Bringing dietary data closer to the truth with statistical adjustment: The 2018 Provincial Dietary Intake Survey as an example
A single 24-hour recall is often used as the primary instrument for measuring dietary intake in large dietary studies. A common concern with a single 24-hour recall is the day-to-day variation in the diet of free-living populations. The magnitude of the mostly random within-person variance varies by nutrients and is largely dependent on cultural and ecological factors. These errors result in large standard deviations in populations. A result of exaggerated variation is that the percentage of respondents below or above specified cut points will be distorted. A method to distinguish within-person from between-person variation, account for extreme intakes and allow for adjustment for covariates was applied.
Results indicated that the overall unadjusted and adjusted means of the nutrient intakes were mostly similar. However, large differences between unadjusted and the adjusted values of the percentiles of the intake of some nutrients were observed. The method compressed the distribution of nutrient intake towards the mean. This compression was most apparent when examining the percentage of the population below or above AMDR/DRI limits, especially for nutrients with skew distributions. In these cases, the adjusted percentage below or above the cut points was substantially smaller, demonstrating that the prevalence of dietary insufficiency or excess could be overestimated if intakes are not adjusted. Importantly, the method allows for estimation of population exposure without conducting repeated 24-hour recalls on the total sample. This study shows that a small subsample (11%) may be sufficient to conduct the described adjustments.