Biostatistics and QC Flashcards
Sensitivity
% of sick who test positive
(TP/(total sick)) x 100
Specificity
% of healthy people who test negative
(TN/(total healthy)) x 100
Predictive value (+)
Probability that positive result predicts patient has disease
(TP/(total positives)) x 100
Predictive value (-)
Probability that negative result predicts patient doesn’t have disease
(TN/(total negatives)) x 100
Efficiency
Percent of tests that correctly reflect patient wellness or disease
((TP + TN)/total tests) x 100
Prevalence
Percent population who has disease
(Total sick/total tests) x 100
Proportional Error (PE)
(Slope - 1) x 100
Affects graph more at higher values
Constant Error (CE)
Y-intercept
Affects graph more at low values
Bias
Average of the differences between 2 sets of data (new method & old method)
Random Error (RE)
Error that cannot be predicted
One day of QC could be much higher or lower while all other days are normal
Systematic Error (SE)
SE = Constant Error (CE) + Proportional Error (PE)
Usually seen as trends or shifts in LJ charts
Standard Deviation (SD)
Describes “spread”, only applies to Gaussian spread
SD = sqrt((sum of ((data point - mean)^2) / (n - 1)))
Variance
SD^2
Coefficient of Variation (CV)
Helps compare SD’s of data with different units/different orders of magnitude
CV (%) = (SD/mean) x100
Precision
As little spread of data as possible, small SD
Use F-Test to compare precision of 2 sets of data
Accuracy
Measured values as close to true values as possible
Use paired T-test to compare accuracy of 2 sets of data
Reference Ranges
Expected range of values of an analyte in a healthy population
Reference Range Partitioning Factors
Age, Sex, Ethnicity, Diet, Pregnancy, Blood Group, Tobacco Use, Exercise, Geographic Location, etc
What value of R is acceptable?
Greater than .95
Analytical Measurement Range (AMR)
Smallest and largest values that can be reported from instrument without dilution, make sure smallest amount can be distinguished from a blank
Reference Interval Verification when given manufacturer’s existing reference interval
20-30 healthy individuals, if less than 10% fall outside proposed limit then reference interval is verified
Reference Interval Validation when establishing new interval with a Gaussian distribution
Test at least 40 healthy patients, reference interval is (+/-) 2 SD from mean
Reference Interval Validation for establishing new interval for non-Gaussian distribution
At least 120 healthy patients, discard outliers if necessary, middle 95% of patient values is new reference interval
Lower Limit of Detection (LLD)
Lowest concentration that can be detected
Biological Limit of Detection (BLD)
Lowest concentration that can be detected in blood
Functional Sensitivity (FS)
Lowest concentration that can be accurately measured
What is good criteria for a screening test? (Sensitivity and Specificity)
High Sensitivity (catch all people who are sick) Low Specificity (may catch some false positives)
What is good criteria for a Confirmatory Test? (Sensitivity and Specificity)
Sensitivity good (still want to catch all sick people) Specificity is high (weed out any false positives)
What are some possible reasons for Systematic Error on a LJ chart? What can be done to fix it?
New reagent lot (try a different lot), wrong or deteriorating reagent (replace), out of reagent (replace), incorrect calibration (re-do)
What are some causes of Random Error on an LJ chart? What can be done to fix it?
Erratic volume change of sample/reagents (replace analyzer pipette tips, check for loose component), temperature changes (use heater or AC)
Paired T-test
Compares the accuracy of 2 sets of data, can confirm if new test is as accurate as old one
F-test
Used to compare precision of 2 sets of data, can be used to determine if new test is as precise as old one
Pre-analytic process
Ordering, collection, identification, transportation, separation
Analytic process
Actual testing done in lab (usually less error than Pre- and Post-analytic)
Post-Analytic Process
Reporting, interpretation, action
Centers for Medicare and Medicaid Services (CMS)
CMS is federal agency that enforces CLIA, gives proxy status to others like CAP and JCAHO to inspect and accredit in order to give CLIA certification to receive repayment for Medicare/Medicaid testing
Clinical Laboratory Improvement Amendments (CLIA)
Federal regulations, first passed 1967, updated 1988 and 2003, must have CLIA certification to get repaid by CMS
College of American Pathologists (CAP)
Act as a proxy for CMS, regulate proficiency testing for labs
Joint Commission on Accreditation of Healthcare Organizations (JCAHO)
Act as a proxy for CMS, regulate proficiency testing for laboratory and hospital
CLIA Requirements for Waived Tests
Only requirement is to follow manufacturer’s written instructions
CLIA requirements for Non-Waived tests
Moderate Complexity must show: accuracy, precision, reportable range, reference interval
High Complexity must show: all above, analytical sensitivity (detection limit), analytical specificity (interference and recovery)
JCAHO requirements for Waived tests
Accuracy, precision, reportable range
JCAHO requirements for Non-Waived tests
Moderate Complexity: accuracy, precision, reportable range, reference intervals
High Complexity: all above, analytical sensitivity, analytical specificity (including interferences)
CAP Requirements for Waived tests
None, except verification of reference range if practical (manufacturer’s document)
CAP requirements for Non-Waived tests
Moderate and High Complexity: accuracy, precision, reportable range, reference range, sensitivity, specificity (interferences)
12s
Warning flag, can be systematic or random depending on other data
13s
Automatic reject, can be random or systematic depending on other data
22s
Can be 2 consecutive runs above 2s or 2 simultaneous runs on the same day on the same side of the mean, usually systematic error
R4s
When there is a 4s difference within the same run, across 2 levels of control, always random error
41s
Can be 4 consecutive measurements above 1s on 1 level of control, or 2 consecutive measurements on 2 levels of control, systematic error
10x
Can be 10 consecutive measurements on 1 side of the mean on 1 level of control, or 5 consecutive measurements on the same side of the mean across 2 levels of control, systematic
7T
7 control measurements that trend in one direction (progressively higher or lower)
Equivalent Quality Control (EQC) vs. Individualized Quality Control Plans (IQCP)
IQCP are standards that must be follow/risk factors checked in order to use analyzers internal QC
Otherwise can use EQC which required 2 levels of external QC every day (more expensive)
Rules for CAP proficiency testing
- Test samples as you would any patient sample
- Do not communicate results or share samples with other labs
- If you get a test you normally send out, don’t sent it out
- If you work at 2 labs you may not test the same sample at both labs
- Can’t fail 2 consecutive surveys
- Can’t fail 2 surveys per year
- If you fail, documentation of corrective action
- Records must be kept for 2 years
