stress and Eating Behaviors by Sex, Obesity Status, and Stress Vulnerability: Findings from the Vitamins and Lifestyle (VITAL) Study. J Acad Nutr Diet. 2014-11-01 2014;114(11):1791–1799. 10.1016/j.jand.2014.03.015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Fatin Izzaty M, Noraida O, Zulfitri A, Zakaria N. Factors associated with food choices among elderly: a scoping review. Malays J Nutr. 2019;25(2):185–198. 10.31246/mjn-2018-0133 [DOI] [Google Scholar]
- 50.Stark JH, Neckerman K, Lovasi GS, et al. Neighbourhood food environments and body mass index among New York City adults. J Epidemiol Community Health. 2013-09-01 2013;67(9):736–742. 10.1136/jech-2013-202354 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51.Afshin A, Peñalvo JL, Del Gobbo L, et al. The prospective impact of food pricing on improving dietary consumption: A systematic review and meta-analysis. PLoS One. 2017-03-01 2017;12(3):e0172277. 10.1371/journal.pone.0172277 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52.Lovasi GS, Johnson NJ, Altekruse SF, et al. Healthy food retail availability and cardiovascular mortality in the United States: a cohort study. BMJ Open. 2021;11(7):e048390. 10.1136/bmjopen-2020-048390 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 53.Cobb LK, Appel LJ, Franco M, Jones-Smith JC, Nur A, Anderson CA. The relationship of the local food environment with obesity: a systematic review of methods, study quality, and results. Obesity. 2015;23(7):1331–1344. 10.1002/oby.21118 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54.Vlahov D, Galea S, Freudenberg N. The urban health “advantage”. J Urban Health. 2005;82(1):1–4. 10.1093/jurban/jti001 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 55.Freudenberg N, Galea S, Vlahov D. Beyond urban penalty and urban sprawl: back to living conditions as the focus of urban health. J Community Health. 2005;30(1):1–11. 10.1007/s10900-004-6091-4 [DOI] [PubMed] [Google Scholar]
- 56.Wang H, Stokes JE. Trajectories of rural-urban disparities in biological risks for cardiovascular disease among Chinese middle-aged and older adults. Health Place. 2020;64:102354. 10.1016/j.healthplace.2020.102354 [DOI] [PubMed] [Google Scholar]
- 57.Richardson AS, Meyer KA, Howard AG, et al. Neighborhood socioeconomic status and food environment: a 20-year longitudinal latent class analysis among CARDIA participants. Health Place. 2014;30:145–153. 10.1016/j.healthplace.2014.08.011 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 58.Stark JH, Neckerman K, Lovasi GS, et al. Neighbourhood food environments and body mass index among New York City adults. Journal of epidemiology and community health. 2013-09-01 2013;67(9):736–742. 10.1136/jech-2013-202354 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
RECVD/CHS linked data underlying the findings cannot be shared publicly due to the terms and conditions of licensed RECVD data. Linked data are available for researchers who meet the criteria to work with the licensed data. For more information regarding access to linked data, contact gslovasiresearch@gmail.com for more information. Researchers can request access to unlinked parent CHS data through the Cardiovascular Health Study, at https://chs-nhlbi.org/. All other questions contact Gina S Lovasi PhD, MPH at gsl45@drexel.edu.