Abstract
The "Dairy Diary" is a user-friendly web-based dairy intake screener. The reliability and validity are unknown. We aimed to evaluate the screener in terms of test-retest reliability and comparative validity. In a diagnostic accuracy study, a purposefully recruited sample of 79 (age: 21.6 ± 3.8 years) undergraduate dietetics/nutrition students from three South African universities completed 3 non-consecutive days of weighed food records (reference standard) within a seven-day period (comparative validity), followed by two administrations, 2 weeks apart, of the screener (index test) (reliability). For the four dairy product serving scores (PSSs) and the summative dairy serving scores (DSSs) of the screener and the food records, t-tests, correlations, Bland-Altman, Kappa, McNemar's, and diagnostic accuracy were determined. For reliability, mean PSSs and DSSs did not differ significantly (p > .05) between the screener administrations. The mean PSSs were strongly correlated: milk (r = .69; p < .001), maas (fermented milk) (r = .72; p < .001), yoghurt (r = .71; p < .001), cheese (r = .74; p < .001). For DSSs, Kappa was moderate (k = 0.45; p < .001). Non-agreeing responses suggest symmetry (p = .334). For validity, the PSSs of the screener and food records were moderately correlated [milk (r = .30; p = .0129), yoghurt (r = .38; p < .001), cheese (r = .38; p < .001)], with k = 0.31 (p = .006) for DSS. Bland-Altman analyses showed acceptable agreement for DSSs (bias: -0.49; 95% CI: -0.7 to -0.3). Categorized DSSs had high sensitivity (81.4%) and positive predictive value (93.4%), yet low specificity (55.6%) and negative predictive value (27.8%). The area under the receiver operating characteristic curve (0.7) was acceptable. The "Dairy Diary" is test-retest reliable with moderate comparative validity to screen for dairy intake of nutrition-literate consumers.