Post on 13-Dec-2015
1
Class 5
Additional Psychometric Characteristics: Validity and Bias, Responsiveness, Sensitivity to Change
October 22, 2009
Anita L. Stewart Institute for Health & Aging
University of California, San Francisco
2
Overview
Types of validity in health assessment– Focus on construct validity
How bias affects validity– Socially desirable responding and culture as
sources of bias Sensitivity to change
3
Validity
Does a measure (or instrument) measure what it is supposed to measure?
And…Does a measure NOT measure what it is NOT supposed to measure?
4
Valid Scale? No!
There is no such thing as a “valid” scale We accumulate “evidence” of validity in a
variety of populations in which it has been tested
Similar to reliability
5
Validation of Measures is an Iterative, Lengthy Process
Accumulation of evidence– Different samples
– Longitudinal designs
6
Types of Measurement Validity
Content
Criterion Construct
– Convergent– Discriminant– Convergent/discriminant
All can be: Concurrent Predictive
7
Content Validity:
Relevant when writing items Extent to which a set of items represents
the defined concept
8
Relevance of Content Validity to Selecting Measures
“Conceptual adequacy” Does “candidate” measure represent
adequately the concept YOU are intending to measure
9
Content Validity Appropriate at Two Levels
Battery or Are all relevant domainsinstrument represented in an instrument?
Measure Are all aspects of a defined
concept represented in theitems of a scale?
10
Example of Content Validity of Instrument
You are studying health-related quality of life (HRQL) in clinical depression– Your HRQL concept includes sleep problems,
ability to work, and social functioning SF-36 - a candidate
– Missing sleep problems
11
Types of Measurement Validity
Content
Criterion Construct
– Convergent– Discriminant– Convergent/discriminant
All can be: Concurrent Predictive
12
Criterion Validity
How well a measure correlates with another measure considered to be an accepted standard (criterion)
Can be– Concurrent– Predictive
13
Criterion Validity of Self-reported Health Care Utilization
Compare self-report with “objective” data (computer records of utilization)– # MD visits past 6 months (self-report)
correlated .64 with computer records
– # hospitalizations past 6 months (self-report) correlated .74 with computer records
Ritter PL et al, J Clin Epid, 2001;54:136-141
14
Criterion Validity of Screening Measure
Develop depression screening tool to identify persons likely to have disorder– Do clinical assessment only on those who
screen “likely” Criterion validity
– Extent to which the screening tool detects (predicts) those with disorder» sensitivity and specificity, ROC curves
15
Criterion Validity of Measure to Predict Outcome
If goal is to predict health or other outcome– Extent to which the measure predicts the
outcome Example: Develop self-reported war-
related stress measure to identify vets at risk of PTSD– How well does it predict subsequent PTSD
(Vogt et al., 2004, readings)
16
Types of Measurement Validity
Content
Criterion Construct
– Convergent– Discriminant– Convergent/discriminant
All can be: Concurrent Predictive
17
Construct Validity Basics
Does measure relate to other measures in hypothesized ways?– Do measures “behave as expected”?
3-step process– State hypothesis: direction and magnitude
– Calculate correlations
– Do results confirm hypothesis?
18
Source of Hypotheses in Construct Validity Prior literature in which associations
between constructs have been observed– e.g., other samples, with other measures
of constructs you are testing Theory, that specifies how constructs should
be related Clinical experience
19
Who Tests for Validity?
When measure is being developed, investigators should test construct validity
As measure is applied, results of other studies provide information that can be used as evidence of construct validity
20
Types of Measurement Validity
Content
Criterion Construct
– Convergent– Discriminant– Convergent/discriminant
All can be: Concurrent Predictive
21
Convergent Validity
Hypotheses stated as expected direction and magnitude of correlations
“We expect X measure of depression to be positively and moderately correlated with two measures of psychosocial problems”– The higher the depression, the higher the level
of problems on both measures
22
Testing Validity of Expectations Regarding Aging Measure
Hypothesis 1: ERA-38 total score would correlate moderately with ADLS, PCS, MCS, depression, comorbidity, and age
Hypothesis 2: Functional independence scale would show strongest associations with ADLs, PCS, and comorbidity
Sarkisian CA et al. Gerontologist. 2002;42:534
23
Testing Validity of Expectations Regarding Aging Measure
Hypothesis 1: ERA-38 total score would correlate moderately with ADLS, PCS, MCS, depression, comorbidity, and age (convergent)
Hypothesis 2: Functional independence scale would show strongest associations with ADLs, PCS, and comorbidity
Sarkisian CA et al. Gerontologist. 2002;42:534
24
ERA-38 Convergent Validity Results: Hypothesis 1
ERA-38ERA FunctionalIndependence
ADL .19** .20***
PCS-12 .27** .32***
MCS-12 .35** .30**
Comorbidity - .09* ns
Depressive symptoms - .33** - .28**
Age - .24** - .14**
25
ERA-38: Non-Supporting Convergent Validity Results
ERA-38ERA FunctionalIndependence
ADL .19** .20***
PCS-12 .27** .32***
MCS-12 .35** .30**
Comorbidity - .09* ns
Depressive symptoms - .33** - .28**
Age - .24** - .14**
26
Types of Measurement Validity
Content
Criterion Construct
– Convergent– Discriminant– Convergent/discriminant
All can be: Concurrent Predictive
27
Discriminant Validity: Known Groups
Does the measure distinguish between groups known to differ in concept being measured?
Tests for mean differences between groups
28
Example of a Known Groups Validity Hypothesis Among three groups:
– General population– Patients visiting providers– Patients in a public health clinic
Hypothesis: scores on functioning and well-being measures will be the best in a general population and the worst in patients in a public health clinic
29
Mean Scores on MOS 20-item Short Form in Three Groups
PublicGeneral MOS health
population patients patientsPhysical function 91 78 50Role function 88 78 39Mental health 78 73 59Health perceptions 74 63 41
Bindman AB et al., Med Care 1990;28:1142
30
PedsQL Known Groups Validity
Hypothesis: PedsQL scores would be lower in children with a chronic health condition than without
Child report: Total score
Emotionalfunctioning
Chron ill* 77 (16) 76 (22)
Acutely ill* 79 (14) 77 (20) ANOVA, p = .001
Healthy 83 (15) 81 (20)
* Different from healthy children, p < .05
JW Varni et al. PedsQL™ 4.0: Reliability and Validity of the Pediatric Quality ofLife Inventory™ …, Med Care, 2001;39:800-812.
31
Types of Measurement Validity
Content
Criterion Construct
– Convergent– Discriminant– Convergent/discriminant
All can be: Concurrent Predictive
32
Convergent/Discriminant Validity
Does measure correlate lower with measures it is not expected to be related to … than to measures it is expected to be related to?
The extent to which the pattern of correlations conforms to hypothesis is confirmation of construct validity
33
Basis for Convergent/Discriminant Hypotheses
All measures of health will correlate to some extent
Hypothesis is of relative magnitude
34
Example of Convergent/Discriminant Validity Hypothesis Expected pattern of relationships:
– A measure of physical functioning is “hypothesized” to be more highly related to a measure of mobility than to a measure of depression
35
Example of Convergent/Discriminant Validity Evidence
Pearson correlation:
Mobility Depression
Physical functioning .57 .25
36
Testing Validity of Expectations Regarding Aging Measure
Hypothesis 1: ERA-38 total score would correlate moderately with ADLS, PCS, MCS, depression, comorbidity, and age (convergent)
Hypothesis 2: Functional independence scale would show strongest associations with ADLs, PCS, and comorbidity (convergent/discriminant)
Sarkisian CA et al. Gerontologist. 2002;42:534
37
ERA-38 Convergent/Discriminant Validity Results: Hypothesis 2
ERA-38ERA FunctionalIndependence
ADL .19** .20***
PCS-12 .27** .32***
MCS-12 .35** .30**
Comorbidity - .09* ns
Depressive symptoms - .33** - .28**
Age - .24** - .14**
38
ERA-38: Non-Supporting Validity Results
ERA-38ERA FunctionalIndependence
ADL .19** .20***
PCS-12 .27** .32***
MCS-12 .35** .30**
Comorbidity - .09* ns
Depressive symptoms - .33** - .28**
Age - .24** - .14**
39
Construct Validity Thoughts: Lee Sechrest
There is no point at which construct validity is established
It can only be established incrementally– Our attempts to measure constructs help us
better understand and revise these constructs
Sechrest L, Health Serv Res, 2005;40(5 part II), 1596
40
Construct Validity Thoughts: Lee Sechrest (cont)
“An impression of construct validity emerges from examining a variety of empirical results that together make a compelling case for the assertion of construct validity”
41
Construct Validity Thoughts: Lee Sechrest (cont)
Because of the wide range of constructs in the social sciences, many of which cannot be exactly defined..– …once measures are developed and in use,
we must continue efforts to understand them and their relationships to other measured variables.
42
Interpreting Validity Coefficients
Magnitude and conformity to hypothesis are important, not statistical significance – Nunnally: rarely exceed .30 to .40 which may be adequate
(1994, p. 99) – McDowell and Newell: typically between 0.40 and 0.60
(1996, p. 36) Max correlation between 2 measures = square root
of product of reliabilities– 2 scales with .70 reliabilities, max correlation .70– Correlation of .60 would be “high”
43
Overview
Types of validity in health assessment– Focus on construct validity
How bias affects validity– Socially desirable responding and culture as
sources of bias Sensitivity to change
44
Components of an Individual’s Observed Item Score (from Class 3)
Observed true item score score
= + error random
systematic
45
Random versus Systematic Error
Observed true item score score
= + error random
systematic
Relevant to reliability
Relevant to validity
46
Bias is Systematic Error
Affects validity of scores– If scores contain systematic error, cannot
know the “true” mean score
– Will obtain an observed score that is either systematically higher or lower than the “true” score
47
“Bias” or “Systematic Error”?
Bias implies that the direction of error known
Systematic error – direction neutral– Same error applies to entire sample
48
Sources of “Systematic Error” in Observed Scores of Individuals
Respondent– Socially desirable responding– Acquiescent response bias– Cultural beliefs (e.g., not reporting distress)– Halo affects
Observer– Belief that respondent is ill
Instrument
49
Socially Desirable Responding
Tendency to respond in socially desirable ways to present oneself favorably
Observed score is consistently lower or higher than true score in the direction of a more socially acceptable score
50
Socially Desirable Response Set – Looking “good”
After coming up with an answer to a question, respondent “screens” the answer– “Will this answer make the person like me less?”
May “edit” their answer
Systematic underreporting of “risk” behavior example– A woman has 2 drinks of alcohol a day, but
responds that she drinks a few times a week
51
Ways to Minimize Socially Desirable Responding
Write items and instructions to increase “acceptability” of an “undesirable” response
Instead of:– “Have you followed your doctor’s
recommendations?” Use:
– “Have you had any of the following problems following your doctor’s recommendations?”
52
Acquiescent Response Set Tendency to
– agree with statements regardless of content– give “positive” response such as yes, true, satisfied
Extent and nature of bias depends on direction of wording of the questions
Minimizing acquiescence:– Include positively- and negatively-worded items in
the same scale
53
Example of Systematic Error Due to Cultural Norms or Beliefs
A person feels sad “most of the time” Unwilling to admit this to the interviewer so
answers “a little of the time”– Not culturally appropriate to admit to negative
feelings– Always present a positive personality
Observed response reflects less sadness than “true” sadness of respondent
54
Discrepancies in Information Sources: Systematic Error or Different Perspectives?
In reporting on a patient’s well-being– Patients report highest levels
– Clinicians report levels in the middle
– Family members report the lowest levels No way to know which is the “true” score
– to say one score is “biased” implies another one is the “true score”
55
Overview
Types of validity in health assessment– Focus on construct validity
How bias affects validity– Socially desirable responding and culture as
sources of bias Sensitivity to change
56
Sensitivity to Change: Two Issues
Measure able to detect true changes One knows how much change is meaningful
on the measure
57
Measure Able to Detect True Change
Sensitive to true differences or changes in the attribute being measured
Sensitive enough to measure differences in outcomes that might be expected given the relative effectiveness of treatments– Ability of a measure to detect change
statistically
58
Importance of Sensitivity
Need to know measure can detect true change if planning to use it as outcome of intervention
Approaches for testing sensitivity are often simultaneous tests of – effectiveness of an intervention
– sensitivity of measures
59
Measuring Sensitivity
Score is stable in those who are not changing Score changes in those who are actually
changing (true change)
One method– Identify groups “known” to change– Compare changes in measure across these groups
60
Sensitivity to Change Evidence for PHQ-9 (Short Screener for Depression)
Classified patients with major depression (DSM-IV criteria) over time as:– Persistent depression– Partial remission– Full remission
Examined PHQ-9 change scores in these “known groups”
Löwe B et al. Med Care, 2004;42:1194-1201
61
Changes in PHQ-9 Scores by Change in Depression at 6 Months
Mean change Effect size
Persistent depression -4.4 -0.9
Partial remission -8.8 -1.8
Full remission -13.0 -2.6
Löwe et al, 2004, p. 1200
62
Considerations in Developing CHAMPS Physical Activity (PA) Questionnaire
Needed outcome measure to detect PA changes due to CHAMPS lifestyle intervention– increase PA levels in everyday life (e.g.,
walking, stretching) in activities of their choice
Existing measures designed to capture younger persons’ PA
Stewart AL et al. Med Sci Sports Exerc, 2001;33:1126-1141.
63
Changes in Measure Resulting from Intervention: Validity Evidence for Others
After CHAMPS intervention detected PA change, others used our results as evidence of “sensitivity to change” – Used in Project ACTIVE because of it’s
sensitivity to change in CHAMPS (S Wilcox et al, Am J Pub Health, 2006;96:1201-1209)
64
Sensitivity to Change: Two Issues
Measure able to detect true changes One knows how much change is meaningful
on the measure
65
Relevant or Meaningful Change
Is the observed change important? To clinician:
– change might influence patient management To patient:
– patient notices change
– amount of change matters
66
“Minimal Important Difference” (MID)
The minimal difference that would result in a change in treatment
The smallest change perceived by patients as beneficial
67
Two Basic Approaches to Estimate MID
Anchor-based methods– Require external criterion of change
Distribution based methods– Statistical indicators of change
68
Anchor-Based Approaches to Estimating MID
Requires longitudinal studies Criteria:
– Clinical endpoints
– Patient-rated global improvement
– Some combination
69
Example of Anchor-Based Approach
Identify a subgroup in a study that has changed by a “minimal” amount– Clinical change
– Patient reported change Change score in a relevant health
measure for this subgroup = MID
70
Locating Groups that Have Changed “Minimally”
Administer a global rating of change (perceived change) by patients – the anchor
Select subset that reported “somewhat better” or “somewhat worse” – change in a relevant health measure for this
subset = MID
71
Two Categories Can Define “Minimal Change” Groups
Since your surgery, how would you rate the amount of change in your physical functioning? – Much worse– Somewhat worse– About the same– Somewhat better– Much better
72
Meaning of Change Depends on Direction of Change
A change for the better may result in a different MID than a change for the worse
May need to evaluate these as separate estimates
73
Example: Mean 2-week Change Score in Symptom Measure by Perceived Change
Mean change
Much better 2.25
A little better 1.41 minimal positive change?
About the same 0.42
A little worse -0.29 minimal negative change?
Much worse -0.10
C Paterson. BMJ, 1996;312:1016-20.
74
Distribution-Based Methods
Ways of expressing the observed change in a standardized metric
Three commonly used:– Effect size (ES)
» Mean change divided by SD at baseline– Standardized response mean (SRM)
» Mean change divided by SD of changes– Responsiveness statistic (RS)
» Mean change divided by SD of change for people who have not changed
75
Mean 4-week Change Score in Four Measures and Responsiveness Statistic
Measure Patients who are “about the same”
Patients who are “a little better”
Responsiveness
score
Symptom 1 0.58 1.64 1.14
Activity 0.46 1.64 1.33
Well-being 0.39 0.68 0.39
C Paterson. BMJ, 1996;312:1016-20.
Note: scores range from 1-7; Higher change scores indicate improvement
76
Multi-item Measures: More Likely to Detect Change
Instrument needs to have sufficient variability to detect change– Multi-item scales: many scale levels
Look for evidence of good variability in sample like yours (at baseline)– Room to improve
77
Effect Size of Changes in Health Due to Treatment for Menstrual Bleeding
Drugs Surgery
Self-rated health item -.18 -.10
Health perceptions scale (5 items)
-.03 -.64
Energy/vitality -.23 -.89
Mental health -.14 -.65
Pain -.12 -.73
C Jenkinson et al. Qual Life Res, 1994;3:317-321.
78
Summary: MID of Measures
MID is based on evidence from multiple studies– Over time, learn whether evidence is strong for a
particular MID MID of a measure in one context may not
generalize to another one– e.g. MID for treatment of pain in cancer may differ
from MID for treatment of back pain
79
Readings as a Resource
Farivar et al.– Issues in measuring MID
Stewart et al – Methods for assessing validity (as developed
for the Medical Outcomes Study) Sechrest
– Classic commentary on validation issues
80
Next Class (Class 5)
Factor analysis with Steve Gregorich
81
Homework
Complete rows 20-26 in matrix for your two measures– Validity, responsiveness and sensitivity to
change, scoring, and costs