Table 3.
Summary of findings for all 132 main models investigating the impact of outlet definition.
| Statistically significant difference (Q1 vs Q4) |
Substantive Conclusions |
|||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| No. of associations |
% agreementa |
No. of associations |
% agreementa |
|||||||||||
| + | 0 | – | M | N | Lrg+ | Sml+ | U+ | Null | U- | Sml- | Lrg- | M | N | |
| Fast Foodb | ||||||||||||||
| B | 6 | 10 | 0 | 62.5 | 43.8 | 3 | 5 | 2 | 6 | 0 | 0 | 0 | 68.8 | 25.0 |
| M | 8 | 8 | 0 | – | 31.3 | 2 | 6 | 2 | 6 | 0 | 0 | 0 | – | 25.0 |
| N | 0 | 9 | 7 | – | – | 0 | 1 | 0 | 8 | 0 | 5 | 2 | – | – |
| Convenienceb | ||||||||||||||
| B | 0 | 16 | 0 | 87.5 | 87.5 | 0 | 0 | 2 | 14 | 0 | 0 | 0 | 56.3 | 62.5 |
| M | 1 | 14 | 1 | – | 87.5 | 0 | 1 | 2 | 11 | 0 | 2 | 0 | – | 62.5 |
| N | 0 | 14 | 2 | – | – | 0 | 0 | 4 | 10 | 0 | 2 | 0 | – | – |
| Supermarketsc | ||||||||||||||
| B | 1 | 11 | 0 | 91.7 | 83.3 | 0 | 1 | 3 | 8 | 0 | 0 | 0 | 50.0 | 50.0 |
| M | 0 | 12 | 0 | – | 91.7 | 0 | 0 | 2 | 10 | 0 | 0 | 0 | – | 75.0 |
| N | 0 | 11 | 1 | – | – | 0 | 0 | 0 | 10 | 0 | 2 | 0 | – | – |
Key: +/−: statistically significant positive/negative difference between quartile 1 (Q1) and quartile 4 (Q4). 0: no statistically significant difference between Q1 and Q4. B: broad. M: moderate. N: narrow. Lrg+/−: large positive/negative association. Sml+/−: ‘small positive/negative association’ U+/−: ‘positive/negative U-shaped association.
Percentage agreement between findings from models differing in the definition scope, but being otherwise identical.
Results from 16 models respectively modelling exposures of outlet counts within 3,200m, 1,600m, 800m Euclidian and network buffers and presence/absence of outlets within 400m Euclidian and network buffers (8 exposures), against the respective outcomes of BMI and obesity (2 outcomes).
Results from 12 models respectively modelling exposures of Supermarket counts within 3,200m & 1,600m Euclidian and network buffers and presence/absence of supermarkets within 800m Euclidian and network buffers (6 exposures), against the respective outcomes of BMI and obesity (2 outcomes).