54 Revista Română de Medicină de Laborator Vol. 8, Nr.

3, Septembrie 2007

Quality Control for Chemistry Laboratory – Dynacare

Kasper Laboratories Procedures
Lorena Ivona Stefan1*, Deborah Reid2
1. Emergency Clinical Hospital, Laboratory Medical Analysis, Craiova, Romania
2. Dynacare Kasper Laboratories, Clinical Chemistry Department, Edmonton, Canada

Abstract
This paper presents the quality control procedure (internal quality control, Westagard rules and exter-
nal quality control) for chemistry in Dynacare Kasper Medical Laboratory (DKML) from Edmonton, Canada
and provide a practical approach to quality control.
Keywords: quality control, clinical chemistry, medical laboratory.

Goals for a quality control program precision and accuracy demanded by Clinical
Laboratory Improvement Amendment of 1988
The first step in establishing a laborato- (CLIA’88) regulations. Second are the preci-
ry quality control program is to develop criteria sion and accuracy that appear to be attainable
for acceptable laboratory performance. How ac- performance by most laboratories. This infor-
curate and precise should the laboratory be? mation can be obtained by communication with
How precise and accurate must it be?. These other laboratory professionals or from data de-
considerations include the determination of rived from proficiency surveys, such as that of
what constitutes acceptable analytical error the College of American Pathology (CAP).
based on the use of the test result in clinical Third and probably most important, it is essen-
care. Control beyond that required for medical tial to determine the precision and accuracy re-
purposes can waste time and materials; hence it quired by the clinical staff, the users of data
is important to evaluate whether error reduction produced by the laboratory. In general, a testing
improves medical diagnosis, treatment or prog- system’s analytical error should be much small-
nosis. er than the allowable error in the regulatory re-
Several bases exist upon which perfor- quirements. Otherwise, the laboratory may not
mance criteria can fe formulated. The first is the meet its regulatory and perhaps medical re-
body of regulatory standards; for example, the quirements.
*
Corresponding author: Lorena Ivona Stefan. Home address: Cart. CraioviŃa Nouă bl.8, Ap.32, Craiova, Dolj,
1100, Romania. Phone: +4 0726/314845; E-mail: [email protected] or [email protected].
Office: Spitalul Clinic de UrgenŃă Craiova, Sectia Laborator Analize Medicale, Bulevardul 1 Mai, Nr.60, Craio-
va, Dolj 1100, Romania.
 Revista Română de Medicină de Laborator Vol. 8, Nr. 3, Septembrie 2007 55

Medical Decision Limits changes in test values. One key element in in-
terpreting the medical usefulness of a test result
For true control of quality it is neces- is an estimate of the magnitude of an analytical-
sary to evaluate, from the customer’s perspec- ly significant change in concentration. This esti-
tive, the performance required for each aspect mate is called the significant change limit
of the clinical laboratory’s operation. The ele- (SCL).
ments of a good quality control program in- The significant change limit is a deci-
clude establishment of analytical accuracy and sion – making tool that helps physicians distin-
precision performance criteria based on medical guish day-to-day changes in results that are
usefulness requirements. Each laboratory caused by the inherent variability of the ana-
should consult the appropiate users or clinicians lytical procedure from changes that are caused
to obtain their estimate of allowable error based by modifications in the patient’s physiology
on their particular medical practice. If their sug- and pathology. The significant change limit is
gestions are reasonable and would not place the based on the assumption that the usual stan-
laboratory conflict with regulatory require- dard deviation (USD) represents day-to-day
ments, the laboratory should try to attain these method variability. As an approximation, the
limits of error. If no information is available significant change limit is three times the usual
about the precision and accuracy targets needed standard deviation. Changes greater than the
for medical decision making, one can then esti- significant change limit are likely to represent a
mate a theoretical error based on the degree of real change in the patient.
intra-individual and inter-individual variation
for each analyte. Control of quality (process control)
Reference intervals for laboratory tests and error detection
describe the expected values for carefully se-
lected groups of individuals determined by test- Once a laboratory’s performance crite-
ing systems that are assumed to be performing ria are established, a process control system
appropriately. Increased bias will cause a shift must be put into place. The purpose of this sys-
in test values and will thus invalidate the medi- tem is to allow continuous monitoring of the
cal usefulness of the established reference inter- testing process (including preanalytical and
vals and may in fact lead to inappropriate pa- postanalytical testing) to ensure that either the
tient care. performance goals are met or that steps are tak-
en to achieve the goals. It is important to recog-
Meeting Medical Usefulness Criteria nize the key role of laboratory personnel in the
by Calculating the Significant Change Limit quality process.
The day-to-day medical usefulness of Levels of Activity in the Control Pro-
clinical laboratory tests depends on maintaining cess
the accuracy and precision of the testing sys-
tem. Physicians make many clinical decisions The control process that we call quality
on the basis of the day-to-day differences in pa- control (QC) is designed to detect error in the
tient test values, assuming that the day-to-day measurement system. There are at least three
accuracy and precision are maintained at the level in this process, each the responsibility of
same level from month to month and year to different individuals. For the control process to
year. Thus the actual accuracy and precision of be most effective, active communication among
the measurement procedure directly influence the individuals within each level of responsibil-
the medical interpretation of these day-to-day ity is crucial.
The first level of the process is the re-
56 Revista Română de Medicină de Laborator Vol. 8, Nr. 3, Septembrie 2007

sponsibility of the bench medical technologist to determine the actual level of each constituent
and the supervisors. At this level, the control using the procedures employed for routine anal-
process includes the daily analysis of quality ysis. The laboratory can estimate the target val-
control specimens (discussed in this section) ues of the control samples by repeated analysis
and the review and verification of patient re- (the „true values” being estimated as the mean),
sults (results verification) and reports. The tech- use the manufacturer’s estimates of the values,
nologist is responsible for performing quality or ideally, determine the values by definitive or
control analyses at the appropriate intervals and reference methods. The frequency of analysis of
for determining that, during any given run, the QC material is established by each laborato-
there is no significant systematic error in that ry for each method. CLIA’88 requires the anal-
run. Both the technologists and the supervisor ysis of at least two controls of different values
are responsible for reviewing patient data to en- for each run (defined as up to 24 hours of sta-
sure that no random error exists. ble operation) as do other accrediting bodies
The second level of control ensures that with deemed status from CLIA.
minimal systematic bias enters into the system Most laboratories use two different
over a relatively short period of weeks to pools, one normal and one abnormal. A normal
months. The responsibility for this level of error pool contains constituents at concentrations
control is usually shared by supervisors and the within the non-diseased reference interval,
laboratory director, though technologists often whereas an abnormal pool contains the analytes
contribute greatly. The control process at this at concentrations outside the reference interval.
level requires timely review of the quality con- Some laboratories may employ three pools –
trol data and proficiency testing that have accu- low abnormal, normal and high abnormal – es-
mulated over that period of time. pecially when medically significant decisions
The third level of the control process are made at each level. CLIA allows each labo-
ensures that the analytical systems are as pre- ratory to set its own protocols for chemistry
cise and accurate as possible. This is responsi- testing assay control samples as long as at least
bility of the laboratory director or technical two control samples of different concentrations
consultant. are assayed every 24 hours. Some states man-
The control process at this level re- date three pools for certain tests. CLIA man-
quires review of proficiency testing results, dates special rules for blood-gases, requiring as
knowledge of the levels of precision and accu- a minimum the analysis of one QC sample ev-
racy achievable by other laboratories, and, ery 8 hours of testing and the use of combina-
when applicable, the use of accuracy-based tion of QC samples and calibrators that includes
standards to verify or correct errors. This level samples with both high and low concentration
of quality control review occurs over a longer each day of testing.
period of time, from months to years. The Clinical Laboratory Improvement
Amendment also requires the use of one cali-
Testing Quality Control Specimens – brator or control each time a patient sample is
Daily Decision Making analyzed, unless the blood-gas instrument is
The daily preparation and analysis of calibrated at least every 30 minutes. Because of
quality control samples is a regular responsibili- this complexity of blood-gas quality control,
ty of the analyst. The quality control pools are some manufacturers have included QC reagents
analyzed as „known” controls during analysis as part of the reagents resident on blood-gas an-
of patient samples. The values are considered alyzers and have thus assumed a more active
„known” because some attempt has been made role in the QC process. More commonly, how-
 Revista Română de Medicină de Laborator Vol. 8, Nr. 3, Septembrie 2007 57

ever, the manufacturer of a testing system rec- cluded in the analytical run.
ommends the testing frequency that should be Because the quality control material is
used as a basis for a laboratory’s quality control analyzed in every run along with patients’ spec-
policy. imens, large amounts of control material are
Testing personnel must use the data needed each year. Several sources currently ex-
from each quality control analysis to make a de- ist from which a laboratory can obtain sufficient
cision about the validity of patients’ test data. quantities of quality control material: (1) com-
Generally, if the results for a quality control mercial lyophilized pool material; (2) commer-
sample are within the accepted target range, cial stabilized liquid pools; and (3) frozen,
technologists may assume that the patients’ re- pooled, patient specimens. Patient serum is
sults obtained during the same run are equally more frequently used than plasma because it is
valid and can „accept” the run. more readily available and is less likely to in-
On the other hand, if the results for the clude precipitated material. Frozen liquid or
quality control pool are unacceptable, the run is pools that have been clarified (with material
not acceptable. The decision to accept or reject that reduce turbidity) generally show smaller
an analytical run should be documented and in- standard deviation than do lyophilized pools.
clude the decision (either accept or reject), the The smaller imprecision errors of the liquid
analyst’s name (or code number), and the date pools derive, in part, from the absence of the er-
on a work sheet, in a separate log book, on a rors involved with the lyophilization and recon-
data sheet, or in the laboratory information sys- stitution processes. However, the liquid pools
tem (LIS). Usually, the process of verification may experience greater instability errors associ-
of patient data in the LIS by technologists is re- ated with shipping batches of a lot to the cus-
garded as implied acceptance of the associated tomer. It is important to select a pool with a ma-
quality control data included in the run. Al- trix that interacts least with the methods em-
though the term run implies a batch process, ployed in the laboratory. Notice that control
current laboratory practice usually has the mea- pools prepared in the laboratory from pooled
surements continuously performed in real-time patient samples (serum, plasma, urine and CSF)
on automated analyzers. That is, the run is more can be contaminated with viruses; thus it is es-
generally associated with the 8-hour work shift. sential to test each specimen or group of speci-
Although daily bench-level quality con- mens and the final pool for harmful viruses.
trol testing is most useful for detecting system- Therefore, the following statements apply to all
atic errors, it can also be used to detect increas- specimen pools used for quality control. First,
es in imprecision. However, random errors, all pooled human material should be monitored
which occur unpredictably, are not usually de- for the human immunodeficiency virus and the
tectable by a quality control system. Random hepatitis B virus. No pools should be used if
errors can be detected only by review of report- there is evidence of either virus. Second, all
ed problems and patients’ results. control material requires refrigerator or freezer
space for storage of a 1- to 2- year supply. Al-
How to choose a quality control pool ternatively, commercial distributors may supply
Quality control material should have a quantities from a single lot number of stored
matrix that closely matches that of the speci- material on a monthly or quarterly basis so that
mens in the analytical run. This means that if the laboratory can use the same lot number over
the run includes cerebrospinal fluid, serum and 1 to 2 years. This helps bring long-term stability
urine, then controls composed of cerebrospinal to the quality control process, though the possi-
fluid (CSF), serum, and urine should also be in- bility of shipment-to-shipment variations within
58 Revista Română de Medicină de Laborator Vol. 8, Nr. 3, Septembrie 2007

the lot must be considered. significant change limits, external quality con-
Some professional groups and manu- trol and accuracy comparisons, and quality
facturers offer participation in regional quality control performance.
control programs in which laboratories use the Three approaches can be used to estab-
same batch of pooled serum. This offers both lish the limits of acceptable values for a control
scientific advantages and cost benefits. The pool. One method is to use the medically values
comparison between laboratories can help pre- for a control pool. One method is to use the
dict how similar testing systems (peer groups) medically allowable error for choosing the
will perform in proficiency testing. This com- range. Another, more usual, approach is to esti-
parison becomes more valuable when the accu- mate the target value and usual standard devia-
racy of the quality control pool is established by tion (SD) for the method and use some number
reference or definitive methods. of SDs to establish the range. The third tech-
nique is to employ the more statistically accu-
Preliminary Considerations for Esti- rate method of power curves.
mating Limits for Quality Control Pools
Setting Quality Control Limits by
Unless the true value of a pool is estab- Power Curves
lished by definitive or reference methods, the
target values are only averages of repeated mea- The design of specific control rules for
surements of the pool. The average temporary a laboratory requires a five-steps process that
or average final target values of the quality con- includes (1) defining total allowable analytical
trol pool are the estimated concentrations of error, (2) estimating the method’s actual stan-
each analyte within the pool. Each laboratory dard deviation and bias at the medical decision
usually establishes its own average target val- concentrations, (3) determining the systematic
ues for the analytes by performing the laborato- and random error that must be detected by the
ry’s test procedures on each pool. CLIA’88 al- control system, (4) determining the probability
lows the pool’s manufacturer to establish target level used for error detection (i.e., do you want
values, with the laboratory confirming that each to detect 90%, 95% or 99% of errors?) and (5)
target value is applicable to its testing system. plotting and inspecting the power curves to de-
When new target values are established termine the number of control specimens that
for a new lot of quality control material, it is should be tested per run. In general, the most
important to be sure that, during the data collec- difficult part of these evaluations is determining
tion period, the analytical systems perform ac- how much error is allowable.
cording to normal performance specifications. Westgard (www.westgard.com) used
The new lot of quality control material should these power curves to develop a series of spe-
be tested in parallel with the current lot of qual- cific control guidelines, popularly called the
ity control material. If the analytical data from „Wesrgard rules”. The rules, which are used to
the current quality control material indicate sat- determine whether an analytical run is out of
isfactory performance of the methods, the data control, are written in shorthand as follows: (1)
for the new lot can be assumed to be valid. 12S, 12.5S, and 13S mean one control value ex-
When a quality control system is being set up ceeding two, two and one half, or three standard
for the first time, the current methodology is ac- deviations, (2) 22S means two control values ex-
cepted as valid if the method meets perfor- ceeding two standard deviations, and (3) R4S
mance specifications. The choice of the labora- means the range of two control specimens ex-
tory’s testing method (or testing system) is ceeds four standard deviations. For many test-
based on experience with medical usefulness, ing situations the sequential application of the
 Revista Română de Medicină de Laborator Vol. 8, Nr. 3, Septembrie 2007 59

13S/22S/R4S set of control rules allows two con- Technologists must be directed to document
trol specimens to give sufficient error detection their response to every control value that ex-
for a single run. These rules mean that the run is ceeds the established limits.
rejected if any of the following happen: (1) 13S, 2. A method is determined to have an
if one control value differs by more than three inappropriate reference interval; if the range is
standard deviations from the mean value, (2) not immediately correctable, the method is „out
22S, if two control values differ by more than of control”.
two standard deviations from the mean value, 3. A method demonstrates unacceptable
and (3) R4S, if the range between two controls in imprecision, nonlinearity or interferences. Inter-
the same run exceeds a combination of four ferences usually are limited to specific speci-
standards deviations. The first two rules will men types or substances.
detect excessive bias, whereas the last rejects 4. The laboratory director, section di-
the run because of excessive imprecision. rector, or technical supervisor declares the
Notice that, for rejection, the control method out of control for other reasons.
value should exceed the control limit, not just
be equal to that value. For many chemistry Detection of Quality Problems
tests, power curves allow cost-effective detec- Computer assistance. The target val-
tion of significant total errors (based on clinical ues and limits for acceptable results that are es-
usefulness) when two controls are used and the tablished for each control pool are used in daily
limits are set somewhere between 2.5 to 3.5 practice to detect analytical problems. A control
standard deviations. For this reason, many use result can be reviewed in a variety of ways by a
3.0 usual standard deviations as a generalized technologist to evaluate acceptability. The tech-
control limit. For best implementation of West- nologist can simply compare the result with the
gard rules, the laboratory information system posted range. This limits the technologist’s abil-
(LIS) or the analyzer must have the proper soft- ity to employ the Westgard rules or to evaluate
ware present to support this level of quality the trend of previous results. More complex se-
control checks. A more sophisticated approach lection rules are now available as part of some
to quality control will minimize run rejection computer programs, either on the instrument or
and, at the same time, ensure the quality of pa- as part of the laboratory’s information system
tient results. (LIS). Computer assistance allows real-time re-
view of control results, early detection of QC
Detection and resolution of quality prob- problems, and better documentation of the qual-
lems ity control process.
Levey-Jennings plot. Current quality
The out – of – control decision control data are best interpreted in the context
of previous QC results as described above. In
A testing system is designated as „out order to facilitate this goal, the data obtained
of control” when the validity of the results is from daily analysis of quality control pools can
not considered to be appropriate. be plotted to give a visual presentation of the
The conditions for an out-of-control de- data. The most common visual analysis is the
termination should be set by each laboratory; as Levey-Jennings plot. Levey-Jennings plots are
a minimum, the criteria for an out-of-control usually available on the LIS or on the instru-
decision include the following elements: ment performing the assays, obviating the need
1. Control values exceed predetermined to plot these QC results manually. Levey-Jen-
out-of-control limits within a specified period. nings plots should be routinely evaluated by
60 Revista Română de Medicină de Laborator Vol. 8, Nr. 3, Septembrie 2007

technologists and supervisory personnel look- the mean value of quality control pools or pa-
ing for trends or shifts in the data that could in- tients test results. Some testing systems do not
dicate problems in the testing system. allow calibrators (especially calibrators from
other system) to be run as an unknown because
of matrix mismatch. Often a calibrator will
Calibration and quality control
have assigned values that don’t represent actual
analyte values. These assigned-value calibrators
Controls may not be used as calibrators.
are designed to calibrate testing system to pro-
Controls and calibrators must be different be-
duce accurate test values when patients samples
cause each has a separate and important func-
are used.
tion. Calibrators set the reported values accu-
A laboratory that wishes to change a
rately, whereas controls verify the stability and
manufacturer’s calibration set point must docu-
accuracy of the calibration and the testing sys-
ment that the change does not adversely affect
tem. However, for those tests that do not have
the method’s performance specifications. Some
suitable controls available, CLIA’88 allows cal-
of the newer analyzers employ a two-dimen-
ibration materials to be used as controls. For
sional bar code with specific calibration associ-
evaluation of the system’s stability, in these
ated with the particular lot of the reagents. In
cases it is best to find calibrator materials other
this case, the manufacturer provides one or two
than those used for calibration of the testing
point „adjusters” or calibrator-like reagents to
system.
refine the calibration curve prior to placing the
A commercially available calibrator has
reagents into use. Of course, controls are subse-
an assigned value that the manufacturer estab-
quently run to verify the accuracy of such facto-
lishes by using a definitive or reference method
ry calibrations.
or by using reference materials (traceable to
primary standards).The calibrator is then used Quality Control of Reagents Changes
to set the value reported by the laboratory’s and Instrument Maintenance
method or instrument. This process establishes
correspondence of the instrument output signal Each lot of reagent or separate ship-
with known concentrations. Differences be- ment must be evaluated for quality before it is
tween an aqueous and serum matrix can affect put into use. The laboratory can show that new
the transfer of known concentrations to a re- lots or shipments of reagents (including calibra-
ported patient result. These matrix differences tors and quality control pools) are acceptable if,
include turbidity, surface tension, which can af- after their use, the control values do not change
fect sample pipetting, interactions between ana- significantly. It is also a good practice, after any
lytes and proteins, and the effect of the volume maintenance is performed, to test a set of con-
fraction occupied by protein or the other large trols and run several patient samples from a
molecules (especially lipoproteins) on the actu- previous batch before testing is resumed. Main-
al concentration of the analytes. tenance problems can lead to an „action-limits”
Calibrators are usually purchased in situation because operating parameters may be
lots large enough to last 12 or more months. It changed. A chronological record of all reagent
is recommended that a new lot of calibrator ma- changes, instrument repaires, and maintenance
terial be tested 6 weeks before it is used. This procedures along with any calibration verifica-
time delay allows the laboratory to detect any tion tests must be kept.
systematic bias between the values of the cur-
rent and the new calibrator. Bias in a new lot of
calibrator is detected when changes are seen in
 Revista Română de Medicină de Laborator Vol. 8, Nr. 3, Septembrie 2007 61

External quality control programs to the mean of all participants doesn’t establish
accuracy. These comparisons show bias only
Accuracy Control Is Required by when the comparison value is the true value.
CLIA’88 Certainly repeated bias on proficiency tests
must raise the suspicion of a true bias and will
CLIA’88 requires that all laboratories require that additional steps be taken to either
holding a certificate that allows testing of mod- prove or disprove a real bias.
erately or highly complex tests must participate
successfully in proficiency testing. Proficiency
testing (PT) specimens are used to evaluate the Quality Control Procedure for Chem-
adequacy of laboratory performance in all labo- istry – Dynacare Kasper Medical Labo-
ratory specialties. The analyst must test these ratories
specimens in the same manner as patients’
specimens. Historically, PT has been part of a Dynacare Kasper Medical Laboratories
volunteer peer review and educational process. is a full-service laboratory covering all labora-
Proficiency testing is now regulatory, and fail- tory medicine and is accredited by the Ameri-
ure on PT has serious penalties. However. the can College of Pathologists (CAP) and the Al-
value of proficiency testing is the provision of berta College of Physicians and Surgeons. The
independent validation of the internal quality chemistry department is highly automated with
control programs. Because the analyst does not many analyzers, some of which are attached to
know the target value of the PT sample, it is an automated sample delivery system.
difficult for the operator to influence the results.
These programs, if properly used, can give an Internal Quality Control
estimation of the inherent accuracy of a system, 1. All controls received are:
at least as compared with a peer group or to the a. labeled as to date of receipt,
overall mean. b. logged into control inventory log book;
Continued or significant deviations c. stored as per manufacturer’s specifica-
from the PT target levels, even if there is no tion.
failure, should alert the laboratory to a possible 2. Reconstitute lyophilized controls as fol-
accuracy problem. If a method’s USD is not lows:
significantly smaller than the comparative a. consult manufacturer’s specifications
group’s SD, that method is at increased risk for and add the required amount of Type I
PT failure. reagent grade water using a Class A volu-
An estimation of a system’s bias can metric pipette,
also be made from proficiency testing perfor- b. label the vial with the date and time (if
mance. To do this, evaluate the specific test time sensitive) of reconstitution, and initials
method’s observed values against a comparison of person reconstituting the control,
value, which is either the mean value reported c. allow reconstituted controls to equili-
for all similar methods (peer group mean), the brate for the time stated by the manufactur-
mean value for all methods, or the definitive er,
method value. Bias is calculated by subtracting d. visually check control before use to en-
the comparison value from your method’s val- sure that controls are dissolved. Check for
ue. The algebraic sign shows whether your clarity and that contamination or turbidity
method’s value is higher (positive bias) or low- are not present.
er (negative bias) than the group mean. Notice 3. Store controls as per manufacturer’s rec-
that comparison to a peer group mean or even
62 Revista Română de Medicină de Laborator Vol. 8, Nr. 3, Septembrie 2007

ommendations when not in use. Discard con- a. Reagent lot number changes
trols according to manufacturer’s stability. b. Instrument changes
4. Quality control ranges for new unassayed c. All other changes that may affect the
controls are established by parallel analysis results obtained.
with current controls. New control means are The Tech II or designate must:
established with a minimum of 20 values. As- 1. Under function QC, pull Levey Jennings
sayed control values must be verified corre- graphs for each control and analyte.
sponding to the methodologies by the laborato- 2. Review the statistics on Levey Jennings
ry for quantitative tests. graphs:
5. Depending upon availability more than one a. Compare established mean to monthly
level of control must be used for each analyte mean.
(i.e. low, medium and high). b. Trend or shift in values.
6. Controls are usually positioned after cali- c. %CV – changes in CV from previous
brators and prior to patient samples. Additional month.
controls are assayed at intervals depending on d. Evaluate precision compared to expect-
type of analysis and number of specimens to be ed precision.
analyzed. 3. Review external QC results for trends,
7. All QC values are entered into the MySys shifts and/or outliers.
laboratory and are monitored for trends or shifts 4. Review patient means and instrument cor-
using rules defined in MySys laboratory. relations and linearities.
5. Document any remedial action required or
Daily quality control routine make note if further evaluation to follow.
1. Review control results for the following: When a test suddenly and repetitively
a. values are within acceptable limits, yields unacceptable values, that cannot be cor-
b. review controls which are not within rected, check with senior staff and review
acceptable limits and add appropriate com- methodology.
ments. Notify senior tech. The Clinical Chemist reviews inter-
2. Document any reagent lot number changes, nal QC every 3 months.
instrument changes, or any other changes that Quality Control- Westgard Rules
may affect the results obtained.
Westgard Rules are designed to detect
Internal Quality Control random and systematic errors in analytical test
All staff must follow this procedure: systems. The rules are used to evaluate data
1. Verify that control results are within ac- both from within run (data in same run) and be-
ceptable limits. tween run (data from previous run), Table 1.
2. If control values are not within acceptable If all normal and abnormal control re-
limits, add appropriate comments and trou- sults are within ± 2SD of the mean established
bleshoot as required. Notify Tech II or Team for the test, patient results may be reported.
Leader. Random error is detected with R13S
3. When a test suddenly and repetitively and R12S rules.
yields unacceptable values, that cannot be cor- Systematic error is detected with R22S,
rected, check with the Tech II and review R412 and R102 rules.
methodology. R12S. Warning Rule
4. Document the following in the appropriate One control is outside ± 2SD
log book along with the date and initial: • Possible error within the method but is not
 Revista Română de Medicină de Laborator Vol. 8, Nr. 3, Septembrie 2007 63

considered cause for run rejection. • Proceed with troubleshooting. Notify Tech
• When violated, the remaining rules must be II if the problem is not resolved.
assessed. R 412. 4 results in a row outside of
• Violation of any of the remaining rules is 1SD with one of these results outside of 2SD.
cause for rejection of the run and patient results • Indication of a systematic error and trend,
are not released. • Some action is required as this trend may be
R22S. Two consecutive control re- more serious than R102,
sults exceed ± 2SD on the same side of the • Repeat control, review performance, trou-
mean (usually systematic error). bleshoot and document.
• If two consecutive control results exceed ± R10X. Warning rule
2SD on the same side of the mean, hold subse- Ten consecutive controls on the same
quent results. side of the mean (systematic error).
• Proceed with troubleshooting the procedure R102. Ten in a row on one side of the
or instrument. Notify Tech II or Team Leader if mean for either Level 1 or 2 (within) OR ten in
the problem is not resolved. a row on same side of mean combining Level 1
• Do not report patient values until the con- & 2 (across) PLUS one of these 10 is > 2SD.
trols are within ± 2SD and the patient samples • Repeat the control, but more importantly re-
have been repeated. view performance history checking manual
• Repeat random specimens whose results plots or by using function Levey-Jennings.
have been previously released. • Systematic error or trend indicator.
• Patient results may be sent in the control Rules violated
range where the controls are acceptable. Accept or reject control results as indi-
R13S. One control is outside ± 3SD. cated in Table 2.
• Usually an indicator of a random error.
External Quality Control
• Repeat control, troubleshoot if necessary
and document. External quality control surveys are an
RR4S. Range rule more than 4SD integral part of the quality assurance program.
difference in consecutive controls. The Chemistry department is enrolled
• Usually an indicator of random error or in the following programs:
poor reproducibility. • College of American Pathologists (CAP)
R41S. Warning Rule • Accutest
Four consecutive controls with a com- • Bayer Urinalysis Survey
bined range of greater than 4SD in either direc- • HealthMetrx
tion of the mean. • Centre de Toxicologie du Quebec
• Accept the run. • Centers for Disease Control and Prevention
• If the another rule is also violated, reject the
run. Patient results must not be released.
Table 2. Acceptance or rejection of control results.
Rules Violated Accept Reject
Table 1. Westgard rules used to detect errors R12S X
from within and between run
R13S X
Within Controls Across Controls R22S X
R12S, R13S, R22S, R22S, RR4S, R412, RR4S X
RR4S, R412 R102 R412 X
64 Revista Română de Medicină de Laborator Vol. 8, Nr. 3, Septembrie 2007

(CDC Atlanta) d. Submits the original form to the Chem-

• Health Canada istry manager for review and signing.
• Wisconsin State Laboratory of Hygiene
(NSLH). Acknowledgments
An external QC survey is processed as
follows: All materials including textbooks, pro-
1. Quality Assurance (QA) Department re- cedure manuals and other documentation was
ceives and delivers the survey samples, survey given by Dynacare Kasper Medical Laborato-
instructions and forms to the department. ries, Edmonton – Canada. Access was possible
2. Chemistry staff: with support of Dr. Trefor Higgins MSc,
a. Processes the survey samples as patient FCACB – Director of Chemistry Department –
specimens. Dynacare Kasper Medical Laboratories and
b. The samples are repeated only when with 2006 International Exchange Award spon-
they fall within the same criteria where a sored by Roche Diagnostics obtained from
patient would be repeated. No extra or spe- Canadian Society of Clinical Chemists. During
cial steps are taken. my time at Dynacare, I observed the perfor-
c. Fill out the survey form and forwards mance of serum and urine protein electrophore-
the form and raw data to the Tech II or sis as well as CSF, cryoglobulin and
Team Leader for review. hemoglobin electrophoresis. I spent time with
Acceptable results Dr. Higgins interpreting these electrophoretic
If results are within the allowable limits patterns. Also, I spent time learning how to in-
and accepted by the survey program: terpret HPLC chromatograms of hemoglobin
1. Quality Assurance Department sends a for the investigation of thalassemia and
copy of the final results to the department. hemoglobin variants. I observed quality control
2. Chemistry Manager and Tech II or Team procedures and I had access to all materials in
Leader review and sign the final data and file it the laboratory. With Dr. Trefor Higgins I visit-
in the appropriate department survey binder. ed the offices of Chenomx, an Edmonton-based
Unacceptable results company specializing in NMR spectroscopy of
If results are outside the allowable limit biological fluids and we discussed areas of mu-
and rejected by the survey program: tual interest with the CEO President and Scien-
1. Quality Assurance Department sends out tist.
result summary sheets to the department. Special thanks to Dianne and Trefor
2. Tech II or Team Leader: Higgins, Edmonton, Canada.
a. Directs staff to repeat the survey mate-
rial (if applicable).The following analytes
are unstable and are not repeated: Cl, CO2, References
ALP, TBIL, CBIL, CK, OSMO, LD, Mi-
croscopic Urinalysis, Macroscopic Urinaly- 1. Lawrence A. Kaplan, Amedeo Pesce, Steven
sis. Kazmierczak – Quality Control for the Clinical
Chemistry Laboratory. In: Lawrence A. Kaplan,
b. Fills in the result summary form and
Amedeo Pesce, Steven Kazmierczak - Clinical
signs it. Chemistry: Theory, Analysis, Correlation, 2002,
c. Makes a photocopy of the completed 390-405.
signed form and files it along with the re- 2. Dynacare Kasper Medical Laboratories Chem-
peat raw data in the appropriate department istry – Quality Control Procedures, June, 2007.
survey binder.