QC Definitions Flashcards
% Error is used when looking for the…
Reportable Range
F test - does it measure accuracy or precision?
Precision
- compares 2 data sets
- small SD (small spread of data)
- confirm if new test is as accurate as old one
Paired T test - does it measure accuracy or precision? What type of error does it measure?
Accuracy
- compares 2 data sets
- confirm if new test is as accurate as old one
Systematic Error (SE)
A normal/Gaussian/parametric distribution is when these three values are similar
Mean
Median
Mode
How are reference ranges (interval) established?
Mean +/- 2SD
How do we find the area under a normal curve at a particular Z score?
Look at Z score table
- area from negative infinity to less than that z score = whatever value the z score is
- area above z score = 1 - z score value
Proportional Error - affects graphs more at low or high values?
High
Constant Error - affects graphs more at low or high values?
Low
Define Random Error (RE)
Error that cannot be predicted
Define Systematic Error (SE)
Constant Error (CE) + Proportional Error (PE)
- CE = Y-intercept
- PE = (slope - 1) * 100
- bias is also a SE (combine effects of CE and PE)
Shifts and trends in a Levey-Jennings chart are what type of error?
Systematic Error (SE)
Define Reference Range
Expected range of values of an analyte in a healthy population
Factors that contribute to varying Reference Ranges (Reference Range Partitioning Factors)
Age, sex, ethnicity, diet, pregnancy, blood type, tabacco use, exercise, geographic location, etc
Another name for Analytical Measurement Range (AMR). Define AMR
Reportable Range
Smallest and largest values that can be reported from instrument without dilution
-smallest amount must be distinguishable from blank
When would 20-30 healthy individuals be sufficient for establishing Reference Interval validation? How is the RI verified?
Adopt a manufacturer’s existing reference interval
2 or fewer (<10%) values fall outside the proposed limit
-lab must make own RI if rejected
When would 40 healthy individuals be sufficient for establishing Reference Interval validation?
Establishing a new Reference Interval with a Gaussian/normal distribution
-for a particular analyte in a population
When would 120 healthy individuals be sufficient for establishing Reference Interval validation?
Establishing new Reference Interval with a non-Gaussian/normal distribution
- discard outliers
- choose middle 95% of values is the new reference interval
Define Lower Limit of Detection (LLD)
Lowest concentration that can be measured
Define Biological Limit of Detection (BLD)
Lowest concentration that can be measured in the blood
-always higher than LLD (other stuff in blood)
Define Functional Sensitivity (FS)
Lowest concentration that can be reliably/accurately measured
What are good criteria for screening tests?
High sensitivity - catch all sick people
Low specificity - catch false positives
What are good criteria for confirmatory tests?
Good sensitivity - catch all sick people
High specificity - weed out false positives
Causes of Systematic Error (SE) (4)
New reagent lot number
Wrong or deteriorating reagent
Empty reagent
Incorrect calibration
Causes of Random Error (RE) (2)
Erratic volume change of sample or reagent (check for loose components)
Temperature changes
Pre-analytic process involves… (5)
Ordering Identification Collection Transportation Separation -STOIC
Analytic process involves…
Lab testing
-usually not a source of error
Post-analytic process involves… (3)
Reporting
Interpreting
Action
F test vs T test - which one measures if there’s significant differences between SDs? Which one measures if there’s significant differences between the means?
F test = SD
T test = mean
What does an r > 0.95 signify? What is another name for r?
Good correlation between the two methods
1.0 is best correlation
Correlation coefficient
Bias is a measure of what type of error?
Systematic Error (SE) -bias is combined effect of both PE and CE
Diff SD is a measure of what type of error?
Random Error (RE)
Define CV. Is a low or high CV better?
Measure of dispersion of data points around the mean
Low CV is better = less dispersion
CV = (SD/mean) * 100
Define false reject and false accept
False reject - QC system shows a flag, but no significant change (12S, 13S)
-ideally < 5% of the time
False accept - significant change present, but QC system fails to flag it.
-very problematic, incorrect results
Imprecision, shift, trend - which is associated with SE and which is RE?
SE = shift or trend
RE = imprecision
-increasing large jumps across large SDs
What is the number 1 issue that causes SE?
Wrong or deteriorating reagent
Define ANALYTICAL sensitivity and ANALYTICAL specificity. What type of test are they normally associated with?
Analytical sensitivity - lowest concentration of detection
-includes LLD, BLD, FS
Analytical specificity - things that can interfere with your test
Proficiency Testing: CLIA, JCAHO, CAP - high complexity testing
When designing an in-house test, what complexity level is it considered?
High complexity
Accuracy vs precision experiments
Accuracy - method comparison with gold standard
Precision - replication experiment
- same unknown
- calculate SD, CV
- compare controls
Reportable Range - ideally, plotting the measured value against the assigned value should give a graph that looks… Should reportable range be smaller or larger than reference interval?
Linear
-dilute samples if too concentration
Larger - catch patients with really high or low values
Linear regression - a perfect slope is at a ___ degree angle
45 degree (slope = 1) slope = 1 means there's no PE
The Y-intercept is also known as this type of error…
Constant Error (CE) -ideally 0
The r value is a measure of what type of error?
Random Error (RE)
RE and SE - which one is inaccuracy and which one is imprecision?
RE = imprecision
SE = inaccuracy