CC - Quality Management Part 1

Quality management system

Coordinated activities to direct and control an organization with regard to quality (ISO & CLSI). Examines all of the processes, procedures, and components that lead to quality laboratory results.

CLSI and ISO

Clinical and Laboratory Standards Institute and International Organization for Standardization. They produce different standards for clinical laboratories.

Quality system essentials

• Documents and records
• Organization
• Personnel
• Equipment
• Purchasing and inventory
• Process control
• Information management
• Occurrence management
• Assessment
• Process improvement
• Service and satisfaction
• Facilities and safety

Process control

The quality system essential that subdivides into preanalytical, analytical and postanalytical. The analytical branch covers calibration, maintenance, quality control, and PT/EQA.

Statistical quality control loop

Sample the measurement process, then ask whether it is stable. If yes, report patient results. If no, take corrective actions, repeat patients, and return to sampling the measurement process.

Quality assurance

Part of quality management that includes all the planned activities within the pre-analytic, analytic and post-analytic phases of laboratory testing.

Pre-analytic activities

• Test ordering
• Specimen collection
• Transporting the specimen to the laboratory
• Processing the specimen
• Entering patient information
• Centrifuging
• Separating and aliquoting serum/plasma
• Delivery to proper laboratory location

Analytic activities

• Specimen analysis, manual or automated
• Commercial controls
• Record keeping

Post-analytic activities

• Reporting out specimen results, manual entries or computer system
• Physician contact
• Reference ranges

Quality control

Systematic monitoring of analytic processes to detect errors that occur during analysis and to prevent the reporting of incorrect patient test results. Types: internal and external.

Internal quality control

Monitors the overall reliability of laboratory results in terms of accuracy and precision through the use of controls. Performed within the laboratory.

Accuracy

Closeness of result to the true value.

Precision

Ability to obtain the same result when test is repeated on the same sample. Can be determined by measuring the standard deviation and coefficient of variation.

External quality assessment

Proficiency testing. A series of unknown samples are sent to the lab for analysis from the program offering this testing. The samples are analyzed in the same manner as patient specimens, and the results are reported to the proficiency program, which compiles results from all of the participating laboratories.

Purpose of proficiency testing

Identify areas of improvement, aid in troubleshooting, and validate the lab's measurement method, technical training, and allowable error limits for new tests.

Control

• Analyzed only for QC purposes
• Contain a specific amount of the analyte/s
• Tested in the same manner as patient samples
• Liquid or lyophilized (freeze dried)
• Commercially or non-commercially prepared
• Assayed or unassayed
• Available in normal and abnormal ranges

Characteristics of controls

• Same matrix as the specimen being tested
• Should be stable for long periods of time
• Available in sufficient quantities
• With minimal vial to vial variability
• Should have target values that are close to medical decision points

CLIA 88

Mandates that controls must be run at least once every 24 hours, or more frequently if recommended by the manufacturer.

Mean

Measure of center. The average; most commonly used.

Median

Measure of center. The middle of the data after the data have been rank ordered; used with skewed data.

Mode

Measure of center. The most frequently occurring value in a data set; used to describe bimodal data.

Range

Measure of spread. The difference between the highest and lowest data points.

Standard deviation

Measure of spread. Describes the distribution of all data points around the mean; the square root of the variance.

Standard deviation formula

SD = the square root of the sum of (Xi − mean) squared, divided by n − 1.

Coefficient of variation

Measure of spread. Used to monitor precision.

Coefficient of variation formula

CV (%) = SD divided by the mean, multiplied by 100.

Worked SD example

Samples 2, 3, 6, 6, 8. Mean = 25/5 = 5. Deviations −3, −2, 1, 1, 3; squared 9, 4, 1, 1, 9. SS = 24. s squared = 24/(5−1) = 24/4 = 6. s = the square root of 6 = 2.449.

Comparing precision between assays

Precision is compared by coefficient of variation, not by SD alone, because the CV expresses the SD relative to the mean. The lecture's exercise gives Ca mean 2.5 SD 0.3, K mean 4.0 SD 0.25, and Na mean 140 SD 4.0, and asks which assay has the poorest precision.

Gaussian distribution

Measure of shape, named for Carl Friedrich Gauss. • Mean, median, and mode are identical
• Distribution is symmetric (bell curve)
• 68-95-99 rule

68-95-99 rule

In a Gaussian distribution, 68 % of data fall within ±1 SD of the mean, 95 % within ±2 SD, and 99.7 % within ±3 SD.

Inferential statistics

Used to draw conclusions regarding the means or standard deviation of two or more sets of data. Important consideration: shape.

Parametric and nonparametric tests

Gaussian data take parametric tests; skewed data take nonparametric tests.

Levey-Jennings chart

• Most commonly used histogram in quality control
• Used to graph day-to-day control values and to identify an unacceptable analytical run

Analytical run

A set of patient and control samples that are tested together under the same conditions.

Control limit formula

Upper limit = Mean + (SD × n). Lower limit = Mean − (SD × n), where n = 1, 2, or 3.

Worked control limits

For a mean of 100 and an SD of 5: ±1 SD = 95–105, ±2 SD = 90–110, ±3 SD = 85–115.

Random error

Variation in QC results with no pattern. Examples: pipetting errors, mislabeling of samples, temperature fluctuation, improper mixing of sample and reagent.

Systematic error

Influences observations consistently in one direction. Examples: calibration problems, deterioration of reagents, unstable and inadequate reagent blanks, contaminated solutions, failing instrumentation, sample instability, changes in the standard materials, poorly written procedures or inadequate staff training.

Evidence of systematic error

Shift and trend.

Shift

Formed by control values that distribute themselves on one side of the mean for a period of 6 or more consecutive days.

Trend

Formed by control values that continue either to increase or decrease for a period of 6 or more consecutive days. The lecture's diagram labels the two halves upward trend and downward trend.

Outlier

Occasional error that obviously differs significantly from the rest of the results.

Shift against trend

Both run for 6 or more consecutive days and both are evidence of systematic error. A shift sits on one side of the mean; a trend keeps moving in one direction.

Westgard rules

• Developed by James Westgard
• Uses a multiple QC procedure as decision criteria to determine if an analytic run is in control
• Reduces false rejections and maintains high error detection

12s

One control value exceeds ±2SD. Warning rule that initiates testing of control data by other rules.

12s

About 4.5 % of all QC results will fall between ±2SD and ±3SD limits in the absence of analytical error.

13s

One control value exceeds ±3SD. Allows high sensitivity to random error.

22s

Two consecutive control values exceed ±2SD on the same side of the mean. Allows detection of systematic error.

22s

There are 2 applications to this rule: within run, using 2 levels of control, and across run, using 1 control only.

R4s

One control value exceeds +2SD and another exceeds −2SD. Allows detection of random error.

41s

Four consecutive control values exceed ±1SD and are on the same side of the mean. Allows detection of systematic error.

10x

Ten consecutive control values fall on one side of the mean, with no requirement for SD size. Allows detection of systematic error.

Rules detecting random error

13s and R4s.

Rules detecting systematic error

22s, 41s and 10x.

Westgard decision tree

Control data are tested against 12s first. If 12s is not violated the run is in control and is accepted. If 12s is violated, test 22s, 13s, R4s, 41s and 10x; if any is violated the run is out of control and rejected, and if none is violated the run is accepted.

Within run 22s example

On a two-level Levey-Jennings chart, Level 1 runs 9 and 10 both sit above +2 SD and Level 2 runs 9 and 10 both sit below −2 SD, so the 22s rule is violated on both levels.

Out-of-control procedure

Step 1: control material is tested, analyzed in the same manner as patient specimens.
Step 2: the QC result is plotted on the Levey-Jennings chart and compared with the established mean and SD or control limits.

Out-of-control procedure

Step 3: look for unusual results or patterns — a result beyond a control limit, a shift, a trend, or other unusual QC patterns.
Step 4: apply the Westgard rules to determine whether the analytical process is in or out of control.

Out-of-control procedure

If QC is acceptable, meaning no Westgard rejection rule is violated, accept the analytical run. The analytical system is considered acceptable for reporting, assuming other requirements are met, and patient results from this run may be reported.

Out-of-control procedure

If QC violates a Westgard rule, reject the analytical run. Patient results from the affected run should not be reported until the problem is resolved.

Out-of-control procedure

Investigate the possible source of error: reagents, calibration, instrument, control material, technique or operator, and environmental conditions.

Out-of-control procedure

Correct the problem with appropriate corrective action based on the cause, then repeat QC to determine whether the problem has been corrected. Once QC is acceptable, the analytical run is accepted and patient results may be reported.