CDM Reliability (10/8b) [Integrative] Flashcards
Levels of Measurement
Nominal
Ordinal
Interval
Ratio
Nominal Measurement
OVERVIEW
- names, nonnumeric data with no order or intervals
- EX: sex, insurance, names, disease/condition
DESCRIPTIVE STATS
- frequency, tallies, counts, percentage, mode
Ordinal Measurement
OVERVIEW
- has order, no consistent intervals
- EX: numeric pain scale, poor/fair/good/etc.
DESCRIPTIVE STATS
- same as nominal
- central tendency: mode, median
- range
Interval Measurement
OVERVIEW
- has order and equal intervals
- EX: time intervals, weight
DESCRIPTIVE STATS
- same as nominal and ordinal
- central tendency: mean
- variability: standard deviation, MDC
Ratio Measurement
OVERVIEW
- has order and intervals, true zero
- EX: TUG time intervals, range of motion
DESCRIPTIVE STATS
- same as nominal and ordinal
- central tendency: mean
- variability: standard deviation, MDC
Continuous Data Precision
Digital — important to know how much rounding was done
Analog — measure as precisely as we can to reduce rounding error; usually to the nearest ½ unit displayed
Heart Rate — can be really error prone
Basic Relationship Between Measurement and Error
Measured value = true value + error
an estimate of actual value
Types of Error
Patient
Examiner
Environment
Instrument
Standard Error of Measurement (SEM)
estimate of the average variability expected around a measurement, how error is quantified
SEM = s√(1 - rxx )
Use SEM to determine CI around a measurement
95%CI = X ± ~2 * SEM , (X = measurement)
Confidence in Single Measure (CiM)
95% Confidence in measure (CiM 95)
Measurement ±1.96 * SEM
If wanted a 68% confident would use 1 SEM
Minimal Detectable Change (MDC)
MDC provides more meaningful information than just reliability coefficients
MDC95 = CiM95 x √2 MDC95= (1.96 x √2) * SEM MDC95= 2.77 * SEM
MDC Characteristics
SEM (estimate of the average variability expected around a measurement)
Based on:
- Variability in patient performance
- Variability in measurement process
Specific to a patient population receiving the measurement
MCID Characteristics
Smallest change that would be important to patient
Specific to a patient population receiving the measurement
What do you need to calculate SEM?
Variability (s) of measures for a tested group
Reliability (r) coefficient for tested groups
Reliability (r)
A value that quantifies the consistency of a tool, not accuracy
Necessary (but not sufficient) in order for a measure to be considered valid
Test-Retest Reliability
Assumption that multiple measures of an unvarying characteristic should be equal
SEMs and MDCs are ___ reliable than reliability coefficients like ICCs
more
Intraclass Correlation Coefficient (ICC) - Equation
(Variability btwn subjects - Variability wthn subjects) / (Variability btwn subjects)
High Reliability = closer to 1
Low Reliability = closer to 0
How variability affects ICC
To have a higher ICC (closer to 1)— need higher variability between subjects
This is a con about ICCs because they are very dependent on heterogeneity of the group
Higher variability within subjects leads to poor/low ICCs (closer to 0)
Interrater vs Intrarater
Inter-rater = Between different raters, usually less consistent
Intra-rater or Test-Retest = Within a single rater, usually more consistent
Reliability Quantified by
Reliability coefficient, such as ICC
SEM (standard error of measure)
SEM is in the units of the measure
Use SEM to calculate MDC
Validity
does it measure what it says it measures?
Criterion
Construct validity established by comparing the measure to something else
