Professional Documents
Culture Documents
PT FAILS ROOT CAUSE
PT FAILS ROOT CAUSE
ABSTRACT
Objective: To understand the reasons for proficiency testing (PT) 37.4%; PT Evaluation, 0.5%; Unexplained, 11.5%; and Other, 0.7%.
failures, toward the goal of continuous improvement of laboratory Reasons unique to the PT process accounted for 15.6% of all reasons.
services. Corrective actions mainly included correcting calibration, maintaining
or changing equipment, strengthening reagent management, rewriting
Methods: Laboratories participating in PT of routine chemistry procedures, and retraining staff members.
testing provided by the National Center for Clinical Laboratories
(NCCL) of China were given 3 schemes for handling PT failures in Conclusions: The causes for PT failures relate mainly to laboratory
2016. If they had unsatisfactory PT performance, they were notified problems, not to the PT process itself. Therefore, PT sponsored by the
and requested to investigate failure errors and submit data regarding NCCL is an effective evaluation tool for laboratory performance. Also,
those errors. laboratories should strive to improve analytical performance.
Results: Failure types were distributed as follows, per error group: Keywords: proficiency testing, clinical chemistry test, PT error, quality
Clerical, 10.2%; Methodological, 16.2%; Equipment, 23.6%; Technical, assurance, performance evaluation, corrective action
The scope of proficiency testing (PT) in the clinical labora- Laboratory Improvement Amendments (CLIA) of 19884 in
tory has evolved considerably since Belk and Sunderman the United States, the PT program was voluntary in China.
provided the first PT program in the late 1940s.1 However,
the first PT program in China was not provided until 1982 Because laboratory test reports are released from pro-
by the National Center for Clinical Laboratories (NCCL) of cesses that meet their quality specifications, an evaluation
China. Today, the PT program has become an essential that reveals unacceptable results by a PT provider yields
component of the quality management system in laborato- an unexpected outcome for the laboratory. In response, the
ries; it is required by laboratory accreditation organizations laboratory should investigate the causes of error and correct
in many countries, including China.2,3 Unlike the Clinical the procedure that led to PT failures, to reduce or eliminate
the chances of recurrences. Outcomes of investigations
into the reasons of PT failures can be used by laboratories,
Abbreviations by manufacturers, and by the PT providers themselves in
PT, proficiency testing; NCCL, National Center for Clinical Laboratories; improving their respective products and service.5–8
CLIA, Clinical Laboratory Improvement Amendments; CAP, College of
American Pathologists; EQA, External Quality Assessment; CLSI, Clinical & The results of a survey by Klee and Forsman6 of the Mayo
Laboratory Standards Institute; SDI, standard deviation index; QC, quality
control. Clinic during 1985 and 1986 showed that only 28% of PT
failures were associated with analytic errors, 23% were unex-
1
National Center for Clinical Laboratories/Beijing Engineering Research plained, and the rest were related to the difference between
Center of Laboratory Medicine, Beijing Hospital, National Center of
processes in the PT program and in the handling of patient
Gerontology, Beijing, China; 2Graduate School, Peking Union Medical
College, Chinese Academy of Medical Sciences, Beijing, China specimens. In contrast, the College of American Pathologists
(CAP) Q-Probes study6 of 665 laboratories and the study
*To whom correspondence should be addressed.
by Hoeltge and Duckworth indicated that PT is a good indi-
zgwang@nccl.org.cn
cator of the technical ability of the laboratory to perform a
**
Joint first author: contributed equally to this work. test.7,8 To determine the value of using PT results in judging
© American Society for Clinical Pathology 2018. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com 103
Management & Administration
laboratory performance, it should be demonstrated that the by the NCCL), before the deadline. If the results for a spe-
causes of PT failures relate mainly to laboratory-operation cific analyte were not submitted in a timely manner by a
problems other than the PT process itself. Therefore, we laboratory, that laboratory would obtain a score of 0 for
initiated an investigation into the reasons for PT failures, to the analyte in this testing event. Two weeks after the dead-
determine whether PT sponsored by the NCCL of China is an line for results reporting, participants received a report that
effective evaluation tool for laboratory performance. included an evaluation of whether their performance met
the acceptable criteria of specific analytes from the NCCL.
Materials and Methods score(s) for 1 or more analytes in each testing event of
SDI =
PT result of your laboratory - Target value
(2)
Results
Standard deviation
The 2369 laboratories evaluated 762,120 challenges that
where the target value was determined as the robust mean occurred during 152,424 PT events. Of the 152,424 PT
of the peer group and the SD refers to the robust SD of the events, 4838 were evaluated as showing unsatisfactory PT
peer group; both values were calculated by robust analysis performance for a specific analyte in the entire year. Of these,
via Clinet EQA statistical analysis software, version 5.5.0.46 1186 were evaluated as showing unsuccessful PT perfor-
(developed by the NCCL). For example, if a laboratory mance for an analyte (882 of which failed 2 testing events
reported a serum sodium concentration of 140 mmol per L, [302 failed successively and 580 discontinuously], 304 of
the target value for the peer group was 130 mmol per L, and which failed 3 successive testing events); the rest (3652) only
the peer group SD was 2.7 mmol per L, and then the SDI was failed 1 testing event for a specific analyte. Overall, 1907 par-
determined as ([130 mmol/L―140 mmol/L]/ 2.7 mmol/L) = 3.7. ticipating laboratories (80.5%) had at least 1 unacceptable
PT result (ie, failed at least 1 challenge in a testing event). Of
Classifying PT Failures these, 474 (24.9%) had no unsatisfactory analyte performance
in every testing event; the remaining laboratories may have
Table 1 shows the final version of classification for routine had unsatisfactory PT performance for 1 or more analyte.
chemistry in the study. In the Table, classes of possible
causes of unacceptable results included 7 major groups. In this investigation, participants submitted information,
Clerical problems were related to errors in the process when including reasons and corrective actions taken, for 7883
reporting PT results. Methodological problems were related to unacceptable challenge results (60.0% of all unaccept-
the analytical test system or the documented procedure itself. able results) in 2568 unsatisfactory PT events (53.1% of
Equipment problems were related to analytical instruments or all unsatisfactory PT events). As for problematic submitted
pieces of equipment used as part of the method. Technical information (15.0% of all responding results), 85.2% was
problems were related to personnel and the operation of verified by telephone, and the reasons were all corrected.
Comprehensive Analysis of Reasons for PT consensus by several experts of the NCCL in China, 9 sub-
Failure groups were deemed as being unique to the PT process,
accounting for 15.6% of all reasons, and 5 subgroups were
In Table 1, we show the categories of causes of unaccept- deemed to potentially contribute to multiple analyte failure
able PT results. A total of 9 subgroups accounted for 80.0% within a PT event.
of the reasons for PT failures, with the problem of “Failure
to adhere to scheduled instrument maintenance proce- Laboratories reporting “Failure to test correct PT speci-
dures” being the most common reason. According to the mens” as a reason for failure indicated that they failed 9
to the situation of routine chemistry analytes, and the cat- Of the 9 most-common problems for PT failures, which
egorizations are limited to use in evaluation of quantitative accounted for 80.0% of all the reasons, only 2, namely, “Result
measurement. transcription error” and “Improper reconstitution, preparation,
or storage of PT specimens”, were unique to the PT process.
Our findings indicated that repeated failures of PT events The small percentages (total: 7.8%) of the 2 reasons suggest
within a laboratory were very rare. Most of those failures that most problems identified from unacceptable results would
occurred in the next testing event or with 1 pass between also affect measurement of patient specimens. That is, labo-
2 events. Only a fraction of those events failed 3 PT events ratories were able to detect that analytical problems exist in
in a row. That is, unsuccessful PT performances for specific laboratory operation and quality management.
analytes were few; this result was similar to those from
other studies.7,11 The 3 main reasons for unacceptable PT results—“Failure to
adhere to scheduled instrument maintenance procedures”
Classifying the problem of PT failures was the most impor- (19.5%), “Calibration problems” (17.4%), and “Management
tant part of the study. Two previous studies from China problems for reagent” (10.0%)—can all be corrected funda-
and CAP investigated the reasons for early PT failures. mentally. As in other studies,7,10 we found that many unac-
Details about these 2 studies and the present one are ceptable results were unexplained. Because statistically
shown in Table 2. The percentages in the Unexplained cat- based PT grading systems determine PT failures based
egory during 2013 and 2016 in NCCL were smaller than on the distribution of PT results, statistical failures can be
that those percentages in the CAP study. The change in regarded as a result of random error when detected by this
the Unexplained category percentages may suggest that grading system and may explain why some unexplained
participants have more knowledge of the PT process and discordance exists. However, proper laboratory records
are better able to recognize errors of unacceptable results. may help laboratories determine the reason(s) for unaccept-
The main analytical related reasons for PT failures in these able PT performance, so the reports of some participants
studies were different. Blood gas analytes were included that their PT failures were unexplained may result from an
in the CAP Study, whereas in the others, they were not, so inability to reconstruct what happened when analyzing the
analyte-specific errors may exist. Comparing the results PT specimens.
of the present study with those from the study by Zhao
et al,10 we found that the percentages of problems in the Reasons unique to the PT process (Table 1) comprise 15.6%
Methodological category became smaller and may indicate of all failures; this percentage was smaller than that in the CAP
that laboratories themselves or manufacturers have made Study (which accounted for 24.5%).7 In contrast, most reasons
some improvements in their methodology for routine chem- (72.2%) were related to laboratory analytical performance,
istry analysis, such as using more-accurate methods or which demonstrated that PT, the most extensive external
changing manufacturers of reagents or reference materials. quality-control tool used by an accreditation organization, is a
potential tool to monitor performance improvement over time contribute to a true PT failure but is also in a region susceptible
and to assess overall laboratory analytical performance. This to statistical noise that could indicate false problems.
finding is similar to that stated by many reports.12–14 Moreover,
it is suggested that laboratories have more knowledge of PT Now that we have elaborated the reasons for PT failures, how
and can be familiar with the PT process with time. Reporting can laboratory scientists use this information to help the lab-
reasons for PT failures and corrective actions taken not only oratory investigate PT failures? It is best that laboratories take
can help laboratories detect PT failures and pay more attention limited steps to investigate obvious clerical problems first and
to PT process but also can help the PT provider find problems then take more comprehensive steps to find hidden errors in
unique to the PT process—for example, improving the web- analytical processes, including those in the Methodological,
Equipment, and Technical categories. Perhaps laboratories
We were not surprised that none of the participants reported Despite its promising findings, our investigation has certain
PT Material problems and few reported PT Evaluation prob- limitations. First, because participants are distributed in dif-
lems. As for PT Material problems, such as inappropriate ferent Chinese provinces, it was impossible to perform field
stability or homogeneity, PT specimens that were mislabe- investigations of the reasons, and the information gathered
led, and PT specimen material that had deteriorated in tran- could only be submitted via the online reporting system.
sit, these errors are the responsibility of the PT provider and Second, because each laboratory voluntarily participated in
are often impossible for participants to detect. As for the PT the PT program and there are no punitive measures for the
Evaluation problems, participants having little knowledge of laboratories with unsatisfactory PT performance, it is only
what is required to interpret PT results (including the process required that the laboratory report failing reasons before a
of target value assignment and acceptable criteria) may con- set time, rather than within a specific ranges of dates after
tribute to unquestioning acceptance of all results from PT receiving the PT report. Finally, high percentages of unex-
reports.15–18 So, PT providers should comment on specific plained unsatisfactory performance may limit the use of PT
data-handling processes in the feedback report. results as the unique indicator of laboratory ability to meas-
ure routine chemistry analytes accurately.
As Figure 1 shows, the SDI in the Clerical and PT Evaluation
groups were greater than those in the other groups; the results As a critical evaluation tool for laboratory performance,
of Kruskal-Wallis testing also demonstrate this result. Thus, PT instructed us to routinely monitor laboratory analytic
when the SDI of the unacceptable result was greater than 10, performance and take corrective actions when the causes
the priority of investigation of reasons should be given to these for PT failures had been found to assure the accuracy of
2 groups. Yet, the differences among other major categories measurement of patient specimens. In this study, it was
were not significant. When the SDI of the unacceptable result shown that the reasons for PT failures relate mainly to
was lower than 5, the SDI may not provide help. An SDI in laboratory operation problems rather than the PT process
the range of 2 to 3 implies a developing problem that could itself. Therefore, PT, sponsored by the NCCL of China, is an
effective evaluation tool for laboratory performance. This tool Laboratory Improvement Amendments of 1988 (CLIA)—HCFA. Final rule
with comment period. Fed Regist. 1992;57(40):7002–7186.
may also prove effective in laboratory systems in other parts
5. Clinical and Laboratory Standards Institute (CLSI). Using Proficiency
of the world. LM Testing and Alternative Assessment to Improve Medical Laboratory
Quality. 3rd edn.2016.
Funding 6. Klee GG, Forsman RW. A user’s classification of problems
identified by proficiency testing surveys. Arch Pathol Lab Med.
1988;112(4):371–373.
Funding for this study was provided by the Beijing Natural
7. Steindel SJ, Howanitz PJ, Renner SW; College of American
Science Foundation (grant/award no. 7143182); and Beijing Pathologists. Reasons for proficiency testing failures in clinical
Hospital Foundation (grant/award number: BJ-2015–025). chemistry and blood gas analysis: a College of American Pathologists
Q-Probes study in 665 laboratories. Arch Pathol Lab Med.
puter-technology support to establish the network platform 10. Zhao H, Zhang C, Zeng J, et al. Analysis of error sources for routine
chemistry external quality assessment program in China. Chin J Lab
for the survey and relevant services. Med. 2016;39:376–379.
11. Hoeltge GA, Phillips MG, Styer PE, Mockridge P. Detection
Supplementary Data and correction of systematic laboratory problems by analysis
of clustered proficiency testing failures. Arch Pathol Lab Med.
2005;129(2):186–189.
Supplemental figures and tables can be found in the online
12. Halim A-B. Proficiency testing for monitoring global laboratory
version of this article at www.labmedicine.com performance and identifying discordance. Lab Med.
2013;44(1):e19–e30.
13. Perich C, Ricós C, Alvarez V, et al. External quality assurance programs
as a tool for verifying standardization of measurement procedures: pilot
collaboration in Europe. Clin Chim Acta. 2014;432:82–89.