Exposure and Outcome Measurement OM 53

8/8/2019 Exposure and Outcome Measurement OM 53

1/42

1

Exposure and Outcome

MeasurementSuthee Rattanamongkolgul

M.D., MPH, Ph.D.


2/42

2

Data collection methods

1. Personal interview face-to-face

telephone

2. Self-administered questionnaire

postal direct

3. Physical examination: blood pressure, weight

4. Physical, chemical and biological sample: X-ray, bloodsugar, blood culture

5. Environmental measurement: pesticides, air pollution

6. Observation: race, sex, diet, behavior

7. Diaries: dietary record, body temperature

8. Reference to Record: Medical record

9. Use ofproxy ofrespondents: Relatives


3/42

3

True value (T) bias (b) Subject error (E)

Measurement error

Measured value (X)

Measurement error


4/42

4


5/42

5

Weight (Kg.)student 1 student 2 student 3 student 4

Measure weight (X) 61 50 70 63

True weight (T)(by average form last 5 days)

59 52 69 60

Measurement error 2 -2 1 3

Error of scale/toos (b) 1 1 1 1

Subject error (E) 1 -3 0 2

Body weight of 4 medical students


6/42

6

Errors in clinical research


7/42

7

Scale of measurement

Categorical variable Continuous variable

Nominal Data Ordinal Data Interval Data Ratio Data

Hair color Muscle power Exam score Age


8/42

Frequency

True valueMeasurement

A C

Unreliable

Invalid

B D

DIFFERENCES BETWEEN

VALIDITY AND RELIABILITY


9/42

Frequency

True value

Measurement

A C

Unreliable

Invalid

B D

DIFFERENCES BETWEEN

VALIDITY AND RELIABILITY

A- Valid and reliable

B- Valid but not reliableC- Not valid but reliable

D- Not valid and not

reliable


10/42

1010

Types of ValidityTypes of Validity

Content validityContent validity

Construct validityConstruct validity

Criterion validityCriterion validity


11/42

1111

Content validityContent validity

involves the systematic examinationinvolves the systematic examinationof the test content to determineof the test content to determine

whether it covers a representativewhether it covers a representativesample of the behaviour domain to besample of the behaviour domain to bemeasuredmeasured

(Anastasi & Urbina,(Anastasi & Urbina, 19971997 p.p. 114114).).


12/42

1212

Construct validityConstruct validity

refers to the totality of evidence aboutrefers to the totality of evidence aboutwhether a particular operationalization of awhether a particular operationalization of a

construct adequately represents what isconstruct adequately represents what isintended by theoretical account of theintended by theoretical account of theconstruct being measured.construct being measured.

(Demonstrate an element is valid by relating(Demonstrate an element is valid by relating

it to another element that is supposivelyit to another element that is supposivelyvalid.)valid.)


13/42

1313

Criterion validityCriterion validity

involves the correlation between theinvolves the correlation between thetest and a criterion variable (ortest and a criterion variable (or

variables) taken as representative ofvariables) taken as representative ofthe construct .the construct .


14/42

14

Conceptual Definition vs. Operational Definition

Obesity

Intelligence

Hepatitis B

Conceptual Definition Operational Definition

Body mass index

IQ

HBS Ag+


15/42

15

Bias in data collection


16/42

Information BiasInformation Bias(measurement bias)(measurement bias)

Subject BiasSubject Bias

Recall BiasRecall Bias

Respondent BiasRespondent Bias

Observer BiasObserver Bias

Instrument BiasInstrument Bias


17/42

17

Effects of information bias:Misclassifaction

Caused by lack of sensitivity and specificity

Leads to misclassification of either exposure or

outcome 2 Possibilities:Nondifferential misclassification:

rate of misclassification is the same in both groups(exposed/non-exposed or cases/controls)

Differential misclassification: rate of misclassification differs in both groups


18/42

18

Non-differential Misclassification

Case Control

Exposed 200 100Non-exposed 100 200

Sensitivity 100% , Specificity 60%

Case Control

Exposed 240 60Non-

exposed

180 120

True OR = 4.0 Apparent OR = 2.7- 40 false+

+ 40 false+

- 80 false+

+ 80 false+

Reduction!


19/42

19

Differential Misclassification

Case Control

Exposed 200 100

Non-exposed 100 200

Case Control

Exposed 240 60

Non-exposed 100 200

True OR = 4.0 Apparent OR = 8.0

Sensitivity Specificity

Exposed 100% 60%

Non-exposed 100% 100%


20/42

20

Questionnaire Design


21/42

21

To study income of bank managers

How much have you earned each month?F < 5,000 Baht

F5,000- 10,000 Baht

F 10,000- 15,000 Baht

F 15,000- 20,000 Baht

F20,000 Baht or more

Bad questionnaire design


22/42

22

How often do you eat vegetable?

F No eating

F Not frequentF Frequent

F Very frequent



23/42

23

How often do you eat

sticky rice ?

F No eatingF Not frequent

F Frequent

F Very frequent

How often do you eat

sticky rice

F< 1 time/monthF1-3 time/month

F1-2 time/week

F3-4

time/weekF5-6 time/week

F every day


24/42

24


How often do you intake caffeine ?

How often do you drink coffee ?How often do you drink chocolate ?

How often do you eat chocolate ?

How often do you drink .. ?..


25/42

25


Have you used any pesticide ?

Timing not clear ?What are exactly pesticides?

In house pesticide vs. garden pesticide?

Organic pesticide vs. Inorganic pesticide


26/42

26


Did you used any pesticide in 1999 ?

Difficult to remember!Need introduction questions to recall

If interested in gardening pesticide, may ask

Did you do gardening in 1999? Did you use pesticide when you did

gardening?


27/42

27


Do you like Pad-Thai and Tom-Yum?

Do you like Pad-Thai?

Do you like Tom-Yum?


28/42

28


Do you like Phuket? [ ] yes [ ] no

Have you every been to Phuket?

[ ] yes [ ] no

If yes, do you like Phuket? [ ] yes [ ] no


29/42

29


When was your last time watching movie?

When was your last time watching moviein a cinema?


30/42

30

Format of questionnaire

Using different color for long questionnaire with sections

Look nice, tidy, appealing

Large enough letter type and easy to read font

Clear instruction to answer

Logical order of question

Order of question

Thank you subject at the end

Branching of question

Do you have children?If yes, how many? ______ boy _____ girl


31/42

31

Questionnaire design

1. Remember your survey's purpose;If in doubt, throw it out

2. Keep your questions simple3. Consider alternate ways to ask sensitive

questions

4. Keep open-ended questions to a minimum

5. Response options need to be mutuallyexclusive and exhaustive

6. Provide a meaningful scale

Pamela NarinsPamela Narins


32/42

32

Pilot testing

Using representative of target population

Evaluated by a number of researcher who used

Before testing in target population, it may be

tested using relatives, friends, or colleagues.

Researcher should be critical to their work and

open to suggestion

Written comments should be obtained thatcancer evaluated by a number of researchers


33/42

33

The better pilot testing the less regrets

there will be at the end of the study


34/42

34

What was your body shape when you were 18 years old?

Male

Female

Using picture in questionnaire


35/42

Graphic Rating Scale Stressing Pictorial Visual

Communications

3 2 1Very Very

Good Poor

How much do you like the treatment?


36/42

The Attitude Measuring Process

Ranking - Rank order preference

Rating - Estimates magnitude of a characteristic

Sorting -Arrange or classify concepts

Choice - Selection of preferred alternative


37/42

Likert Scale for Measuring Attitudes

Your skin is very dark ?

___Strongly Agree___Agree

___Not Sure

___Disagree

___Strongly Disagree


38/42

38

Odd vs. Even items

Odd

Strongly Agree _____

Agree _____

Neutral _____

Disagree _____

Strongly disagree_____

Even

Strongly Agree_____

Agree ___

Disagree _____

Strongly disagree___


39/42


40/42

40

Forced vs. Unforced

ScalesForced

Extremely Reliable ___

Very Reliable ___

Somewhat Reliable ___

Somewhat Unreliable ___

Very Unreliable ___

Extremely Unreliable ___

Unforced

Extremely Reliable ___

Very Reliable ___

Somewhat Reliable ___

Somewhat Unreliable ___

Very Unreliable ___

Extremely Unreliable ___

Dont know ___


41/42

41

Labeled vs. End Anchored

Scales

Labeled

Excellent _____

Very Good _____

Fair _____

Poor _____

Very Poor _____

End Anchored

Excellent _____

_____

_____

_____

Poor _____


42/42

42

Numberof Scale Points

5 Point

Excellent _____

_____

_____

_____

Poor _____

10 Point

Excellent _____ __________________________

_____________

_____________

_____________

__________________________

_____________

_____________Poor

Exposure and Outcome Measurement OM 53

Documents

Transcript of Exposure and Outcome Measurement OM 53