INTEGRATIVE CARE 10/8b Principles of Measurement Flashcards
1
Q
questions of measurement
A
- how good is it?
- what can I get out of it?
2
Q
levels of measurements
A
counting
mathematical
3
Q
counting measurements
A
- nominal (names): putting a person in a “bin” (sex, falls risk/not falls risk, race insurance, medications, frequencies, tallies, counts, percentage, mode, bar graph)
- ordinal: bins that have order (MMT, Fim scale for assistance, BMI, amount of comorbidities, all of nominal, central tendency is mode and median, range, bar graphs, pie charts)
4
Q
mathematical measurements
A
- interval: has order and equal intervals (weight, can have a negative measurement, all ordinal, mean, standard dev, can create MDC)
- ratio: has order, interval, and true 0 (TUG, ROM, Goniometry, Heart Rate, VO2max, 6MWT, all can have a negative measurement, all ordinal, mean, standard dev, can create MDC)
5
Q
how do SEM and MDC get calculated
A
- measurement error from reliability coefficients/ICC
- MDC from SEM and CI
6
Q
measurement of error SEM
A
- every measurement has error
- the error can be caused by the patient, the PT the environment, or the instrument
- we use SEM to get a CI around any measurment
7
Q
Confidence interval
A
fuzzball around every measurement that puts a number on all of the errors that could happen
95%CI = X +/- ~2SEM
-Need SEM to get this
8
Q
MDC
A
- From SEM and CI
- continuous (interval or ratio) data and precision (accuracy, resolution)
- Digital: important for rounding
- Analog: measure as precisely as we can to reduce error
- Measure HR (in 60s - precise in the level of bpm; in 30s - precise in the level of 2bpm)
- MDC95 = (CiM95 x SqRt 2) x SEM - MDC is dependent on SEM (based on average variability expected around a measurement)
- var in patient perf
- var in measurement process - specific to ptnt population receiving the measurment
- MCID = smallest chane that would be important to a patient
9
Q
Reliability
A
- a value that quantifies the consistency of a tool
- necessary, but not sufficient for a measure to be valid/accurate
- test-retest reliability assumes that multiple measures of an unvarying characteristic should be equal
10
Q
reliability coefficients
A
ICC = Intraclass Correlation Coefficient
- 0.8 or higher is good, 1 is the best it can be
- not super reliable, should include SEMs, SDs, and MDCs because ICCs can be deceiving
- ICC= (variability between subjects - variability within subjects)/variability between subjects
- want a more variable group to get a higher ICC number
11
Q
1s vs ks and 2 vs 3 in ICC
A
1 (just measured once) vs k (measured multiple tiems and took an average)
2 (generalizable) vs 3 (done with extreme training and intra-rater reliability)
12
Q
Inter rater
A
between raters
13
Q
intra rater
A
a single rater