![]()
30
| 2 | 2005
Klinisk Biokemi i Norden
autoverification rule sets can be changed by our
staff (independently of Mylab Oy, the manufac-
turer of the autoverification system) and there are
no restrictions on when or how often they can be
changed - they can even be changed “on the fly”
while the results are being autoverified.
A detailed history of the autoverification of all
individual results is written to a log file accessible to
the users. The information in the log file is invalu-
able for checking the workings of autoverification
rules and detecting the causes of problems as well
as providing information for further development of
the rule sets and for scientific purposes.
For this study we collected the results of one of
our Roche/Hitachi ModularTM analyzers during a
period of four weeks (from October to November
2004).
The most common laboratory tests of the selected
analyzer were plasma creatinine concentration,
plasma C-reactive protein concentration (CRP),
plasma potassium concentration, plasma sodium
concentration, plasma alanine aminotransferase
activity (ALAT, ALT, GPT glutamic-pyruvic trans-
aminase) and alkaline phosphatase activity (AFOS,
ALP), The autoverification rule sets of these tests are
shown on Table 2.
For each result, we collected the time when the
LIS received the result of the assay from the ana-
lyzer as well as the time when the result was made
available to the person who requested the test. We
use the term verification TAT to describe the dif-
ference between these times. This is the time when
the result is ready but waiting for the autoverifica-
tion program or a laboratory technician to check
it. As such it is the exact measure of the effect of
autoverification on laboratory TAT.
Results
During the four weeks of this study, the Hitachi
Modular analyzer produced a total of 87793 results,
out of which 86% were autoverified. For the ten
most common assays, there were only small dif-
ferences in the autoverification percentages that
ranged from 81 to 89%. The number of results and
autoverification percentage of these assays are
shown on Table 3.
When autoverification percentages were compared
between different days of the week, a somewhat
higher degree of variation was detected: the percen-
The basic logic of the checking process is that a
result is accepted by default but it will be rejec-
ted if any one of the rule chains rejects the result
by returning the logical value true (which in this
context means “do reject”). Hence, the rule chains
are connected by a disjunctive relationship. Each
rule chain consists of one or more individual rules,
which in contrast to the relationship between rule
chains, are connected in a conjunctive relationship,
meaning that for the rule chain to reject a result all
the rules of that chain have to reject it. Hence, a
result will be rejected in those cases where all rules
in at least one rule chain reject the result. The logi-
cal framework of the autoverification system makes
it possible to handle independent reasons for rejec-
ting a result as well as reasons that depend on a
particular condition such as the patient being from
a particular ward or clinic. For example, a delta
check with different parameters can be defined for
patients of one ward and another for patients of all
other wards. In the same way, different rules can
be defined to for patients of different sex and/or
age groups, for patients having abnormal results of
other assays, and even for individual patients. For
practical reasons, there is an exception inconsistent
with the logic of the autoverification system: results
of assays for which there are no rules at all are
rejected by default and always manually verified.
Some autoverification rules can affect also the
results of other assays made from the same sample.
In addition of accepting or rejecting results the
autoverification rules can have a variety of side
effects useful for other purposes. It was already
mentioned that they can be used to collect certain
results into special result groups. They can also
be used to append comments to the results. These
comments or messages can be directed only to the
laboratory staff or also to the requester of the assay.
The autoverification rules are also used to connect
the autoverification system with the program that
handles traditional quality control based on control
samples and Bull’s algorithm (a running average of
patient’s results). Thus, results can also be rejected
by the autoverification system when there is a prob-
lem detected by the quality control program.
Any time, autoverification can be turned on or
off at will by the operators of the analyzers. The
(Fortsat fra side 29)
