week7

Question 1

Measurement errors are either:

Answer 1

Random
or
Systematic

Question 2

What is random error

Answer 2

e.g bias, misrecording of food
Affects the precision (reproducibility)
Can be decreased (but not eliminated entirely) by
- increasing the number of participants surveyed
and/or
- increasing the number of days over which the dietary intake of a participant is measured
Can occur across ALL days of the week and across ALL participants

Question 3

What is Systematic Error

Answer 3

Most serious type as it may introduce significant bias

The bias may be positive or negative i.e dietary intakes may be over-estimated or under-estimated

Direction (positive or negative) and extent (size) of systematic errors are very difficult to quantify
Can’t be minimised by increasing the number of observations
May exist for certain participants e.g: those with obesity are prone to under-estimate their usual food intake
May exist for specific interviewers
May be associated with certain foods only (eg) alcohol

Question 4

The direction (positive or negative) and size of random and systematic error varies with:

Answer 4

The methodology used
The population being studied
The nutrients of interest

Question 5

How to control measurement error

Answer 5

Achieved by proper quality control measures e.g:
Training and re-training sessions for interviewers and coders
Use of standardised interviewing techniques and questionnaires
Pre-tested questionnaires and inclusion of a pilot study
NB: Good quality control will still not eliminate measurement error- particularly systematic error

Question 6

List the 7 major sources of error

Answer 6

Respondent bias
Non-response bias
Interviewer bias
Respondent memory lapses
Incorrect estimation of portion sizes
Supplement usage
Coding/computation errors

Question 7

Respondent bias

Answer 7

May arise because the subject:
- misunderstands what the investigator has requested
- receives non-verbal cues to the “right answers”
- has a need to give “socially desirable” answers

= systematic mis-reporting of foods consumed

Question 8

Under-reported foods

Answer 8

Can involve:
- under-recording → failure to record all the foods consumed during the period and/or under-estimation of their amounts
- under-eating → when subjects eat less than usual
- specific foods and beverages may be selectively under-reported
- weight status – consistently associated with under-reporting (likely to increase as BMI increases)

Question 9

Social desirability and approval biases

Answer 9

Social desirability → the tendency to respond in such a way as to avoid criticism

Social approval → the tendency to seek praise

May affect the way that people report foods which have “good”/ “bad” connotations

Question 10

Non-response bias

Answer 10

Non response or poor compliance from a specific subset of otherwise randomly selected subjects
May bias the dietary data because the sample loses its randomness (ie) is unrepresentative
(eg) busy parents vs retired subjects
Need to minimize non-response bias (eg) simplifying dietary assessment methods

Question 11

Interviewer bias

Answer 11

May include the following errors
incorrect use of probing -questions
-incorrect recording of responses
-biases associated with the -interview setting
distractions
-degree of rapport between investigator and participant
Errors may be random across days and participants and/or
Systematic for a specific interviewer or
Exist as an interaction between certain interviewers and participants only
Efforts should be made to minimize interviewer bias (eg) carefully designed and standardized interviewing protocols
Interviewers should be trained to anticipate and recognize potential sources of bias
Cultural and ethnicity issues

Question 12

Respondent memory lapses

Answer 12

May affect recall methods in two ways:
1. participant may fail to recall foods actually consumed
2. participant may report foods that were not consumed during the study period

Reduced by asking “probing questions” and/or using memory aids e.g: food photographs
It may be reduced if the time period between actual food intake and recall of that food intake is short (24 hrs)
Generally foods contributing to the main part of the meal are remembered better

Question 13

Incorrect portion sizes

Answer 13

Probably the largest source of measurement error
Occurs when respondents fail to quantify accurately the amount of food consumed
Alternatively the respondent’s concept of a normal portion may deviate from the standard (or the interviewer may assume an answer of a normal portion/average serving size
Extensive errors are introduced if standard portion sizes are assumed

Question 14

Supplement usage

Answer 14

May (easily) be omitted from the dietary record
Failure to account for the
- type (brand variability is large)
- frequency
- amount of supplements used
→ systematic under-estimation of the
intake of certain nutrients

Question 15

Coding and computation errors

Answer 15

Arises when portion sizes are estimated from household measures and/or when food items are incorrectly coded
Coding errors are mostly due to inadequate descriptions of foods rather than weight errors
“Rules for coding” should be established at the outset to deal with incomplete or ambiguous description of foods

Question 16

Precision

Answer 16

Is a function of
1.Measurement error (which can be minimised but not eliminated)
2. True variation in daily nutrient intakes (will always remain)
→ therefore
if the results of two dietary surveys disagree, this does not mean the method is imprecise – rather food intakes may have changed

Question 17

True variability - intra and inter subject variation

Answer 17

Due to dietary differences
Among individuals [between or inter- subject variation and
Within one individual over time [within or intra- subject variation]

Intra-subject variation in dietary intake is usually greater than inter-subject variation

Question 18

Estimation of precision depends on various factors:

Answer 18

Time frame of method
Population group
Nutrients of interest
Techniques used to measure quantities of foods consumed
Inter- and Intra-subject variation in intakes

Question 19

Measuring precision

Answer 19

test: re-test design i.e: repeating the same dietary method on the same participants after a pre-selected time interval

Care is needed: The second measurement should not be influenced by the first one

Question 19

Measuring precision

Answer 19

test: re-test design i.e: repeating the same dietary method on the same participants after a pre-selected time interval

Care is needed: The second measurement should not be influenced by the first one

Question 20

24 hr recall

Answer 20

The method can provide a relatively precise estimate of group mean intake

However - if single 24 hr recall methods are used to estimate group mean intake, the number of subjects required is large

For estimating the mean intake of an individual, the 24hr recall is most unlikely to represent usual intake, i.e the method has low precision. This is due to a large intra-subject variation in food intake

Question 21

7 day food records

Answer 21

Generally good agreement between group mean values for energy and most nutrients when tested on two separate occasions

Question 22

Diet histories

Answer 22

The precision depends on:
The time frame of the method (eg 1 month…or 3 months)
The time interval between the records
The technique used of measuring amount of foods consumed
The population group

Generally the dietary history gives good precision at group level- especially over relatively short time frames for:
Energy and nutrients for average week-days
Poorer at week-end days because of greater variation in food intakes

Question 23

Food Frequency Questionnaires (FFQ)

Answer 23

Limited number of studies
but they suggest good precision

Question 24

Sources of true variability

Answer 24

Inter- subject variation
A measure of how subjects differ from each other in their true daily intake
Varies with the nutrient being investigated
For most nutrients- inter-subject variation is usually smaller than intra-subject variation
Result → mean intake of a group can generally be assessed MORE precisely than individual intakes
Inter-subject variation can be overcome if sample size
Is as large as possible
Is representative of the group being studied

Question 25

Age and gender

Answer 25

Energy and nutrient intakes are generally presented separately by age and gender

Gender differences are largely due to differences in amounts consumed, rather than pattern of consumption

If nutrient intakes are expressed as nutrient densities (i.e. amount of nutrient/total grams of food consumed) - differences tend to disappear (as you are adjusting for the amount of energy consumed)

Question 26

Intra-subject variation is a measure of:

Answer 26

TRUE day-to-day variation in intake by the same participant
Cannot be measured by diet history
Statistically assessed by analysis of variance
Significance of intra-subject variation:
Very important if usual intakes are required for correlation with biochemical or clinical parameters
Multiple measurement days are required to characterize usual intake more precisely
Extent partly depends on diversity of food intake

Question 27

Significance of intra-subject variation

Answer 27

Very important if usual intakes are required for correlation with biochemical or clinical parameters
Multiple measurement days are required to characterize usual intake more precisely
Extent partly depends on diversity of food intake

Question 28

Day of the week effects

Answer 28

Group mean nutrient intakes (per day) and individual usual intakes (per day) MAY VARY WITH THE DAY OF THE WEEK
E.g: Women are reported to eat more on Sundays and both genders have lower fibre intakes at the week-ends
NOT ALL NUTRIENTS ARE AFFECTED
Day of the week effects can be accounted for by proportionally representing all days of the week in the study design

Question 29

Seasonal effects

Answer 29

Depends on:
Population group
Socio-economic status
Country
Generally seasonal effects will probably be greater for food items rather than nutrient intakes

Question 30

Training effects

Answer 30

Reaction to repeated interviews
May alter reported intakes over time

Effects are most likely if participants complete recalls on consecutive days (not advised)

Where highly trained interviewers and standardised techniques are used such training effects may not be observed