National Cancer Institute (NCI) Adjustment Method for Nutrition Data
The ABS has recently commenced collecting detailed food consumption data from the Australian population as part of the new expanded Australian Health Survey (AHS), something not attempted since the National Nutrition Survey (NNS) ran in 1995. Randomly selected participants initially undergo a face to face interview referred to as a "24 hour recall" where they are assisted in comprehensively detailing their consumption of solids and liquids over the previous day.
A key issue is how these 24 hour recalls should be used to draw conclusions regarding the population's long-term consumption of foods and nutrients, which is the primary concern of government policy makers in food and nutrition standards.
A general solution, and the one employed by the 1995 NNS, is to collect a second 24 hour recall from participants at a later date so that the component of error due to day to day variation (within-person measurement error) can be modelled across the population and removed to leave the long-term intake distribution as desired.
Methods following this principle are relatively straight-forward for nutrient consumption where data is always non-zero and often well approximated by a normal distribution. However, modelling long-term consumption of certain foods encounters formidable challenges such as high numbers of zero entries (many foods are not consumed on a particular day) and right-skewed consumption patterns (many foods have a smaller number of high-consumers).
Given the lack of developed methods at the time of the 1995 NNS and the small number of second day 24 hour recalls to work with (these were only collected for 10% of the sample), the ABS only published long-term intake distributions for selected nutrients, not foods. Hence quality long-term food intake distributions have long been a significant information gap filled only by food-frequency data with its intractable recall and concept bias issues.
Since this time, methods to overcome such challenges have built on each other to culminate in a comprehensive solution now referred to as the "NCI method" in recognition of the National Cancer Institute in the United States, where the most recent innovation to the method was developed.
Use of the NCI method for the Nutrition component of AHS 2011-12 was proposed by the Household Survey Methodology (HSM) section and justified shedding of the food frequency component of the survey freeing up valuable interview time. HSM has since conducted a trial of the method on the "Kids eat, Kids play" nutrition and physical activity dataset for children collected by CSIRO in 2007 to further understand the method and anticipate issues surrounding its application in practice within the ABS environment in particular.
For further details please contact Ross Watmuff at firstname.lastname@example.org or on (02) 6252 7084