Selecting a Measure of Selecting a Measure of Dietary Intake for Children Dietary Intake for Children Tom Baranowski, Ph.D. Tom Baranowski, Ph.D. Professor of Pediatrics Professor of Pediatrics (Behavioral Nutrition and Physical Activity) (Behavioral Nutrition and Physical Activity) Children Children ’ ’ s Nutrition Research Center s Nutrition Research Center Baylor College of Medicine Baylor College of Medicine University of Washington – March 7, 2006
Transcript
Selecting a Measure of Selecting a Measure of Dietary Intake for ChildrenDietary Intake for Children
Tom Baranowski, Ph.D.Tom Baranowski, Ph.D.Professor of PediatricsProfessor of Pediatrics
(Behavioral Nutrition and Physical Activity)(Behavioral Nutrition and Physical Activity)ChildrenChildren’’s Nutrition Research Centers Nutrition Research Center
Baylor College of MedicineBaylor College of Medicine
University of Washington – March 7, 2006
Common QuestionCommon Question
I want to measure 3I want to measure 3rdrd grade studentsgrade students’’ fruit fruit and vegetable intake as the primary and vegetable intake as the primary outcome for my study. I only have a outcome for my study. I only have a couple of minutes, so I need to do it with couple of minutes, so I need to do it with one or two questions. What do you have?one or two questions. What do you have?
(but impossible)(but impossible)
Diet AssessmentDiet Assessment
Every measure has errorEvery measure has errorDiet measures are fraught with errorDiet measures are fraught with errorHow much error can you tolerate?How much error can you tolerate?
Can you afford not to detect a true difference?Can you afford not to detect a true difference?Can you afford not to detect a true Can you afford not to detect a true relationship?relationship?
(For review of specific scales see: JW McClellan et al, Journal of Nutrition Education 2001, 33: S35-S48)(For review of studies see: T Baranowski et al, Preventive Medicine 2004, 38: S1-S13)
Issues in Selecting a MethodIssues in Selecting a Method
Prior AssessmentPrior AssessmentValidityValidityReliabilityReliabilitySensitivity to changeSensitivity to change
PrecisionPrecisionLevel of detailLevel of detailHabitual behaviorHabitual behaviorReactivityReactivity
00--7 years7 yearsProbably need a proxy reporter or observerProbably need a proxy reporter or observer
88--10 years10 yearsMay need assistance (national surveys)May need assistance (national surveys)
1111--13 years13 yearsCognitive capabilities not clearCognitive capabilities not clear
14 years +14 years +Answer like adults?Answer like adults?
Characteristics of diet among Characteristics of diet among adolescentsadolescents
Rapidly changing eating habitsRapidly changing eating habitsUnstructured eating (snacks, meal Unstructured eating (snacks, meal skipping)skipping)Eating away from home (31% foods)Eating away from home (31% foods)High levels of restrained eatingHigh levels of restrained eating
Characteristics related to diet Characteristics related to diet assessmentassessment
Increased under reporting with ageIncreased under reporting with ageDifficulty with time referenceDifficulty with time referenceGreater under reporting among obese Greater under reporting among obese (high levels of overweight)(high levels of overweight)Data collection in schools Data collection in schools -- challengingchallenging
Most of method research done in Most of method research done in adultsadults
Methods usually work less well in childrenMethods usually work less well in children
Common Methods of Assessing Common Methods of Assessing Validity of Diet AssessValidity of Diet Assess
Compare againstCompare againstObservation of consumption Observation of consumption -- quantitativequantitative
At the food levelAt the food levelReactive?Reactive?
Doubly labeled water Doubly labeled water -- quantitativequantitativeEnergy expenditure/intakeEnergy expenditure/intake2 weeks: urine/saliva2 weeks: urine/saliva
Biomarkers Biomarkers –– non quantitativenon quantitativeNutrient specificNutrient specificBlood sampleBlood sample
What do we usually want to do with What do we usually want to do with dietary intakedietary intake
ClassificationClassificationIs this person in a high/low consuming group? Is this person in a high/low consuming group? (misclassification error)(misclassification error)
Correlation with other variablesCorrelation with other variablesDo children with this psychosocial Do children with this psychosocial characteristic eat more/less? (attenuation)characteristic eat more/less? (attenuation)
In outcome studies, unreliability In outcome studies, unreliability reduces effective sample sizereduces effective sample size
Effective sample = n R
TRUE OBTAINED
Misclassification ErrorMisclassification Error
(de Moor C et al Public Health Nutrition 2003, 6(4); 393-399)
Identification of the errors of assignment of a single cut pointIdentification of the errors of assignment of a single cut pointwhen the true values are knownwhen the true values are known
xo
xT
Obs
erve
d in
take
True intake
Misclassification ErrorMisclassification Error
(de Moor C et al Public Health Nutrition 2003, 6(4); 393-399)
Accuracy (AC) for four percentile cut points as the Accuracy (AC) for four percentile cut points as the correlation between true and observed values increasescorrelation between true and observed values increases
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 10.0
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.0
9.0
10.0
10th percentile
25th percentile
33rd percentile
50th percentile
Correlation
AC
Correction for attenuationCorrection for attenuation
rrTT RRxxRRyy = = rrxyxy
y
xHigh reliability
y
xLow reliability
Most common tools for measuring Most common tools for measuring intakeintake
Food Frequency Questionnaire (FFQ)Food Frequency Questionnaire (FFQ)24 hour dietary recall24 hour dietary recallFood RecordFood RecordInnovative MethodsInnovative Methods
Food Frequency Questionnaire Food Frequency Questionnaire (FFQ)(FFQ)
Age groupAge groupall adultsall adultschildren specificchildren specificseniors specificseniors specificethnic specificethnic specificproxies for proxies for young childrenyoung children
Strengths of the FFQStrengths of the FFQ
Assesses entire diet re: specific nutrients or food Assesses entire diet re: specific nutrients or food groupsgroupsAverages across time interval (<day to day Averages across time interval (<day to day varvar))Done in a short time interval (<1hr?)Done in a short time interval (<1hr?)Requires no special expertise to administerRequires no special expertise to administerStraight forward conversion to foods/nutrientsStraight forward conversion to foods/nutrientsInfrequent updating of nutrient data base or Infrequent updating of nutrient data base or foods neededfoods needed
Major Method IssuesMajor Method IssuesFood Frequency QuestionnaireFood Frequency QuestionnaireAdultsAdults::
Supposed to cover long time, but test retest Supposed to cover long time, but test retest relrel is lowis lowError structureError structure
Correlated errorsCorrelated errorsLonger Longer qq’’nairesnaires –– higher over estimationhigher over estimationRequires year long memoryRequires year long memory
Season effects?Season effects?Requires averaging across irregular consumptionRequires averaging across irregular consumptionPortion size est. adds littlePortion size est. adds littleIdentify under/over reporters & modifyIdentify under/over reporters & modify
*
**
.
.
.
.
.
.
.
.
.
.
. ..
Day 1
XW1
XW2 XW3 XW4
- 3
- 2
- 1
1
2
3
4
5
6
7C
orre
latio
n C
oeffi
cien
ts
. Indicates r is not significantly different from zero
* Indicates significance a < 0.10Indicates r significance < 0.05
Experimental Adults (n=12)
Control Children (n=13)
Experimental Children (n=9)
Control Adults (n=12)
(Baranowski Res Q Exer & Sport 1988, 59(4): 314-27)
Correlations backward in time between a retrospective Correlations backward in time between a retrospective estimate of minutes of aerobic activity per day from estimate of minutes of aerobic activity per day from
selfself--monitoring formsmonitoring forms
Usual Method of FFQ ValidationUsual Method of FFQ ValidationFFQ FFQ vsvs multiple 24hdrmultiple 24hdr
Problem: Common self report errorsProblem: Common self report errorsOPEN (Observing Protein and Energy Nutrition) StudyOPEN (Observing Protein and Energy Nutrition) Study
(among mostly middle class adults)(among mostly middle class adults)
0.220.220.330.33FFQsFFQs
0.260.260.410.4124hdr24hdrProteinProtein
0.100.100.190.19FFQsFFQs
0.240.240.390.3924hdr24hdrEnergyEnergy
WomenWomenMenMenCorrelationsCorrelations
(Subar et al Am J Epid 2003, 158; 1-13)
OPEN Study (OPEN Study (SubarSubar et al)et al)
2727--32%32%3030--34%34%FFQsFFQs
1111--15%15%1111--12%12%24hdr24hdrProteinProtein
3434--38%38%3131--36%36%FFQsFFQs
1616--20%20%1212--14%14%24hdr24hdrEnergyEnergy
WomenWomenMenMen% Under reporting against biomarker% Under reporting against biomarker
- Under reporting increased with level of intake
When is FFQ appropriateWhen is FFQ appropriate
Only need to determine a personOnly need to determine a person’’s relative s relative position in a distribution on a particular position in a distribution on a particular food/nutrientfood/nutrient
Better for group differencesBetter for group differencesWill use large samplesWill use large samples
Time & resource limitedTime & resource limitedAn instrument has been validated for the An instrument has been validated for the variable of interest in the target population variable of interest in the target population under similar circumstancesunder similar circumstances
(See J Cade et al Public Health Nutrition 2002,5: 567-587)
Age groupsAge groups Variables possibleVariables possiblefoods/beverages foods/beverages consumedconsumednutrients consumednutrients consumedmeals/snacks meals/snacks patternpatternwho was therewho was therewhen occurredwhen occurredduration of mealduration of mealwas TV on?was TV on?where consumedwhere consumedfrom where did food from where did food come?come?
Strengths of the 24hdrStrengths of the 24hdrAssesses one day with some accuracyAssesses one day with some accuracyProvides a more detailed data setProvides a more detailed data set
After After WeekendWeekendNext DayNext DaySame DaySame Day
(JADA 1997, 97:1293-8)
Can parents accurately report for Can parents accurately report for their children?their children?
Procedure Design:Procedure Design:Day 1Day 1 Day 2Day 2All dayAll day MotherMotherobservationobservation proxyproxyof childof child 24hdr24hdr(7am(7am––7pm)7pm)
SESSESLow High
At home < 4.5hr
Not at home> 4.5hr
(Baranowski et al JADA 1991, 91: 669-674)
Sample Design:Sample Design:
Can parents accurately report for Can parents accurately report for their children?their children?
At home group: 27 of 29 could reportAt home group: 27 of 29 could reportNot at home group: 13 of 27 could reportNot at home group: 13 of 27 could reportNo No differncediffernce by SESby SES
(Baranowski et al JADA 1991, 91: 669-674)
Agreement & Errors of Mothers against Agreement & Errors of Mothers against ObservationObservation
10%10%5%5%7%7%Partial AgreementPartial Agreement
16%16%19%19%18%18%Under reportingUnder reporting
12%12%9%9%10%10%Over reportingOver reporting
62%62%67%67%65%65%AgreementAgreement
Not at Not at home home
mothersmothers
At home At home mothersmothers
All All mothersmothers
(Baranowski et al JADA 1991, 91: 669-674)
Limitations of the 24hdrLimitations of the 24hdrComplicated to administerComplicated to administer
Require RD, NDS cert?Require RD, NDS cert?Multiple QC reviewsMultiple QC reviewsMultiple nonMultiple non--contiguous days of assesscontiguous days of assessSocial desirability of responseSocial desirability of responseSubstantial under reportingSubstantial under reportingTakes 20Takes 20--45 min/day to administer45 min/day to administerBenefits from NDS to promptBenefits from NDS to promptPortion size estimation problemsPortion size estimation problemsBehavBehav vsvs EpidEpid Coding: What to Count?Coding: What to Count?Complex program to Complex program to estest nutrientsnutrientsRegularly update program to Regularly update program to estest nutrients (20nutrients (20--30,000 30,000 new foods each year)new foods each year)Different software provide different nutrient estimatesDifferent software provide different nutrient estimates
Food Recognition FormFood Recognition Form
List of foodsList of foodsHad yesterday? (yes/no)Had yesterday? (yes/no)
by meal/snackby meal/snack
++++++’’s easy for children to completes easy for children to complete-- -- -- ’’s limited # of foodss limited # of foods
only one dayonly one dayno portion sizeno portion size
(Simons-Morton, Baranowski et al AJPM 1990, 6: 218-27)
Food RecordsFood RecordsTime coveredTime covered
multiple multiple random daysrandom daysmultiple multiple sequential sequential daysdaysone dayone day
Diet assessedDiet assessedeverythingeverythingweighed or weighed or not?not?portion sizesportion sizesadd add onsons
Age groupsAge groups““9 9 yotyot??
Variables PossibleVariables PossibleSame as 24hdrSame as 24hdr
Major Method IssuesMajor Method IssuesFood RecordsFood Records
ChildrenChildren::Requires trainingRequires trainingSpelling problems (Spelling problems (olivnacoqtalolivnacoqtal))Portion size problems Portion size problems –– units/no referentsunits/no referentsRecall problems Recall problems –– if not done right after meal (fill out if not done right after meal (fill out weekweek’’s record in front of center)s record in front of center)Needs daily supervision/prompting/reviewNeeds daily supervision/prompting/reviewSubstantial coding timeSubstantial coding timeNeed multiple daysNeed multiple daysIf weighed If weighed –– need equipmentneed equipment
-- takes special peopletakes special people
Financial CostsFinancial Costs
Paper copiesPaper copiesSoftwareSoftwarePortable hardwarePortable hardwareScalesScalesStaff time administratorStaff time administrator
DietitianDietitianRARA
Staff time QCStaff time QCData entry staff timeData entry staff time
Biomarker solutionsBiomarker solutions
Blood sampleBlood sampleCarotenoidsCarotenoids
DLW DLW –– salivasalivaTotal energyTotal energy
TransdermalTransdermal infrared technologyinfrared technologyIron, Iron, carotenoidscarotenoids (machine can be trained), (machine can be trained), glucoseglucose
(S. Pilotto Lasers Med Sci 2001, 16: 2-9)
Surrogate VariablesSurrogate Variables
Home food availabilityHome food availabilityGrocery store purchases receiptsGrocery store purchases receiptsStore food availabilityStore food availability
Statistical ManipulationStatistical Manipulation
Post adjust distributions of intake from Post adjust distributions of intake from existing methodsexisting methods
Propensity to be missing adjustmentPropensity to be missing adjustment((CarriquiryCarriquiry, , J J NutrNutr 2003, 133: 601S2003, 133: 601S--8S)8S)(Baker et al, (Baker et al, BiostatisticsBiostatistics 2006, 7: 292006, 7: 29--40)40)
Moment reconstructionMoment reconstruction(Freedman et al, (Freedman et al, BiometricsBiometrics 2004, 60: 1722004, 60: 172--181)181)
Innovations in Dietary AssessmentInnovations in Dietary Assessment
Technological solutions Technological solutions –– minimal respondent minimal respondent burdenburdenFoods on trays with numbersFoods on trays with numbers
Take pictures before and after mealTake pictures before and after mealNot catch spills/tradesNot catch spills/trades
Point of purchase/service sales dataPoint of purchase/service sales dataSupermarkets/groceriesSupermarkets/groceriesSchoolsSchools
Not consumptionNot consumptionNot who consumesNot who consumes
Innovations in Dietary AssessmentInnovations in Dietary Assessment
Technological solutions Technological solutions –– more participant effortmore participant effortUse cell phone with videoUse cell phone with video
Take and transmit pictures before and after mealsTake and transmit pictures before and after mealsUse visual pattern recognition technique to identify Use visual pattern recognition technique to identify foods and portionsfoods and portions
Use USP bar codes to identify foodsUse USP bar codes to identify foodsQueries for how much (also can be bar coded)Queries for how much (also can be bar coded)
Diet Self Monitoring PDADiet Self Monitoring PDAAudio enhanced for childrenAudio enhanced for children
Prompted Momentary Time Sampling with PDA Prompted Momentary Time Sampling with PDA or Cell Phoneor Cell Phone
Innovations in Dietary AssessmentInnovations in Dietary Assessment
Food Intake Recording Software System (Food Intake Recording Software System (FIRSStFIRSSt) ) (Baranowski et al (Baranowski et al JADAJADA 2002, 102: 3802002, 102: 380--5)5)ASA24 (ASA24 (SubarSubar et al, 2006 abstract)et al, 2006 abstract)Moore et al (Moore et al (Euro J Euro J ClinClin NutrNutr 2005, 59: 8092005, 59: 809--16)16)TeresitaTeresita HernandezHernandezZoellnerZoellner et al (et al (J ExtensionJ Extension 2006, 44: 12006, 44: 1--17)17)
ConclusionsConclusions
All measures have errorAll measures have errorMeasures of diet have substantial errorMeasures of diet have substantial error
Measures with children Measures with children –– even more erroreven more errorTomTom’’s order or preferences order or preference
24hdr24hdrFFQ (only if have to)FFQ (only if have to)
Gadgets & widgets offer some promise for Gadgets & widgets offer some promise for the future, but no hope nowthe future, but no hope now
