Method comparison & Evaluation Flashcards
1
Q
3 goals of lab dx.stics
A
- correct/appropriate test
- At the right time
- Performed accuratley
2
Q
whys is method comparison & evaluation important?
A
correct & timely dx for imporved/appropriate patient care
3
Q
when/hwat reasons would you evauluate/compare new methods
A
- current method unsatisfactory: in analytical (efficientcy); cost; reliability of suppliers/tech support; Safety
- new instrument
- new disease marker
- lab reorganisation
4
Q
what to consider during evaluation
A
- Analytical performance: which aspects; w/in run/ b/w run (if run on same day = get same results?)
- staffing
- sample vol
- cost
5
Q
important principles to successful evaluation of new method
A
- consider clinical PERSPECTIVE:
- Consider anlystical GOALS required
- Select analytical METHOD
- Conduct experiments/EVALUATION: compare w/ existing method
- Use stat.al tools & methodology to estimate diff b/w methods
6
Q
how to compare results b/w 2 methods?
A
- VISUAL: scatterplot, difference plot
- STAT.AL approaches: test diff. b/w means, regression analysis, correlation coeff. (R^2)
- CONTEXT: understand context when interpreting results
7
Q
2 types of total errors (innaccuracy)
A
- random error (imprecision)
- Systematic error (bias)
8
Q
4 cuases of systematic error
A
- improperly prepped reagents
- deterioration of reagents or calibrators
- inappropriate storage of reagents / calibrators
- variation in procedure b/w technologists
9
Q
4 Causes of random error
A
- air bubbles in reagent
- improperly mixed reagents
- clogged or imprecise pipetter
- improperly fitting pipette tips
10
Q
what is regression
A
- aka std error of estimate
- measures variability of observed values on regression line
- you want to minimise the distance b/w line & dots
11
Q
2 regression methods
A
- Deming regression: (not detect errors to method on X-axis)
- Passing-Bablock: less sensitive to outliers
12
Q
why is measurment uncertainty important?
A
- so we can tell if there’s been a significant change
13
Q
overall uncertainty (OU) equation
A
OU = √(MU^2 + BV^2) MU^2 = variance of measurement uncertainty BV^2 = variance of biological variation
14
Q
what is bland-altman analyses
A
- difference plot
- w/ difference on the y-axis
15
Q
T-test
A
- test for diff b/w means
- paired & unpaired t-test
- summarise overall bias (usually proportional)
