Glossaries - CLSI EP 23-A and Others
Basic Concepts. Performance-Driven Quality Control. 2001 Zoe Brooks
Samples may come to the laboratory from patients, proficiency testing (PT) programs and external quality assessment schemes (EQAS), or manufacturers of quality control (QC) material. When we test a portion of the same QC material each day, we are measuring the same sample.
Samples are tested on an analytical system to produce results that indicate the amount of analyte present. The analyte is the specific substance we are interested in measuring.
In this book, the term PT sample will also apply to EQAS samples. PT samples are sent to the laboratory from external agencies and tested to reflect the laboratory’s performance with patient samples. Erroneous proficiency results indicate that our laboratory is incapable of meeting the accepted standard of performance and can ultimately lead to the loss of our laboratory’s license to perform that specific test or entire class of tests.
The analytical system includes the reagents, calibrators, instruments, disposables, and step-by-step processes needed to produce a result.
Reagents are chemical solutions that react in predictable ways to given analytes; the predicted reaction allows us to measure the amount of an analyte present in each sample.
Calibrators are materials that contain a known amount of analyte.
Instruments mix samples with reagents and compare the reaction of the patient, PT, or QC sample against the known calibrator values or predictable chemical activity of an analyte to produce a result for each sample.
Disposables include the items required to contain samples, deliver set volumes of samples and/or reagents and measure the chemical reaction.
Processes include all of the steps necessary to prepare samples for testing, prepare reagents and calibrators, set up and maintain instruments, combine samples with reagents, calculate the amount of analyte present, and report results.
Results of QC samples are analyzed to assess, and alert us to, changes in method performance. Because of inevitable minor changes in reagents, calibrators, instruments, disposables, and processes, these QC results show a predictable and expected random variation. Measured results on the same QC sample are not always the same; some are higher and some are lower than others due to random variations in the analytical system.
create a bar chart of our QC results with
the frequency of results on the y-axis and value on the x-axis,
we expect a symmetrical bell-shaped distribution of data.
random variation has a defined mathematical
relationship based on random distribution and is
described as Gaussian
distribution. This predictable pattern of data distribution
is the foundation of statistical analysis in laboratory
When analytical systems are stable, they do not experience significant change, and the QC results exhibit Gaussian distribution with the average value at the center. The dispersion of results around the average is determined by the inherent random variation of the test system. We calculate these values as the mean (average) and standard deviation (variation) to assess method accuracy and precision. An accurate method will produce results with a mean value close to the true or target value for the measured analyte.
The true or target value is the best estimate of the correct value for each control sample. The difference between the measured mean and the target is called bias. Method bias changes from time to time when we change reagents or calibrators, components of the instrument, disposables, or processes.
A precise method will show a relatively small standard deviation, or narrow dispersion, of results around the mean.
If a change occurs in the bias or precision of our analytical system, the pattern of data distribution changes from the expected Gaussian distribution observed prior to the change.
Daily QC encompasses preparing and handling QC samples, testing samples to produce QC results, and assessing individual QC results using QC charts and QC rules. In practice, these activities may be performed several times each day for some analytes, and only once or twice each week or month for others, depending on the frequency of testing, the expected error rate of the method, and the number of patient and PT samples tested.
Stability reflects the inherent error rate in an analytical process; it is a measure of the likelihood that a method will experience sudden changes in accuracy or precision that will adversely affect its ability to meet quality specifications. A method that demonstrates frequent significant changes is described as having a high error rate or low stability.
Each analysis of QC samples brackets an analytical run.
The result(s) of the QC sample(s) in each run reflect the accuracy and precision of the method at that time. Any changes in accuracy or precision will affect the QC samples and alert us to similar changes in patient and PT samples. Results from daily QC are plotted on a QC chart—an x–y plot with results from testing of QC samples on the y-axis, and the run number or date on the x-axis. The scale of the y-axis is usually set with the mean at the center and the minimum and maximum value often but not always defined by +/- 4 SD.
QC flags are events that indicate unexpected performance, usually triggered when QC results fall outside expected limits selected to meet the criteria of QC rules. QC rules are usually defined by the number of occurrences of QC values that vary from the mean by more than a defined number of SD. We select QC rules to maximize QC flags when changes in method accuracy or precision cause patient or PT results to exceed quality specifications, and minimize QC flags in the absence of such change.
Summary statistics, the mean and SD, compare method accuracy and precision to the target value and total error allowable (TEa) for each QC sample to assess overall method performance. QC flags in reports of summary statistics alert us to changes in the overall accuracy or precision of a method.
Quality specifications are usually defined as a total error allowable (TEa); these limits specify the maximum acceptable variation of results from the target value.
A total error flag in summary statistics indicates that total error, the combined effect of method bias and precision, exceeds the acceptable variation defined by the total error allowable.
Critical systematic error (ΔSEc) is a measure of the number of SD the mean can shift before values will exceed the TEa limit. ΔSEc flags indicate that a method is close to its quality specification and requires close monitoring.
Performance-Driven Quality Control helps us design QC systems to ensure that our laboratory meets the defined quality specification for each method.