NUTR 400 Part 2

• -whether you want information about dietary
• behaviours or amounts of foods, nutrients, and nonnutrient
• food constituents consumed
• - whether you want to assess absolute or relative
• amounts of foods and/or nutrients
• - whether you want historical or current patterns
• - time, resources, money etc that you hav4e

• Food Frequency Questionaire
• 24Hour Recall
• Food Record
• Direct Observation

• - 24hr recall, food records, food diaries
• - Oftern used in population surverys for monitoring health and nutrition

• - FFQ
• -Often used in large cohort or case control studies to examine die health relationships

• - Trained or untrained respondents
• - Detailed review or not
• - Highly standardized coding rules or not
• - Development of electronic methods: PDA, cell
• phones

• - Intake is quantified, detailed
• - Can include weekends
• - Could be relatively accurate (detail/training)
• - Data are rich in details (Nutrient, Cooking practices, Meal and eating frequency)

• -Recording influences diet (reactive tool)- Biased measurement
• - Requires literacy
• - High respondent and investigator burden
• - Multiple days required to estimate usual intake
• - Sample selection bias
• - Completion worse over time
• - Underreporting is typical- Worse with overweight/obese

• -Ascertain descriptions and amounts of foods consumed
• in 24 hour period prior to interview
• - Can be interviewer administered or self-administered by computer (US National Cancer Institute is developing a
• web-based tool;

• -Training of interviewers
• - Standardization of probing questions
• - Multiple passes through the day
• - Computer vs. paper/pencil administration
• - In-person vs. telephone
• - Portion size models or measurement aids
• - Quality and completeness of nutrient database
• -Development of self-administered automated
• -5 systems administered via computer or internet

• -low respondent burden and high compliance- less
• chance of selection bias due to burden
• - can be suitable for illiterate subjects
• - standardised protocol can be used
• - unlikely to interfere with dietary behaviour
• - can get through relatively large numbers of subjects in
• one day
• ! can get details on cooking practises, brands, meal
• frequency etc.

• -Imperfect knowledge and memory- relies on this
• - Relies heavily on getting a portion size estimation
• - Multiple days required to estimate usual intake- single day is a snap-shot
• - Costly to administer if face-to face- Highly trained interviewers
• - Some evidence that reporting declines with multiple administrations
• - Underreporting is typical- Worse with overweight/obese

• - relies heavily on memory
• - portion size estimation issues
• - a single 24 hour recall only gives a snapshot of that
• persons intake
• (simply having a bad day cannot be used to make judgements that person likely nutrition status

• - Number of foods, clarity of questions
• - Portion size questions: pictures vs. text description
• - Time frame
• - Development of food list, nutrient database
• - Type of administration: computer vs. paper
• - Specificity to population of interest
• - Supplement intake
• - Food preparation

• -Can assess intakes of very large numbers of people
• (can be distributed/collected by mail and self
• administered)
• -Atempts to estimate 'usual' individual intake of foods with one administration
• - Does not affect eating behavior-
• -Lower cost of administering/processing-optic-readable scanned forms and Computer-generated data

• - Lack details ( finite food list, details of cookin methods, specific food type lost
• -Cognitively complex
• - Requires literacy
• - Affected by recent diet
• - Severe measurement error: prone to bias
• - Different FFQs can behave quite differently-FFQ designed to
• assess calcium intake is NOT appropriate if you want to
• measure intakes of fruit and vegetables
• - Different populations respond differently
• -Problems with assessing foods

• 1. What was the original purpose of the
• questionnaire?
• 2. Who was the target population? Are
• questionnaires developed in Europe suitable for
• use in Canada?
• 3. When was the questionnaire developed? Are the
• foods and portion sizes still relevant?
• 4. Has a previous validation been carried out, was it acceptable

• Databases ( USDA, CANDI)
• Questions:
• Foods away from home?
• Foods prepared at home?
• Fortified/enriched foods?
• How good is it at accurately determining nutrients?
• Updating of the systerm with changes to food supply?
• New food compound of health interest?
• Supplement use?

• - What database was used?
• - Are valid estimates possible - intake data collected to obtain a reliable and valid estimate of food in the population of interest., completion of the database
• -Is the correction level of precision reported

• - Used to assess dietary intake (food records, FFQ)
• for case-control, survey & cohort studies
• - Cluster, Factor or Index analysis
• - Cluster & Factor determine patterns from collected
• food intake
• - Indexes are pre-determined based on definition of
• healthy eating
• - Healthy Eating Index, Dietary Quality Index,
• Alternate Healthy Eating Index
• -limitations in using these?

• -Very few short term available, few long term
• EXAMPLES: see readings
• - Concentration: reflect the biological response to intake - Cannot be used to assess amount consumed, Correlate to amount consumed, May reflect short or long term intakes, Some are appropriate for risk assessment
• -Most nutrients tightly controlled: no relationship to intake

• -Web-based, automated FFQs
• - Self-administered, automated, web-based,
• 24-hour dietary recalls
• - Real time data collection of food records
• - Voice recognition software
• - Combining methods: FFQ with 24-hour recalls

• study design)- extent to which results from a study can be extended to universe that is broader than population
• studied.
• Increased by:
• a) subject selection - random selection, differences in confounding recruitment methods (including incentives)
• b) subject number - sufficient statistical poer to detect differences between groups

• -- Still need quality control of measures and judgment
• about degree to which results can be extrapolated

• adequacy of measurement - how well does it measure what it is intended to measure (index or indicator reflects what it is intended to
• measure)

• 1) Reliability (reproducibility)- Degree to which repeated measures give same value
• -Affected by:
• a) True variation in the measurement
• b) Random error (or random variation)
• c) Systematic errors
• 2) Face Validity - judgment (often by experts) that tool measures what it is supposed to
• 3) Content Validity - are all components of the
• construct (e.g. diet) measured by the tool?
• 4) Criterion validity - how well the tool performs
• compared with another tool (gold standard?) known
• to measure the same thing?

• 1) Reliability
• - Standardize protocols; calibrate instruments
• - Use protocols suitable for situation
• - Include test-retest as part of study design
• 2) Get expert opinion before implementing tool
• 3) Conceptualize content of construct and choose tools accordingly (e.g. based on literature)
• 4) Compare between and among tools

A special type of bias (therefore affects validity)1 A characteristic or variable that is distributed differently in the study and control groups, and affects the outcome being assessed. Eg. (Randomization, stratification)

• Ways to control
• 1) study design - randomization, stratificaton, recruitment restirction, matching
• 2) Post Data Collection - stratification (in analysis) - multivariate analysis

• -complete and accurate record of all food consumed, where the choice of food consumed has not been influenced by the act of recording
• Valid FFQ
• - accurate reflection of typical food intake over
• a designated period of time - undistorted by false memory or desire to appear "norma"

Difference between the true value and the value obtained from a measurem

-Source of error in which measurements consistently depart from the true values ( if a instrument has a systematic error then the average of many repeat measures does not approximate the true value)

• Bias is either negative or positive
• 1)Negative bias - means that dietary intakes are
• systematically under-estimated
• 2) Positive bias - means that dietary intakes are
• systematically over-estimated

• Eg. Doubly labelled water studies suggest that obese
• people are likely to systematically under-report
• energy intake

• -Unpredictable error
• affects reproducibility (or precision) of a method
• - caused by true variation in intake + measurement error
• - reduces the strength of associations observed between diet and markers for disease
• - reduces the statistical power of studies

• The ability of a measurement to be reproduced - i.e.
• produces the same answer on repeat administrations

• -Almost impossible to achieve at the individual level food intake varies widely over time
• i.e. if you did 24 hour recalls on 1 person on 4
• different occasions, you would be unlikely to get
• exactly the same energy and nutrient intakes on each
• occasion

• 1) Respondent error: either random or systematic
• OR inability/unwillingness to comply with protocols
• 2) Interviewer bias:
• - Monday interview vs friday, interviewers with desires to promote healthy eating may obtain biased results, interviewed may not click with certain individuals but have very good rapport with others
• 3) Respondent memory lapses
• -use memory aids (e.g. food models, photographs)
• - not such a big problem if the time period between food intake and recall of the intake is short
• - foods contributing to the main part of the meal are remembered better than minor or odd foods
• 4) Incorrect estimation of portion size
• - probably a large source of measurement error
• -using average portion weight instead of a estimate can result in substantial error
• 5) Honest vs desirable responses
• 6) Foods served vs foods consumed
• 7) Subject characteristics (obesity, age,
• motivation, etc)

• Random error- within person error
• Systematic error- Person-specific bias
• - Constant additive error
• - Intake-related bias

• - Inaccurate conversion from estimates or household measures to actual weights
• - May not have appropriate tables for food composition data
• - Systematic mistakes when entering data - e.g. entering data - eg. orange juice as concentrated OJ can result in vitamin C estimates 5x greater than they should be