Fall 17 Midterm Flashcards
Internal Consistency
Consistency of construct across individual items of an outcome measure.
Like is there consistency among the questions in the survey.
Test-Retest
Consistency of a test when given to a person, unchanged in the outcome, on two different occasions.
Intra-rater
Consistency of raters, compared to themselves, on two different occasions.
Inter-rater
Consistency of raters compared to each other.
Internal Consistency Basic Study Design
Conduct the outcome measure on a group of people. Then analysze the intra-subject correlation between items.
Test-Retest Basic Study Design
One person gives test to same people on different days.
Intra-rater Basic Study Design
Several therapist give the test to the same people at different times.
Inter-rater Basic Study Design
Therapist meausre the same participants and their scores are compared.
Internal Consistency Common Statistical Results
Chronbach Alpha - a measure of correllation (ideal is between .7-.9 because if it were one then the questions would be too similar).
Test-Retest Common Statistical Results
ICC or Kappa (closer to 1 is better)
Intra-rater Common Statistical Results
ICC or Kappa (closer to 1 is better)
Inter-rater Common Statistical Results
ICC or Kappa (closer to 1 is better)
Internal Validity Appraisal Considerations
Sample size
Participants have a range of diversity in the outcome measure.
Participants are stable in characteristics of interest.
Same circumstances (consistent manner) for assessment each time.
Time appropriate between measurements (1 day)
Construct Validity
Does the outcome measure measure what it is intended to measure.
Construct Validity must have a theory behind it.
Therefore Convergent, Discriminate and Known groups has a theory behind it.
Criterion: Concurrent
Two measures correlate at the same time point. Did it at the same time.
Criterion: Predictive
The outcome measure of interest is correlated with another outcome measure or outcome at a later point. Therefore it has to deal with Time. Did something change over TIME.
Construct: Convergent
Does the outcome measure of interest correlate with another. Remember that this is very similar to concurrent but this is a criterion so there is a theory behind it.
You have two measures that your expecting to measure the same thing.
Construct: Discriminative
Does the outcome of interest NOT correlate with a measure known to measure a different construct.
Your measuring one thing and then measuring something else that you would expect to have different results to show that it is not measure the same thing.
Construct: Known Groups
Does the outcome measure of interest produce different results for groups of people known to be different on the construct the outcome is supposed to test.
Basically, you KNOW they will be different.
Like teacher taking a test vs students taking a test the teachers haven’t taught yet.
Criterion Basic Study Design
measure the outcome of interest and established outcome measure.
Construct Basic Study Design
Comparing the outcome measure of interest to the established gold standard of reference criterion.
Minimal Detectable Change
The minimum amount of change required on an outcome measure to exceed anticipated measurement error and variability.
Your actually seeing change, and not change due to error.
Responsiveness
An outcome measures ability to detect change over time.
Minimal Clinically Important Difference
The minimum amount of change on an outcome measure which patients are likely to perceive as beneficial.
Minimal Detectable Change Basic Design Study
Conduct a test-retest reliability study and then calculate the MDC
Responsiveness Basic Design Study
Measure a population at two time points, in between which you expect them to change.
Minimal Clinically Important Difference
Measure a group likely to experience change and concurrently measure a gold standard representing meaningful change. Find the cutoff score that best detects meaningful change.
Minimal Detectable Change Common Statistical Results
Points on the outcome measure scale
Responsiveness Common Statistical Results
Effect Size
Minimal Clinically Important Difference Common Statistical Results
Cutoff score and its sensitivity, specificity, likelihood ratio.
Spearman Rho (Correlation -1 to 1)
Criterion Concurrent Criterion Predictive Construct Convergent Construct Discriminitive Criterion Construct
Criterion Common Statistical Results
Spearman Rho (Correlation -1 to 1) OR Pearson correlation Coefficient (-1 to 1)
Known Groups Statistical Results
Analysis of variance for linear trends (p VALUE)
