264 Introduction to Human Nutrition
have not consumed. For this reason, 24 hour recall
studies often provide estimates of food intake that are
lower than food records obtained over the same
period. The size of the error incurred by the omission
of one or more food items clearly depends on what is
omitted and not only on the proportion of food
omitted. For example, the effect, on 24 hour energy
intake, of omitting a cup of black coffee, a glass of
milk, or a bar of chocolate is quite different.
The omission of food items in 24 hour recall studies
can be reduced by appropriate probing by the inter-
viewer in relation to meals, between-meal snacks, and
other activities on the previous day, but even when
respondents have previously weighed their food the
average energy intake may still be underestimated by
as much as 20%.
Long-term memory
The diet history and most food frequency ques-
tionnaires set out to measure the habitual intake of
an individual over a period of weeks or months.
Individuals are not asked to recall their food intake on
specifi c days, but to construct a picture of their “usual”
food consumption pattern over a specifi ed reference
period. To provide reliable information individuals
thus need to be able to remember the range of foods
that they usually consume, to judge the frequency of
consumption on a long-term basis, and to be able to
estimate correctly the average amount that is usually
consumed. These are complex cognitive tasks.
As in the case of 24 hour recalls, no attempt is
usually made to assess how well individuals are able
to perform these various tasks. From the limited
amount of data available from comparative studies
between diet histories and long-term diet records, it
appears that the two methods do not give concordant
results in individuals. Food frequency questionnaires
are subject to the same diffi culties, and have the added
problem that estimates of portion size are based on
standard measures or, in the case of mailed question-
naires, are made in the absence of visual aids such as
food models or photographs.
When respondents are asked to report their intake
over a period of weeks they rely largely on generic
knowledge of their diet and tend to report items that
they are likely to have eaten or items that they rou-
tinely eat, rather than items that they specifi cally
remember having eaten during the reference period.
This tendency increases with the time interval between
the recall and the reference period. The accuracy of
frequency estimates also deteriorates with time. While
individuals appear to report more frequently eaten
foods with greater frequency than less frequently eaten
foods, there are differences between individuals in the
way that they report the same frequency of consump-
tion. Ranking of individuals on the basis of the usual
frequency of intake is thus likely to lead to misclassifi -
cation unless the extent of the differences between
individuals is known and can be taken into account. It
is diffi cult to see how such misclassifi cation can be
reduced unless it is possible to classify individuals, in
some way, in terms of their ability to provide reliable
information on habitual long-term intake.Day-to-day variation in intake
We have already seen that individuals vary consider-
ably in their intake of nutrients from day to day (see
Figure 10.2). In addition, the extent of day-to-day
variation differs between nutrients. The implication
of the fi rst observation is that short-term intake data
(e.g., 24 hour recall data) are unlikely to provide a
reliable estimate of habitual intake for most individu-
als. The implication of the second observation is that
the length of time for which dietary data need to be
collected, in order to estimate habitual intake with
any given level of confi dence, varies with the nutrient
of interest.
Table 10.4 expresses the impact of this variation in
terms of the number of days of dietary information
needed to classify 80% of individuals into the correct
third of the distribution. It is clear from this table that
not only 24 hour recalls but also 7 day records are
likely to be inadequate to classify 80% of individuals
correctly into the appropriate third of the distribution
for most micronutrients. This is an important reason,
although not the only reason, why short-term records
are only rarely used for epidemiological studies, in
preference to food frequency questionnaires, despite
the loss of detail and precision inevitably associated
with the use of the latter.Effect on usual diet
Recall methods clearly cannot change what has already
been eaten, but what has been eaten can be misre-
ported either consciously or unconsciously. When
individuals are asked to keep records, however, they
may also alter their normal habits as a consequence
of the recording process. One obvious reason for