Table 2.
Results - observational studies identified in Stage 1 that used more advanced analytical techniques specified in MRC guidance (n = 8)
| Study details | Description of variables | Results (for two different methods of analysis, when reported) | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Independent variables | Dependent variables | Main method of analysis: | Alternative method of analysis: | |||||||||
| More advanced analytical technique | Single equation analytical technique | |||||||||||
| First author, date, journal | Study population | Description | Time varying | Areal unit precision | Description | Source | Description of analytical technique | Data type (time periods) | Effect sizes (95% confidence interval) 1 | Method | Effect sizes (95% confidence interval) 1 | |
| Results where no statistically significant differences are observed between main and alternative analyses | Results where a mismatch between results is observed 2 | |||||||||||
| Cross sectional studies | ||||||||||||
| Anderson, 2011, American Economic Journal [30] | U.S. adults (11 States) | Miles between home and fast-food restaurant | N/A | Telephone/ZIP codes | BMI | BRFSS | Instrumental variable derived from distance to the interstate highway | Cross sectional (1) | 0.09 (−0.17, 0.17) | Not reported | ||
| Chen, 2012, Health Economics [31] | U.S. adults (Indianapolis, Indiana) | Number of | N/A | Individual addresses | BMI | Obesity Needs Assessment survey | Instrumental variable derived from distance to arterial roads and non-residential zones | Cross sectional (1) | OLS | None | Under-estimates: | |
| (a.) restaurants, | (a.) 0.37* (confidence interval missing) | (a.) 0.06 (−0.03, 0.14) | ||||||||||
| (b.) chain grocery stores, and | (b.) 0.90* (0.12, 1.682) | (b.) 0.14 (−0.21, 0.50) | ||||||||||
| (c.) proportion of park land, within a 0.5 mile radius | (c.) 2.85* (0.03, 5.67) | (c.) 2.39 (−0.66, 5.45) | ||||||||||
| Dunn, 2010, American Journal of Agricultural Economics [32] | U.S. adults (all States) | Number of fast food restaurants (at county level; author collected) | N/A | County level | BMI | BRFSS, 2004-2006 | Instrumental variable derived from number of interstate highway exits in the county | Cross sectional (1) | No statistically significant results were reported, except in two subgroup analyses: | OLS | No statistically significant results were reported, except in two subgroup analyses (see right). | Under-estimates were reported in two subgroup analyses: |
| Female participants in medium density counties: 0.06* (0.01, 0.11) | Female participants in medium density counties: −0.01 (−0.02, 0.01) | |||||||||||
| Non-white participants in medium density counties: 0.20* (0.02, 0.38) | Non-white participants in medium density counties: 0.01 (−0.02, 0.04) | |||||||||||
| Dunn, 2012, Economics and Human Biology [33] | U.S. adults (Brazos Valley, Texas) | N/A | Individual addresses | Obesity likelihood | A mail survey | Instrumental variable derived from distance to nearest highway | Cross sectional (1) | No statistically significant results were reported, except in two subgroup analyses: | Probit model | No statistically significant results were reported, except in two cases (see right). | Under-estimates in just two cases: | |
| e.g. Non-white participants: | Non-white participants: | Non-white participants: | ||||||||||
| (a.) miles to nearest fast-food restaurant, and number of fast-food restaurants within a | (a.) -0.100* (−0.178, −0.022) | (a.) -0.088 (−0.188, 0.012) | ||||||||||
| (b.) 1 mile and | (b.) 0.189* (0.030, 0.348) | (b.) 0.052 (−0.021, 0.125) | ||||||||||
| (c.) 3 mile radius | (c.) 0.058 (0.005, 0.121) | (c.) 0.014 (−0.004, 0.032) | ||||||||||
| Fish, 2010, Am J Public Health [34] | U.S. adults (Los Angeles County) | Resident perception of neighbourhood safety (self-reported dichotomous variable where 1= extremely or somewhat dangerous and 0=fairly or completely safe) | N/A | Individual level survey data | BMI | Los Angeles Family and Neighbourhood Survey | Instrumental variable derived from measures related to social cohesion and experience of household crime | Cross sectional (1) | 2.81* (0.11, 5.52) | OLS (using first wave 2001/2 data) | None | Under-estimate: -0.07 (−1.07, 0.93) |
| Zick, 2013, IJBNPA [35] | U.S. females (Salt Lake, Utah) | Neighbourhood walkability | N/A | Census block (typically 1,500 people) | BMI | Utah Population Database | Instrumental variable derived from neighbourhood characteristics e.g. churches and schools | Cross sectional (1) | −0.24* | OLS | None | Under-estimate: 0.00 |
| Longitudinal studies | ||||||||||||
| Courtemanche, 2011, Journal of Urban Economics [36] | U.S. adults (all States) | Number of Walmart Supercenters per 100,000 residents (these stores provide low cost food and encourage sedentary lifestyles) | Yes | County level | BRFSS, 1996-2005 | Instrumental variable derived from distance to Walmart head office (expansion over time of Walmart stores was shown to be correlated with distance from the head office) | Repeated cross sectional (10) | OLS | None | Under-estimates: | ||
| (i.) BMI | (i.) 0.24* (0.06, 0.41) | (i.) 0.02 (−0.00, 0.05) | ||||||||||
| (ii.) Obesity likelihood | (ii.) 0.023* (0.011, 0.035) | (ii.) 0.001 (−0.001, 0.003) | ||||||||||
| Zhao, 2010, Journal of Health Economics [3] | U.S. adults (all States) | Proportion of people living in densely populated areas with >9000 people per square mile | Yes (4; every 10 years) | MSA level (366 of these in U.S.) | (i.) BMI | National Health Interview Survey, 1976-2001 | Instrumental variable derived from exogenous expansion over time of the U.S. interstate highway system | Repeated cross sectional (25) | (i.) −0.01 (−0.03, 0.01) | Not reported | ||
| (ii.) Obesity likelihood | (ii.) −0.0013* (−0.002, 0.000)3 | |||||||||||
BMI: Body mass index measured in kg/m2 BRFSS: Behavioural Risk Factor Surveillance System dataset. MSA: Metropolitan Statistical Area.
OLS: Ordinary-Least-Squares.
1 * indicates statistical significance at the p < 0.05 level.
2 when compared to results in the main analysis: “Under-estimate” if statistically significant results in the main analysis were not statistically significant the cross-sectional, single equation analysis; “Over-estimate” if statistically insignificant results in the main analysis were statistically significant in the cross-sectional, single equation analysis.
3 The interpretation of this result is that for each additional percentage point decrease in the proportion of population living in the densely populated area, obesity is approximately 0.1–0.2 percentage points higher.