KVALITA INOVCIA P ROSPERITA / QUALITY INNOVATION PROSPERITY XIII/1 2009 1

PROCESSES ASSESSMENT AND MONITORING IN

A CLINICAL LABORATORY
DANA TRVNKOV

1 INTRODUCTION
Nowadays demands for quality continue to grow not only in production, but also
in services. Service is an activity which takes place between a customer and a
provider. There is a disadvantage compared to a product service is more
difficult to provide and mainly, it is more difficult to set measurable parameters
of a service. Quality of health care services is a very sensitive subject, it is
important for not only the health care providers. It is very important for state
administration, health insurance payers and mainly the public, the potential
patients.
There are three dimensions of the healthcare services quality (Madar, 2004):
Quality of the service from the clients view point;
Quality of the service from the view point of management the most
economical and most efficient use of resources within the framewrk of
directives and limits, set by superiors or payers;
Quality of the service from the professional view point if the services
fulfill the needs in the way as they are defined by professionals who
execute them or who send patients to take the service and if the services
contain suitable techniques and procedures, which are necessary for
fulfilment of clients needs.
The article focuses on the services provided by a clinical laboratory from the
professional view point. Its performance can be measured only when the
performance indicators are correctly set and measured.
The objective of this article is: To set indicators for measuring the ancillary
processes capability and to amend the quality indicators in a clinical laboratory
with capability indexes as far as processes assessment is concerned, and with
control charts with moving limits for analytic phase processes monitoring .
Contribution of the article is expected in following areas:
Assessment of the analytical phase processes performance;

ISSN 1335-1745
2 KVALITA INOVCIA P ROSPERITA / QUALITY INNOVATION PROSPERITY XIII/1 2009

Based on the processes performance, setting up of a metodology for usage

of the tools to determine frequency of running control samples within the
internal quality control;
Reduction of the laboratory services costs by setting up a suitable
frequency of running control samples;
Assessment of the use of moving control charts for processes monitoring.

2 METHODOLOGY
The services provided by clinical laboratories are unavoidable in the health care
system. The results of analytical testing have a strong impact in medical
treatment. The results can influence (even fatally) the patients health, quality of
life and sometimes the life itself. The results of tests are the basis for important
decisions on diagnosis, prognosis and the way how to proceed with medication.
That is why the quality of the tests results is so important (precise and accurate)
and the laboratory response time (time from receipt of samples to despatch of
results) minimal. Nowadays, when laboratories use fully automated analysers,
the response time is not a problem any more. The quality of results is closely
connected with assessment of capability of all processes and sub-processes in a
laboratory.
The quality level assesment is based on comparison of what really is with the
vision what should be the optimum of quality. The result of the assessment
therefore has influence on determination of what should be and detection of what
really is. In the health care this problem is long time focused on by the Joit
Commission on Accreditation of Healthcare Organization (JCAHO) and the
national or international standards are considered as the vision of the optimum of
quality (Zgodavov, 2006).
In the Czech Republic the capability of clinical laboratories is being assessed
according to the standard SN EN ISO 15189:2007 Clinical laboratories
Special demands on capability and quality. This European standard has been
approved by CEN and it is used by clinical laboratories to develop their quality
management systems and self-assessing of their capability. It is also used by
accreditation bodies to assess the clinical laboratories capability.
Clinical laboratories can supply 70% of information about patients, When these
information are irrelevant, they cannot help neither the doctor, nor the patient
more to the contrary. How to define quality in a clinical laboratory? The
American Institute for Quality (Richardson, 2003) suggested to define the quality
in a clnical laboratory as the Laboratory system for collection, examination and
issuing of results of human samles, which :
Supports diagnosis, prevention and management of ill conditions;
Gives information of a clinical importance about patients health status;

ISSN 1335-1745
 KVALITA INOVCIA P ROSPERITA / QUALITY INNOVATION PROSPERITY XIII/1 2009 3

Meets the requirements on accuracy, repeatibility and traceability;

Tries to minimise mistakes;
Is quick, safe, efficient and is not expensive;
Is focused on patients satisfaction and continual improvement.

Laboratory examinations should fit to the clinicians needs, laboratories should

ensure the confidence of doctors and patients to the examinations results and to
guarantee that the costs were spent efficiently. The definition of quality differs
according to view point and needs of stakeholders (Westgard, 2008).
The quality of laboraory examinations depends on many factors. Some of them
can be influenced by the laboratory managemet, some originate outside the
laboratory, mostly within the pre-analytical phase. Suitable quality indicators can
be selected on the basis of three different principles, which represent three
different concepts (approaches, models) of quality. The concepts are based on
three models: analytical, biological and clinical ones (Hyloft, 1994).
In the laboratory it is possible to identify processes and class them into groups:
Managing processes are used to control the laboratory functioning;
Main process examination of biological samples, the output of which is
determined for laboratory customers;
Processes of resources management are used to control resources the
laboratoty uses;
Ancillary processes these processes support the above mentioned groups
of processes.

The main process examination of biological samples can be devided into

three sub-processes: pre-analytical phase, analytical phase and post-analytical
phase. For the main process to functionate well it is necessary to ensure that also
the ancillary processes are set and function well (Fig. 1).

ISSN 1335-1745
4 KVALITA INOVCIA P ROSPERITA / QUALITY INNOVATION PROSPERITY XIII/1 2009

Organisation Quality Policy

Organisation Quality Objectives

Laboratory Quality Objectives

Sub-process Sub-process Sub-process

preanalytical analytical postanalytical
phase phase phase

EXAMINATION OF BIOLOGICAL SAMPLES

Sampling, Validation, Results,

transport, internal, external interpretation,
pjem... control... complaints...
Ancillary
processes
Main Laboratory Process Performance
Indicators

Fig. 1 - Algorithm of processes performance indicators (modified with author

permition) (Nenadl, 2001)

Sub-process Analytical phase performance indicators

The performance of the sub-process Analytical phase is measured through
assessment and monitoring of examinations performed. After the Internal Qualty
Control has been mastered, the process performance is measured also by the
External Quality Control. Monitoring of the examinations is a logical follow-up

ISSN 1335-1745
 KVALITA INOVCIA P ROSPERITA / QUALITY INNOVATION PROSPERITY XIII/1 2009 5

of validation, resp. verification. When the IVD (In-vitro diagostics) validated by

their manufacturer are used, verification only is enough to measure examinations.

Verification
To measure the sub-process Analytical phase by the ancillary process
Verification we use following parameters: accuracy, precission, process
capability.
Accuracy indicators: SD, CV
Precission indicators: Bias %
Capability indicators: sigma capability, capability indexes

For measuring monitoring of the ancillary process internal quality control

the Control Charts are used.
Parameters: control charts,
Indicator: SD, data mean

For measuring of the ancillary process external quality control we focus on

how the laboratory succeeds in the given cycle of controls:
Parameters: success in the given cycle, success in the last 2 years
Indicators: TE, Z score

Verification of the analytical process contains, according to recommendation of

the professional associations, the parameters accuracy and precission. In this
article also capability will be included to these parameters and it will be
described by capability indexes.

Processes monitoring
For monitoring of the analytical processes laboratories use control charts, which
are valuable quality control tools. They are very significant, because when they
are rightly chosen and interpreted, they give valuable information about the
process (examination procedure, testing of a sample) behaviour and performance.
Basically, the control charts should be used as a diagnostic tool to assess if the
process tested behaves in the way we expect. Analysis of the control charts can
detect in advance significant deviations of the process from the set levels, find
and explain the causes and perform corrective actions.

ISSN 1335-1745
6 KVALITA INOVCIA P ROSPERITA / QUALITY INNOVATION PROSPERITY XIII/1 2009

The choice of a control chart depends on character of the measured data and their
probability distribution. Then we are able to calculate the control limits. When
the control chart is chosen incorrectly, the probability of detection of process
deviations decreases. In reality it means that, for example, we can get data points
out of control limits even in cases of no change in the process. To be able to
fully utilise the advantages of control charts, the data distribution must be
Gaussian (normal). This is the requirement also of the Shewharts i Levy
Jennings control charts, which are mostly used in clinical laboratories (ISO
8225).
When the data distribution is not Gaussian, it is necessary to use an alternative
control chart instead of the Shewharts one, for example the EWMA chart with
moving control limits. As shown on Figs 2 and 3 (analyte fT4), when the EWMA
is used, the number of rules violation is significantly reduced. The charts have
been constructed with the use of real data, analyte fT4 (free-Thyroxine).

Fig. 2 Shewharts chart x individual Fig. 3 EWMA chart with moving limits,
with dependant data fT4 analyte fT4

Ancillary process Verification the use of Capability Indexes

Processes capability
Capability of a process is a measure of the process real quality compared to a
standard (specification). We asses it after all the systematic effects have been
removed, i.e. in the status when the process is under statistic control. When
assesing the process capability, we then assess only the variance caused by
random effects. When the variance is too high, process cannot have results which
would be permanently within range. Such process must be examined and after
that corrective action(s) must be taken.
As for verification the Cpk indexes were used, other indexes are not mentioned
in this article. Index Cpk takes to account not only the variability of tested quality
parameter, but also the real capability of the process to keep within prescribed
tolerance limits. Its value then reflects the ratio of distance of the mean of the
tested quality parameter from the closer tolerance limit to a half of real data

ISSN 1335-1745
 KVALITA INOVCIA P ROSPERITA / QUALITY INNOVATION PROSPERITY XIII/1 2009 7

variability. Index Cpk can be calculated for both one-side and two-sides
tolerances.
USL - x x - LSL
C pk = min C pU ; C pL , C pU = , C pL = (1, 2, 3)
3 SD 3 SD
It is possible to say, that analytical processes with Cpk between 1,0 and 1,33 are
reliable. Processes with the Cpk value below 1,0 are less reliable and the
probability of incorrect result (non-conformity) occurance is higher. Processes
with the Cpk value above 1,33 are well reliable and with the value above 1,67
they are highly reliable with a very low probability of incorrect result (non-
conformity) occurance (Plura, 2001). Example shown on Fig.4.

Fig. 4 Probability density curves for Cpk = 1,33

For verification, which is sufficient for an analytical process assesment, it is

nowadays enough to asses accuracy (trueness) and precision, or good results in
the External Quality Control. To be able to calculate these indicators, it is
necessary to obtain data, which have usually no further use. But, the data could
be used for calculation of other processes performance indicators, like, for
example, the suggested capability indexes.
The capability indexes provide information about another property of the
analytical process. The calculation takes to account tolerable and real process
variability, like the Six sigma metrics. The Cpk capability index contains
information about accuracy and trueness together. The accuracy is described by
means of the standard dviation SD or the coefficient of variability CV% and
gives information on accuracy achieved only, but without assessment, if it is still
acceptable or not. But the required accuracy of analytes in the biological samples
differs according to the biological variability. If the required accuracy is added to
the capability criterion, for example min. Cp 1,3, the capability index value
returns clear information about acceptability of the variability.

ISSN 1335-1745
8 KVALITA INOVCIA P ROSPERITA / QUALITY INNOVATION PROSPERITY XIII/1 2009

3 APPLICATION
The application part of this work contains evaluation of the use of capability
indexes, which were incorporated into verification parametres. The capability
indexes were calculated and evaluated for 98 tests at two or three levels. It was
found out that 41,06% of all the 263 tests have not reached required capability.
Such a high number of incapable tests was caused by including of
immunochemical methods of testing. These methods have usually higher
variability than other tests (analyzers Architect, Unicel a Stratec). When the
capability indexes were assessed without these methods, the number if incapable
processes was significantly reduced, down to 27,51%. A very good capability
(Cpk 1,33) was achieved at 40,1% of all tests, respectivelly 53,44% without the
immunochemical methods. The worst results were noted at the low levels of the
analytes, the best results at the high levels.

Fig.5 - The Cpk values rate per analyzer

Further on the analyzers have been assessed independently. As we can see at

Figs. 5 and 6, there is a significant difference betwen immunochemical methods
and the other principles. As far as all other principles than the immunochemical
methods are concerned, the differences among them are minor. For the Synchron
Z 1 a Z 2 analyzers, where tests are performed the same way, there is a difference
visible on Fig. 5 in the number of tests with Cpk 2,0. This difference is caused
by different age of the analyzers. The analyzer Synchron Z 2 has been in use for
significantly shorter time than the Synchron Z 1. It means, that capability of
analytical processes is indispensably influenced also by the age of equipment.

ISSN 1335-1745
 KVALITA INOVCIA P ROSPERITA / QUALITY INNOVATION PROSPERITY XIII/1 2009 9

Fig.6 - The rate of Cpk < 1 and 1 per analyzer

The Cpk value depends a lot on the required variability. The tolerance range was
set according to the recommendation of SEKK (www.sekk.cz). In some cases
(for example ALT) it would be wise to adapt the tolerances to concentration
levels. If the required test variability is set correctly, it is possible to use the
capability indexes for determination of frequency of the control tests
performance, as recommended in the Table 1.

Tab. 1 Calability indexes and frequency of control tests

Cpk Value Capability Control testing frequency
> 1,0 low In each series (for example of 20
samples), if the level ifs physiologically
important
1,01 1,32 acceptable 2 x in each series (for example at the
beginning and at the end). Series up to 1
day.
1,33 1,67 very good 1 x in each series (serie 1 day)
> 1,67 high each 2 to 3 series (series 1 day)

Frequences, shown in the table, are valid for particular levels of concentration of
the control sets. In the 1 day series minimally one level must be measured.
Correct application of the capability indexes into the clinical laboratories
operation can bring significant quality improvement and also reduction of costs,
connected with possible reduction of running control tests.

ISSN 1335-1745
10 KVALITA INOVCIA P ROSPERITA / QUALITY INNOVATION PROSPERITY XIII/1 2009

The ancillary process Internal Quality Control monitoring by classical EWMA

chart and EWMA chart with moving limits.
In the last few years we can see comments in professional literature and also
comming from laboratories calling attention to frequent cases of automated
processes, where regulation by the classical control charts (Shewharts type) is
inadequate or impossible. These charts were introduced at the break of the
twenties and the thirties of the last century and were intended for controlling
measurable and attributive parameters (Michlek 2003, Montgomery 2001,
Zvrov 2002).
Classical SPC methods, developed for the conditions prevailing in production,
perform well only if following criteria are met:
The details about process are obtained in regular intervals by collection of
data in selections of the range n>1
Collected data are statistically independent within the selection and among
themselves
The selections are made the way to form logical sub-groups (it must be
ensured that among the elements of the selection there are no determinable
causes of variability)
Data come from identical statistical distribution (for continuous stochastic
variables we usually expect Gaussian normal distribution).

When these assumptions are not adhered to, the classical SPC methods fail. The
SPC system in this case more frequently than it was estimated falsely signals that
there are determinable causes influencing the process and therefore that the
process is not under statistic control (Noskieviov 2003; Michlek 2003).

EWMA Charts
EWMA charts were introduced in 1959. It is abbreviation of Exponentially
Weighted Moving Average (sometimes called exponential forgetting). Its use is
simmilar to Shewharts charts. With advantage it is used in cases when we
cannot guarantee conditions necessary for use of the Shewharts charts (normal
distribution, independent data).
In situations, when data are dependant, it is possible to use a modification of the
EWMA chart the EWMA with moving limits, a chart with one step prediction
of mean and variability.
The EWMA chart with moving limits is suitable for processes, where the
parameters show possitive autocorrelation and the proces has a non-constant
mean which changes slowly.

ISSN 1335-1745
 KVALITA INOVCIA P ROSPERITA / QUALITY INNOVATION PROSPERITY XIII/1 2009 11

Interpretation of the EWMA chart with moving limits is identical to

interpretation of other control charts. If all the values of measured parameter xk
are within the control limits, the process is considered under statistic control and
changes in the process are slow. If some of measured values lies outside the
control limits, it should be considered a signal that the process is out of statistic
control (Toenovsk, 2000).
On Figs. 7 and 8 there is an example shown where for the same dependant data
the classical EWMA chart and EWMA chart with moving limits are used. The
classical chart requires higher necessity of intervention to the process, which
proves the false signal. Also the quickness of detection is different (Kupka,
2001).

Fig. 7 Classical EWMA Chart, analyte Fig.8 EWMA Chart with moving
fT4 limits, fT4

To compare the Shewharts control charts X-individual + R and the EWMA

charts several representatives of clinical examinations were chosen on the basis
of the explorative analysis. The data have been processes by means of statistical
software QC-Expert. Creation of the Shewharts regulation charts depend on
data Gaussian distribution and data inependency. The explorative analysis found
several variants of the requirements violation. In some cases the data had
Gaussian distribution and were independent, which is the requirement for
Shewharts charts creation, in some cases the data did not have Gaussian
distribution but were independent. The last case was that the data did not have
Gaussian distribution and were dependant. Each of the mentioned variants has
been processed. Further on the examinations were chosen, for which both of the
chart types were created. The examinations were chosen taking in mind that they
should cover most of the analytes, according to their structure and function
(enzymes, sacharides, lipides, proteins, ionts, tumor markers a hormons).
For 29 compared analytes the Shewharts charts, classical EWMA charts and
EWMA charts with moving limits were created. With the total number of 4605
data, the numbers of data outside limits for Shewharts and classical EWMA
charts do not differ significantly (229, resp. 239); the classical EWMA chart
refused more data only by 4,4%. There is a significant difference between those

ISSN 1335-1745
12 KVALITA INOVCIA P ROSPERITA / QUALITY INNOVATION PROSPERITY XIII/1 2009

two charts and the EWMA chart with moving limits. When it was used, the
number of refused values is only 53, i.e. 4,4 times less.

4 CONCLUSION
Analytical processes assessment
Within the frame of verification, which is enough to assess the analytical
processes, nowadays only accuracy, resp. trueness are evaluated, eventually also
the success in the External Quality Control. To be able to assess the parametres
we need to measure data which are not used further more. The same data can be
used for calculation of other process performance indicators, like the suggested
capability indexes, namely the Cpk, which contains both the information about
accuracy and presission. If the required accuracy is included into the capability
criterion, for example min. Cp ? 1,3, the capability index can give a clear
information that the variability is acceptable or not. Using the Cpk capability
index we get information how the process performs against the required target
value, i.e. about the process trueness, which is normally described by bias%.
Incorporation of the capability indexes to the parametres of
validation/verification would mean that more information about analytical
processes could be obtained.
Verification of analytical processes was extended by the use of capability
indexes. 98 different examinations have been assessed, each at min. two levels.
The assessment of analytical processes by means of capability indexes can be
used for determination of control samples testing frequency. Correct monitoring
of the processes is very important for the quality of the examinations results,
which are used for treatment of both ill and healthy patients.
Monitoring of the analytical processes
The assessment of EWMA chatrs useability for monitoring of processes was
performed at different combinations of dala properties. When the conditions for
Shewharts charts were met (i.e. independent data with Gaussian distribution),
the classical EWMA chart was more sensitive to a systematic error. In case of
dependent data, we can estimate that signal given by this chart is false, idle,
because the chart with moving limits does not detect the limit violation. The
variation was small, it is so unimportant and the process remains in the stable
status. In the two cases data measured within one month were used. In the third
case for the charts creation data measured within six months were used. The
difference between various charts is higher in this particular case. While
Shewharts x-individual and classical EWMA charts signal more frequent
violation of the limits, the EWMA chart with moving limits signals the same
only once.
Further it was discovered and compared, that data comming from the same
process, but on different concentration levels of the measured analyte, have in the
same time period different character. For this assessment data from one month

ISSN 1335-1745
 KVALITA INOVCIA P ROSPERITA / QUALITY INNOVATION PROSPERITY XIII/1 2009 13

only were used, but tested twice a day, which meant doubled number of data
compared to previous months. The higher the number of data was, the higher was
the difference among usage of various control charts. The reason was that data
change their character in time.
The performed analyses lead to conclusion that the usage of EWMA charts in
clinical laboratories is possible, but it is not necessary for monitoring of the
processes. In laboratories the values are recoded to the control charts once or
twice a day, only rarely more frequently. The time period, which is important for
an analyst, is maximally one month. With this volume of data there was found no
significant difference among the compared control charts. The false signal about
the limits violation comes out in a longer time period (4-6 months), but it is
worthless information for the laboratory operation. The usage of EWMA charts,
namely the chart with moving limits, is suitable for processes, where the test
measuring is performed more frequently, apparently at production processes.
The alert to the incorrect use of the Shewharts control charts due to
inappropriate data properties has not been confirmed in a clinical laboratory
operation. The false signal moreover occures only in case when no deviation
happened. The case, when all the points would be falsely within the control
limits, has not occured. This means that any intervention to the process would be
needless, not that the violation would not be detected. The number of false
signals in a clinical laboratory is minimal, therefore also the needless
interventions costs are minimal.

REFERENCES
Hyloft, G. (1994): Setting quality standards in clinical chemistry: Can competing
models based on analytical, biological and clinical outcomes be harmonized?
Clin Chem 40/10, 1994. s. 1865 1868
Kupka, K. (2001): Statistick zen jakosti. TriloByte, Pardubice 2001. 183 s.
ISBN 80-238-1818-X.
Kupka, K. (2003): Metody statistickho zen jakosti. AUTOMA . 7-8, 2003. s.
13-17.
Madar, J. a kol. (2004): zen kvality ve zdravotnickm zazen. Grada
Publishing, Praha 2004. 174 s. ISBN 80-247-0585-0.
Michlek, J. (2003): Nov pohled na Shewhartovy diagramy. AUTOMA . 7-8,
2003. s.10-12.
Montgomery, D. C. (2001): Introduction to statistical quality control. 4th edition.
Wiley & Sons Inc., New York 2001. 748 s. ISBN 0-471-31648-2
Nenadl, J. (2001): Men v systmech managementu jakosti. Management
press, Praha 2001. 300 s. ISBN 80-7261-054-6.

ISSN 1335-1745
14 KVALITA INOVCIA P ROSPERITA / QUALITY INNOVATION PROSPERITY XIII/1 2009

Noskieviov, D. (2003): Automatizovan vroba a SPC. AUTOMA . 7-8, 2003.

s. 5-9.
Plura, J. (2001): Plnovn a neustl zlepovn jakosti. Computer Press, Praha
2001. 244 s. ISBN 80-7226-543-1.
Richardson, H., Gun-Munro, J. (2003): What does quality mean to you? 2003.
www.qmpls.org/library/newstetters/newsletters.html, accessed: 15.7.2009.
Toenovsk, J., Noskieviov, D. (2000): Statistick metody pro zlepovn
jakosti. Montanex, Ostrava 2000. 362 s. ISBN 80-7225-040-X.
Westgard, J. O. (2008): Six sigma: Quality design and control processes.
www.westgard.com/lesson67.htm, accessed: 15.7.2009.
Zvrov, J. (2002): Zklady statistiky pro biomedicnsk obory. Karolinum,
Praha 2002. 201 s. ISBN 80-7184-786-0
Zgodavov, K. (2006): Manarstvo kvality v zdravotnckych organizcich.
VET Koice 2006.

ABOUT THE AUTHOR

Dana Trvnkov, achieved her secondary education as the clinical
laboratorian and later won her masters degree from the VB-Technical
University Ostrava, in the field of quality management. For many years she has
been working in clinical and analytical laboratories as the quality manager. With
her capacity she prepared the laboratories for accreditation according to the
standards EN ISO IEC 17025 and EN ISO 15189. She also prepares and lectures
specialized courses for the educational centre Dm Techniky Ostrava. She
activelly participates at professional actions and publishes in conferences
proceedings and journals.

ISSN 1335-1745