2. Usual intake more important than one day snapshot
Methods of measuring dietary intake include
◦ Diet History
◦ 24hr recall
◦ Food frequency questionnaire
◦ Food diaries
◦ Duplicate meals
Used to investigate quality of diet (core-foods) and
energy intake
3. Present/prospective
◦ Diaries
Household measures
package weights
numbers of items
size of items
Weights
◦ Chemical analysis
Duplicate meals
Past/retrospective
◦ Actual
24-hour recall
Food Frequency
Questionnaires
◦ Usual
Dietary Histories
4. Prospective (current)
Retrospective(past)
• Actual
• 24-hour recall
• Food Frequency Questionnaires
• Usual
• Dietary Histories
•Food diaries
• Household measures or weighed
•Chemical analysis
•Duplicate meals
5. Current diet: Heisenberg uncertainty principle:
as soon as you stop something to measure it you
change the behaviour
Past diet: relies on memory, conceptual and
cognitive ability and honesty on part of
respondents
No direct measure of what people eat will
provide a true picture of their dietary habits
Task is to understand and define the errors,
try to minimise, manage statistically and
incorporate in interpretation of results.
5
6. Is information needed about foods, nutrients, other
food constituents, dietary behaviours (eg eating
meals)?
Is the average intake of a group required?
eg. do Australian Vietnamese babies have lower iron
intakes than Caucasian Australian babies
Is the average intake of each individual needed ?
eg is iron intake of Australian Vietnamese babies
related to the prevalence of anaemia in this group.
6
7. Using household measures
◦ Set time period for recording (minimum 3 days)
◦ Record all food consumed & drink over there days in
diary
◦ Include a non-working day
Using weights
◦ Similar recording to above but use a set procedure
for weighing using scales & measuring utensils
8. Requires the participant
to record food and
drink eaten and
estimate the amount
9. supplement use recorded
number of days varies, traditionally 7, 3 or 4
consecutive days, including a weekend day for 3/4 day
records
number of days required can be determined statistically
and depends on the nutrient of interest, objective of
the study etc
9
10. Potential to provide quantitatively accurate
information on food consumed during
recording period
‘Gold-standard’
Less likely for foods to be omitted
Accuracy of portion sizes
Information RE meal patterns and specific
foods
10
11. Motivation and high levels of literacy
required
High respondent burden
Change in intake (result of monitoring)
Bias in representativeness of respondents
11
12. Gold standard’ status now questioned
(systematic under reporting)
major issue is how representative the
recording period is of ‘usual’ diet, apart from
the issues of reporting errors. ie natural
variation in diet.
high coding burden, especially in large studies
12
13. Observed or plate waste method
◦ respondent researcher burden
◦ under reporting
Estimated portions
◦ hh measures, plate pictures, food models, photos –
mobile phones
13
14. Collection of an identical
meal/food from client for
analysis
◦ useful for analysing food
composition
Limitations
◦ expensive & time consuming
◦ impractical outside confined
setting
15. Ask individual to remember what food &
drinks they have consumed in previous 24
hours or day
◦ usually performed by trained interviewer
Let’s try a 24-hour recall
1. Write down on a piece of paper what you have eaten in
the last 24 hours
2. Do you think there is sufficient detail for you to
analyse it in detail?
3. Do you think you have been completely honest?
4. Could you remember exactly?
5. Could you estimate quantities accurately?
6. Do you think this record is typical for you?
16. Low respondent burden
Quick, inexpensive
Low levels of literacy required
Potential for greater representativeness
Does not change dietary habits as
retrospective
Coding easier
Provide some information on meal patterns/
intake of specific foods
16
17. relies on memory
portion estimation
relies on interviewer skill, potential for observer bias
can only provide group means, assess proportion of
population with adequate/inadequate intakes
does not provide data on usual intake of individual, no day to
day variation – repeat recall 10%
multiple 24h recalls have been used to provide usual intake
of individuals.
Criteria used to decide no of days of records apply
non-consecutive days, no subject notice
17
18. Qualitative & semi-quantitative
survey of:
◦ how often foods are eaten +/- how much
◦ usually self administered, can be interview
◦ general or specific FFQ
Developed to be used for large groups
Information about usual food consumption patterns
19.
20. 3 sections:
(i) food list
(ii) frequency of use categories
(iii) estimates of quantities usually consumed
May be administered by self (post),
interviewer and now often optically
scannable
20
21. Estimates usual intake of foods
Rank individuals according to their usual
intake of foods or groups of foods (and
nutrient intake if portion sizes included)
Widely used
Generally not adequate to estimate absolute
intake
Can behave unpredictably in subpopulations,
must be validated for study population
21
22. Easy and cheap to self administer
Avoids observer bias
Suitable for geographically widespread
sample as can be done by telephone or post
Small burden for respondent
22
23. Cannot use FFQ designed in US here, even UK
to here difficult, adult FFQ cannot be used
for children or adolescents
No information on food intake and meal
patterns;? use supplementary questions
23
24. Amount of work required to develop and validate
Quantification of intake not as accurate as
recalls/records.
Errors
incomplete listing of foods
errors in frequency estimation
errors in portion size estimation
Memory and period of recall required has major
issues eg 12 months, three months etc.
24
25. Seasonal variation difficult to capture eg
mangoes
Tend to overestimate, particularly fruit and
vegetable intake
Cognitive difficulties with reporting foods eaten
alone and in combination, risk of double counting
25
26. Foods grouped (eg beefburgers, steak, steak
pie) - assumptions made on contribution each
food makes to the overall reported
frequency and portion size
FFQ designed for one ethnic group are
unlikely to be suitable to others due to
major differences in food types, recipes,
food combinations & frequency
26
27. Requires nutritionist to
interview participant
and to record detail and
estimate the quantities
of usual/ typical intake
Original adaptation - 3 components
1. a usual 24-hour recall
usual takes away atypical errors
food amounts enhanced by using food models
2. a checklist
3. a short (3 days) diary
28. Original methodology more commonly now adapted -
Multiple variations
An investigation of habitual intake over a
specified period of time – usually 1 mo to 1 yr
24 hour recall + variations
Details on types (e.g. low fat), quantities,
brands)
Indicates patterns of eating
Able to estimate energy, macro & micro
nutrients
Able to estimate food groups
Can be targeted to particular nutrients e.g.
carbohydrate in diabetes
Administered by a trained health professional
29. a usual 24-hour
recall
◦ would you normally
consume breakfast?
◦ what would you
normally have?
◦ what type of bread
would that usually
be?
◦ etc
32. honesty, memory & frankness of client
need highly skilled interviewer
expensive & time consuming
may underestimate intake
Most useful information for
assessment
33. Depends on memory
Bias towards those with regular meal intake
and pattern
Exaggeration of meal regularity
Underestimation of snacks
No indication of day to day variability
Requires highly trained and skilled staff -
high risk of interviewer bias
More likely to recall immediate past than
period of interest
33
34. Single interview gives detailed information on meal patterns,
food consumption, usual nutrient intake
Information on usual intake over long periods can
theoretically be obtained
Particular attention can be given to nutrients of interest and
goods/ drinks that are often under reported eg alcohol,
confectionary
Detailed food preparation can be obtained
Seasonal variation can be assessed
34
35. duplicate portion methods
PETRA (Portable Electronic Tape Recorded
Automatic) Scales - linked scales & tape recorder
photographic methods - atlas of photographed
food potions
video recording of meals – use of mobile phones
computer assisted automated diet histories/ 24
hour recalls
35
40. Respondent bias
inaccurately report income and age
over report consumption of ‘good foods’
under report ‘bad’ foods - alcohol, snack
foods, confectionary etc
overweight respondents may under report
total food intake
40
41. Respondent memory lapses
Incorrect estimation of portion size
◦ concept of ‘average’ serving size differs
◦ weighed vs estimated does not seem to introduce
bias
Daily variation
in one respect not an error but does introduce
misclassification
41
42. Supplement usage
fail to record gives underestimation of nutrient
intake
Frequency food consumption
FFQ tend to overestimate intake, especially fruit
and veges
No. of food items can have an effect, shorter list
tends to increase bias associated with estimation
of single food items
42
43. Certain groups, eg overweight, systematically under
report their habitual food intake; diabetics tend to
report prescribed diet
Recently identified problem that has called into
question the ‘gold standard status’ of the 7 day
weighed food record.
Must demonstrate if present/not
Must identify under-reporters for analysis
? Exclude, depends on aims
43
44. Day of the week affect
significant week/weekend day differences for
some foods and nutrients
likely to be culturally and economically
significant and change over time
need to proportionately represent weekend
days in design
44
45. Seasonal effects
can be very specific especially in developing
countries and poor communities
variation tends to be greater for foods than
nutrients
some micronutrients significantly effected by
fruit and vegetable consumption
survey across seasons if possible
45
46. Essential to help understand the relationship between
what we have measured and the ‘truth’
Describe and quantify error as far as possible to enable
consideration in
analysis
interpretation of results
To ensure the tool will allow us to detect the diet
disease relationship should it exist
Inadequate attention to validation results in wasted
resources and failure to detect true relationships
between diet and disease`
46
47. Reliability
reproducibility; repeatability
if we did it again would we get the same answer
measure reliable but not valid; if valid must be
reliable
Validity: are our findings true?
requires a true external reference measure or
standard
47
48. Not possible to have absolute direct measure
of dietary intake, ie errors in all methods
aim to establish ‘relative validity’
compare test method with reference method
does not necessarily indicate method valid, may
indicate both methods have the same errors
Relative validity major issue in nutritional epi
48
49. validity and reproducibility of reference method
should be high
7 day record has been the classical reference
method - not error free
errors should be independent
eg recall method validated by prospective method
eg FFQ against Diet history -can reflect same
exposure period but both rely on portion size and
frequency estimations and recall skills
49
Notes de l'éditeur
Measuring food intake is one of the most difficult tasks undertaken by health professionals because:Food supply is extremely variedIndividuals eat widely varying amounts and combinations from day to dayVariations exist in recipes and nutrient content of different versions of the same foodTask of recording foods in detail is time consuming and requires a lot of co-operation from respondents
potential to provide quantitatively accurate information on food consumed during recording period. historically viewed as the ‘gold standard’.omission of foods eaten less likely than recall methodsmore accurate information on portion sizes as most not estimatedgives information re meal patterns and specific type so foods eaten
respondents need to be motivated, numerate and literate respondent burden high, especially for 7 day record? change intake to simplify the weighing procedure or to impress investigatorsbias in representativeness of respondents who successfully undertake the rigours of the 7 day record
Gold standard’ status now questioned; serious systematic under reporting has been documented.
respondent burden small and compliance generally highmethod quick, relatively inexpensive, useful for subjects with low literacy? greater representativeness as burden is small and suitable for wider cross section of populationdoes not change dietary habits as consumption has already occurredcoding easier due to standardisation and certain amount of precoding can be done, ? direct entry into computersome information on meal patterns and intake of specific foods eg fruit and vegetables
Widely used in large scale epidemiological studiesGenerally not adequate to estimate absolute intake, better suited to comparisons and ranking
Assessment of usual intake over a specific time period not widely used although some increased interest latelyuseful for assessing energy and macronutrient intake of groups and to rank individualsmainly clinical use
dependant on memory and ability to give historybias towards those with regular meal intake and patternmay exaggerate the regularity of meals and underestimate intake of snack foodsno indication of day to day variability in intakerequires highly trained and skilled staff - high risk of interviewer biasmore likely to recall immediate past than period of interest
Variation in diet - natural phenomenon not error, but must be accounted for in design and choice of methodsinter-subject: between persons intra-subject: within subjects over time ie from day to day, season to season, year to year. degree of variation depends on the nutrient. Nutrients found in wide range of foods have lower variation than nutrients concentrated in a few specific foods eg carotene/Vitamin A.some nutrients within-person >>between person - difficult to detect difference between groups/individuals
Validity: For the study subjects; internal validityfor the relevant others outside the study - external validity