Quality Control in clinical

laboratory

Kanit Reesukumal, M.D.

Assistant Professor Clinical Pathology
Mahidol University
 Outline
• Internal Quality Control
• Calculation
• Levey-Jennings Charts
• Westgard Rules
• Six sigma
• Additional Quality Control Statistics
• Out of control limit corrective action
• Quality control products
 Introduction
• Achieving quality in the medical laboratory
requires the use of many tools
– Procedure manuals
– Maintenance schedules
– Calibrations
– A quality assurance program
– Training
– Quality control
 Precision vs Accuracy
Internal quality control External quality assessment
• Repetitive assay • Test result compared to
consensus value
• Assess precision • Assess accuracy
 Internal Quality Control
• Internal quality control is a statistical process
used to monitor and evaluate the analytical
process that produces patient results

• The question of reliability for most testing can

be resolved by regular use of quality control
materials and statistical process control
 Regular Testing
• Good laboratory practice requires testing
normal and abnormal controls for each test at
least daily to monitor the analytical process.

• If the test is stable for less than 24 hours or

some change has occurred which could
potentially affect the test stability, controls
should be assayed more frequently
 Comparison of Quality Control Results
to Specific Statistical Limits
Test Potassium Instrument Unit
LV1 control LV2 control Patient Results
Range 3.7-4.3 mmol/L 6.7-7.3 mmol/L
1 May 4.0 7.0 4.2, 3.8, 5.0, 5.8
2 May 4.1 6.9 3.8, 4.4, 3.9, 4.6
3 May 4.2 7.1 4.4, 3.9, 3.7, 4.3
4 May 4.0 7.0 4.2, 4.5, 3.7, 5.6
5 May 4.2 7.1 3.7, 4.5, 3.8, 4.3
6 May 4.1 8.0 2.8, 3.6, 6.0, 7.0

Out of control for abnormal high K on May 6

The range defined for each level of control is fundamental
of the quality control system
 Calculations

• Where:
– s = standard deviation
– x = mean (average) of the QC values
– Σ(xn - x)2 = the sum of the squares of differences
between individual QC values and the mean
– n = the number of values in the data set
Gaussian (bell-shaped) curve

4.5% of value will fall outside 2 SD range

 Levey-Jennings Charts
• Using a Levey-Jennings Chart to Evaluate Run
Quality
• Systematic Error
• Random Error
Levey-Jennings chart
Systematic error
 Trend
• Deterioration of the instrument light source
• Gradual accumulation of debris in
sample/reagent tubing or electrode surfaces
• Aging of reagents
• Gradual deterioration of control materials,
incubation chamber temperature (enzymes
only), light filter integrity, and calibration
 Shift
• Sudden failure or change in the light source
• Change in reagent formulation or reagent lot
• Major instrument maintenance
• Sudden change in incubation temperature
(enzymes only)
• Change in room temperature or humidity
• Failure in the sampling system, reagent dispense
system
• Inaccurate calibration/recalibration
 Random error
• Any deviation away from an expected result
• There is acceptable (or expected) random
error as defined and quantified by standard
deviation.
• There is unacceptable (unexpected) random
error that is any data point outside the
expected population of data (e.g., a data point
outside the ±3s limits).
 Westgard rules
• Dr. James Westgard, 1981
• Based on principles of statistical process control
used in industry nationwide since the 1950s
• Six basic rules
– 12S
– 13S
– 22S
– R4S
– 41S
– 10x
 12S
• Warning rule
• About 4.5% of all results
fall between 2s and 3s
limits
• Acceptable if
– No source of error
identified
 13S
• Unacceptable random
error
• The beginning of a large
systemic error
 22S
• Criteria
– Two consecutive QC
results
– Greater than 2s
– On the same side of the
mean
• Systematic error
– Within run: both levels
– Across run: previous run
 R4S
• Identified random error
only
• Applied only within the
current run
• At least a 4S difference
between control values
within a single run
 41S
• Criteria
– 4 consecutive results
– Greater than 1s
– On the same side of the
mean
• Systematic bias
– Within control material:
single area of the
method curve
– Across control materials:
broader concentration
 10X (or 8X)
• On the same side of the
mean regardless of the
specific standard
deviation in which they
are located
• Systematic bias
– Within control material:
single area of the
method curve
– Across control materials:
broader concentration
https://www.westgard.com/westgard-sigma-rules.htm
Out of control limit corrective action
Additional Quality Control Statistics
• Coefficient of Variation [CV]
• Coefficient of Variation Ratio [CVR]
• Standard Deviation Index [SDI]
 Coefficient of Variation [CV]
• The ratio of standard
deviation to the mean

• Compare precision for CV = (SD/mean) x 100

two different methods
– SD increases as the
concentration of the
analyte increases

• CV = Total allowable error/3

Monthly CV review
 Comparative evaluations
• Instrument manuals and test method
description
• Proficiency surveys
• Interlaboratory QC programs
• CLIA proficiency limits
 Instrument manuals and test method
description
• Published expectations for between-run and
within-run precision
• Reflect ideal condition
• Results higher than specifications may indicate
a possible problem exists
– Should compare to proficiency reports which are
more indicate of “real world” experience
 Proficiency surveys (EQA)
• Receive a set of unknown sample
• Report results to the proficiency agency
• Test result reported by each laboratory is
compared to consensus value
• Laboratory is graded for accuracy
• Can be used to compare and assess day-to-day
laboratory precision
 Interlaboratory QC programs
• Monthly data
• The data are combined with data from other
laboratories which use the same instrument
• Benefit over a proficiency program
– Statistics collected from repeated daily testing
whereas proficiency program provides statistics
collected from single events that occur only 3
times a year
 CLIA proficiency limits

• Published performance limits for commonly

tested analytes in the United States CLIA
regulation
• CLIA proficiency testing criteria for acceptable
analytical performance, as printed in the
Federal Register February 28,
1992;57(40):7002-186.
Coefficient of Variation Ratio [CVR]
• Less than 1.0
– Precision is better than
the peer group
• Greater than 1.0
– Larger imprecision
• Greater than 1.5
– Need to investigate
• Greater than 2.0
– Need for trouble
shooting and corrective
action
 Standard Deviation Index [SDI]
• 1.25 or less
– Acceptable
• 1.25-1-49
– Marginal performance,
some investigation may be
required
• 1.50-1.99
– Marginal performance,
investigation is
recommended
• 2.0 or greater
– Unacceptable
 Quality Control Products
• A quality control product is a patient-like
material ideally made from human serum,
urine or spinal fluid.
• Control product can be a liquid or freeze dried
(lyophilized) material and is composed of one
or more constituents (analytes) of known
concentration.
• Control products should be tested in the same
manner as patient samples.
 Choosing a Quality Control Products

• Selecting a Control Product

• Shelf Life
• Box Pricing
• Clinically Relevant Decision Levels
• Interlaboratory Comparison Programs
 Third party control
• A control that has not been designed for a
particular test system or analyzer
– An independent control
– Manufacturer independently of reagents and
calibrators
– Provides an unbiased assessment of performance
 Selecting a Control Product
• Cheaper products
– Short shelf life after opening
– Not sufficient similar to patient specimens (serum,
urine, spinal fluid or plasma)
– Not have all analytes at medically relevant
decision levels
– Misled by box pricing
 Shelf Life
• It is necessary to know the approximate
volume of control to be used each day

• Shelf life becomes an important issue when

low volume of control is used
– Waste
 Clinically Relevant Decision Levels
• Some analytes need 3 level of quality control
– Low level
– Normal level
– High abnormal level
• Example: TSH with AMR 0.1-50 µIU/mL
TSH (µIU/mL) Vendor 1 Vendor 2
Low level 1.03-1.23 3.0-5.0

Normal 7.5-9.6 8.0-10.0

High abnormal level 27.9-34.5 45-55
 Box Pricing
• Different volume and vials

• Always ask for quality control product quotes

on a per mL basis and not box price
Interlaboratory Comparison Programs

• Highly recommend

• Easy method to assess reliability and

imprecision
– Compare the within-laboratory method means
and standard deviations to other laboratories
using the same instrument and method (peer
group)
 Conclusion
• QC: monitoring reliability of the test results
• IQC: precision
• EQA: accuracy
• Systematic error: 22s, 41s, 10x
– Trend
– Shift
• Random error: 13s, R4s
• Coefficient of variation: compare precision for two
different methods
• Peer group: CVR and SDI
• Good QC materials: matrix, shelf life, price per mL,
clinical relevant, interlab QC