Table 2.
Summary of results across the five parameters used to assess construct/convergent validity (n 15 227*)
| NP model | ||||||
|---|---|---|---|---|---|---|
| Parameters | Statistical test or analysis | FSANZ | Nutri-Score | HCST | EURO | PAHO |
| Association between the proportion of ‘less healthy’ foods classified by model and quartiles of Ofcom scores | Cochran–Armitage trend test | Positive association (P<0·001) | Positive association (P<0·001) | Positive association (P<0·001) | Positive association (P<0·001) | Positive association (P<0·001) |
| Agreement with Ofcom | κ statistic† | Near perfect (κ=0·89) | Near perfect (κ=0·83) | Fair (κ=0·26) | Moderate (κ=0·54) | Fair (κ=0·28) |
| Discordance with Ofcom (% of foods) | McNemar’s test | 5·3 (P<0·001) | 8·3 (P<0·001) | 37·0 (P<0·001) | 22·0 (P<0·001) | 33·4 (P<0·001) |
| Linear association between FSANZ and Nutri-Score scores v. Ofcom scores | Pearson’s correlation coefficient | r=0·973 (P<0·001) | r=0·926 (P<0·001) | N/A | N/A | N/A |
| Cross-classification analyses: agreement and misclassification between quartiles of FSANZ scores and 4 HCST tiers v. quartiles of Ofcom scores, or between 5 Nutri-Score classes v. quintiles of Ofcom scores (% of foods*) | Exact agreement | 95·2 | 76·1 | 32·6 | N/A | N/A |
| Agreement ±1 quartile/tier or class/quintile | 4·2 | 16·1 | 48·8 | |||
| Disagreement | 0·3 | 7·1 | 15·9 | |||
| Gross misclassification | 0 | 0·3 | 2·3 | |||
| Overall convergence with Ofcom | Convergent | Convergent | Not convergent | Moderately convergent | Not convergent | |
NP, nutrient profiling; FSANZ, Food Standards Australia New Zealand; HCST, Health Canada Surveillance Tool; EURO, WHO Regional Office for Europe; PAHO, WHO Regional Office for the Americas/Pan American Health Organization; N/A, not applicable.
*
Across the different comparisons, data were missing for the following proportion of foods: FSANZ and Nutri-Score v. Ofcom, 0·29 % (n 44); HCST v. Ofcom, 0·41 % (n 62); EURO and PAHO v. Ofcom, 0·30 % (n 45).
†
Agreement was assessed using the κ statistic as follows: 0·01–0·20 ‘slight’; 0·21–0·40 ‘fair’; 0·41–0·60 ‘moderate’; 0·61–0·80 ‘substantial’; 0·81–0·99 ‘near perfect’( 39 ).