Unadjusted and adjusted mixed models regression predicting change in effect size from study and effect size characteristics, k=285
| Models 1–14 | Model 15 | |
| Unadjusted b (95% CI) | Adjusted b (95% CI) | |
| Operationalisation of outcome | ||
| Smoking | Reference | Reference |
| Quitting (reverse coded) | 3.81 (−2.09 to 9.71) | −1.17 (−7.49 to 5.16) |
| Relapse | −11.41 (−21.81 to −1.01) | −10.49 (−20.51 to −0.48) |
| Disease | −0.85 (−8.36 to 6.65) | −3.14 (−10.17 to 3.89) |
| Young adult initiation | −8.08 (−16.56 to 0.40) | −14.32 (−22.25 to −6.39) |
| Other (eg, readiness) | −7.60 (−14.68 to −0.53) | −13.97 (−20.84 to −7.10) |
| Operationalisation of predictor variable | ||
| Proximity | Reference | Reference |
| Density: count | 0.52 (−4.60 to 5.65) | 1.35 (−3.42 to 6.11) |
| Density: per land area | −1.41 (−7.93 to 5.11) | 5.20 (−3.22 to 13.62) |
| Density: per people | −11.37 (−25.18 to 2.44) | −11.78 (−23.96 to 0.41) |
| Density: other | 5.44 (−1.69 to 12.58) | 9.06 (−0.71 to 18.83) |
| Study size | ||
| Study size (per 100 n) | −0.02 (−0.05 to 0.01) | −0.02 (−0.07 to 0.03) |
| Study design | ||
| No use of activity space | Reference | Reference |
| Activity space | 10.81 (1.20 to 20.42) | 7.34 (−1.31 to 16.00) |
| Cross-sectional | Reference | Reference |
| Longitudinal | 10.38 (5.29 to 15.47) | 11.09 (5.95 to 16.23) |
| Implementation of control variables | ||
| Unadjusted result | Reference | Reference |
| Adjusted for individual characteristics | 1.76 (−3.11 to 6.62) | 1.17 (−3.13 to 5.47) |
| Adjusted for neighbourhood characteristics | −5.93 (−26.83 to 14.97) | −4.11 (−22.67 to 14.46) |
| Adjusted for both | −0.39 (−6.13 to 5.35) | 1.38 (−4.33 to 7.08) |
| Operationalisation of predictor variable | ||
| No delinearisation of predictor | Reference | Reference |
| Delinearisation of predictor | 5.33 (0.34 to 10.32) | 11.31 (4.80 to 17.82) |
| Operationalisation of outcome variable | ||
| No ordinal or nominal outcome | Reference | Reference |
| Ordinal or nominal outcome | 7.82 (−0.64 to 16.28) | 3.63 (−7.14 to 14.41) |
| Study setting/Context | ||
| Policy environment | ||
| Point of sale advertising policy* | ||
| Without retail advertising restrictions | Reference | Reference |
| With retail advertising restrictions | 3.64 (−2.46 to 9.74) | −15.21 (−44.45 to 14.04) |
| Retail product display policy† | ||
| Without retail display restrictions | Reference | Reference |
| With retail display restrictions | 6.75 (0.87 to 12.63) | 2.42 (−19.00 to 23.84) |
| Country (weighted effect coded)‡ | ||
| Average country | Reference | Reference |
| Canada | 4.03 (0.47 to 7.58) | 6.80 (−13.47 to 27.07) |
| Finland | −1.29 (−8.06 to 5.48) | −6.16 (−25.42 to 13.09) |
| New Zealand | −6.14 (−18.35 to 6.08) | 1.74 (−20.79 to 24.28) |
| UK | −1.88 (−9.92 to 6.15) | −17.34 (−39.84 to 5.16) |
| USA | 0.67 (−4.73 to 6.06) | 2.11 (−14.38 to 18.60) |
| Participant sex | ||
| Results from analyses aggregating participant sex | Reference | Reference |
| Results specific to female participants | −2.39 (−12.18 to 7.41) | 4.01 (−9.78 to 17.80) |
| Results specific to male participants | 15.55 (5.00 to 26.11) | 21.82 (8.26 to 35.37) |
| Participant income | ||
| Results from analyses aggregating income levels | Reference | Reference |
| Results specific to lower-income participants | −1.74 (−15.70 to 12.21) | −3.92 (−16.59 to 8.75) |
| Results specific to higher-income participants | −5.53 (−18.49 to 7.43) | −6.05 (−18.37 to 6.28) |
| Participant tobacco use intensity | ||
| Aggregated levels of tobacco use intensity | Reference | Reference |
| Results specific to light tobacco use | −10.70 (−24.88 to 3.48) | −26.02 (−49.25 to −2.80) |
| Results specific to heavy tobacco use | −0.20 (−14.38 to 13.98) | −36.61 (−59.89 to −13.33) |
| Intercept | – | 4.93 (−11.37 to 21.24) |
A larger number indicates a larger estimated reduction in the behavioural outcome (eg, smoking) from reducing exposure to tobacco retailers as defined in the study. Models were constructed with a random intercept for study to address nesting of results within research studies. Bold indicates statistical significance at the historical p<0.05 threshold.
*Presence of an advertising policy includes results from Australia, Canada and New Zealand.
†Presence of a display policy includes results only from Canada.
‡Country results are weighted effect coded with Australia as the omitted category. Regression results from weighted effect coding can be compared against an average reference category instead of a single reference category.