| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Articles |
1
Acquired Intelligence Inc., No. 205-1095 McKenzie Ave., Victoria BC V8P 2L5, Canada.
2
MDS Metro Laboratory Services, 4489 Viewmont Ave.,
Victoria BC V8Z 5K8, Canada.
a Author for correspondence. Fax 250-479-0764; e-mail bsmith{at}aiinc.ca
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Methods: To study the influence of the LAS on laboratory investigations, a repeated-measures experiment using clinical vignettes was conducted. To collect baseline data on how laboratory investigations are currently conducted, clinicians investigated one-half of the vignettes using a conventional (noncomputer) approach. To determine the influence of the LAS on clinicians' behavior, the other half of the vignettes were investigated using the LAS.
Results: Clinicians using the LAS (compared with conventional practice) ordered fewer laboratory tests during the diagnostic process (mean, 17.8 vs 32.7), completed the diagnostic workup with fewer sample collections (mean, 5.8 vs 7.5), generated lower laboratory costs (mean, $194 vs $232), shortened the time required to reach a diagnosis (mean, 1 day vs 3.2 days), showed closer adherence to established clinical practice guidelines, and exhibited a more uniform and diagnostically successful investigation.
Conclusion: The LAS enhances the outcome of the investigation and improves laboratory utilization.© 1999 American Association for Clinical Chemistry
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
A further drawback to the use of guidelines for utilization control is that guideline dissemination alone does not sustain clinician behavior change even when clinicians are in agreement (7)(8)(9). Guideline implementation is more likely to succeed when accompanied by changes in the practice environment that make it easier for physicians to maintain a new desired behavior (8). The work environment must provide opportunities to practice the new behavior and provide continuing reinforcement and feedback. In their discussion paper, Elson and Connelly (8) stated that: "Physicians' behavior cannot be systematically corrected merely by changing knowledge, attitudes and/or motivation ... they need new tools that will make it easier for them to do the right thing". An example of this is a recent study investigating the laboratory testing of 969 newly diagnosed cases of essential hypertension (10). The authors found that 24% of the cases did not receive the laboratory testing recommended by the guidelines of the Canadian Hypertension Society, whereas >50% received tests that were not specifically recommended. The Canadian Hypertension Society guidelines were widely disseminated: they were mailed to every licensed physician in the country and published in the Canadian Medical Association Journal. The authors noted that "it is unlikely that any guideline could be more widely disseminated without extraordinary expense" (10).
An expert system implemented as an interface between the clinician and the laboratory offers the possibility of addressing some of the difficult issues of developing, disseminating, and adhering to guidelines. The major benefits of an expert system interface are (a) the ability to represent more sophisticated and widely applicable protocols than can currently be implemented with traditional approaches, (b) the ability to make that information available to clinicians at the time of ordering and viewing the results, (c) the ability to make the information specific to the clinical circumstances of the patient, and (d) to provide a complete record for retrospective review of the clinical problem, test orders, and results. Expert systems have been applied to specific clinical problems for 15 years (11). Recently, these techniques have been directed toward more generalized clinical assistance (12)(13). Advances in office computerization and in expert systems technology (in both integration and representation schemes) have made the development of an expert system interface timely. Today, many physicians use computers as an integral part of their office practices. Some physicians work interactively with a computer during their office activities to review their patients' medical records. It is expected that this trend will continue. Peters (14) argues that computer technology is now able to facilitate the routine use of clinical guidelines and should contribute proactively to the delivery of medical care.
The Laboratory Advisory System (LAS) is an expert system program that works interactively with clinicians to assist them with test selection and result interpretation throughout the laboratory investigation of a patient. More than simply a tool for electronic ordering and results receipt, the LAS uses its underlying knowledge bases to optimize the laboratory investigations for better care and lower costs by (a) optimizing patient-specific test ordering strategies, (b) providing patient-specific result interpretation, and (c) offering context sensitive assistance throughout the process.
To study the influence of the LAS on clinicians' pursuit of laboratory investigations, a repeated measures experiment using clinical vignettes was conducted. To collect baseline data on how laboratory investigations are currently conducted, physicians investigated one-half of the vignettes using a conventional (noncomputer) approach. The other half of the vignettes were investigated using the LAS. Each vignette appeared an equal number of times in the conventional condition as in the LAS condition. The influence of the LAS on clinicians' pursuit of laboratory investigations was measured by the number of tests ordered and the number of sample collections required. Additional observations were made of the laboratory costs, the turnaround time, the adherence to established clinical practice guidelines, and the success of the investigation.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
), tests can be requested by name or through the pursuit of a
clinical problem (e.g., thyroid function). The diagnostic approach to
each clinical problem is guided by the expert system, which may use
several steps in which clinical information is elicited, testing
approaches are presented, and further information is available. For
example, if the clinician pursues a thyroid function investigation,
clinical information is collected to make the appropriate
recommendation (see Fig. 2
).
|
|
In the result receipt mode (see Fig. 3
), the test results are displayed in a standard fashion. Below
the numerical values an interpretation is presented. The knowledge base
of the expert system is able to combine the clinical information
elicited during the ordering phase with the numerical value(s) of the
tests to derive a specific interpretation. In addition, an option for
further testing (on the same sample) may be provided. This feature
allows additional testing to be carried out when it will enhance the
diagnosis. Because the second round of testing is conducted on the same
sample, it is possible to reduce costs and turnaround time by avoiding
a second venipuncture. Interpretation of the results of additional
tests takes into account the clinical information provided and the test
results from earlier rounds. Fig. 4
shows examples of a recommended follow-up test and of an
interpretation of results based on information from all rounds of
testing. An additional feature of the system is context-sensitive
"help", which provides explanatory information about each test,
diagnostic pursuit, and clinical problem.
|
|
The LAS acts as an intelligent interface between the clinician and the laboratory information system. It is built using the ACQUIRE® suite of expert system tools (16)(17)(18)(19).1 This includes a development environment that consists of a knowledge acquisition system, an inference engine, and an application programmer's interface. The knowledge acquisition system provides a structured approach to developing the expert system application. In addition to the standard production rule representation of knowledge, ACQUIRE offers a knowledge representation scheme based on a pattern recognition approach that conceptualizes expertise as a highly developed pattern-recognition skill. Following from this approach, the specialist's knowledge is captured in "pattern-consequence" relationships. In the LAS knowledge bases, patterns are made of relevant clinical information, and the consequences are testing recommendations and interpretations. Representing expertise as pattern-consequence relationships provides greater specificity in the knowledge base and is particularly useful for representing the exceptional cases not handled well by production rules. The LAS user interface is built using Visual Basic, and the underlying relational database is implemented in Microsoft Access. The LAS provides a well-defined programmable interface to integrate with other compatible software systems such as patient management systems and laboratory information systems. In this experiment, the LAS ran under Microsoft Windows 95 on a Pentium laptop with 16 megabyte memory.
the trial
Subjects.
Physicians with between 10 and 20 years of
experience, working in private, general practice were contacted. Six
physicians with varying familiarity with computers were able to
volunteer their time during the experimental period. (Procedures using
human subjects were in accordance with the current revision of the
Helsinki Declaration of 1975.)
Vignettes.
Vignettes of 14 cases representing standard
clinical problems found in ambulatory practice were prepared (i.e., 4
thyroid, 4 hepatitis, and 6 autoimmune). For example, one example
states: "The patient has been on thyroid replacement therapy for 3
months. The dosage was increased 4 weeks ago. She continues to complain
of occasional palpitations. You wish to monitor the status of the
therapy. What do you order?" Test results returned to the physician
were indicative of excessive replacement therapy, i.e., a raised free
thyroxine result (26 pmol/L; reference range, 10.5–20.0 pmol/L)
and a suppressed thyrotropin (TSH; 0.02 mIU/L; reference range,
0.5–3.59 mIU/L). Another example states: "A previously well 37 years
old high school teacher with no previous history of drug or alcohol
abuse or transfusion, attends your office with a seven day history of
malaise. One week ago he noted his urine to be `very dark'. On
examination he is slightly jaundiced, and his liver is slightly
enlarged, tender and smooth. You wish to pursue the possibility of
hepatitis. What do you order?" The test results rule out hepatitis,
i.e., the aspartate aminotransferase is within the reference
range (30 U/L; reference range, 
40 U/L). A final example states: "A
45 years old woman has a several year history of gradual thinning of
the skin on her fingers. She has had Raynaud's phenomena since she was
a teenager but these symptoms have become occasionally debilitating in
the past three years. She has also experienced episodes of malaise and
in the past 6 months has experienced shortness-of-breath on exertion.
You wish to pursue a diagnosis of scleroderma (disseminated sclerosis).
What do you order?" The laboratory results confirm the diagnosis of
scleroderma, i.e., the antinuclear antibody titer was a high
positive, the antinuclear antibody pattern was nucleolar, and the Scl70
was positive. In each case a specific diagnostic direction was
suggested. This was done because it was the test ordering and result
interpretation for a given diagnosis that was of interest, not
the diagnostic ability of the clinician. In each case, the laboratory
results were unambiguous in confirming or rejecting the diagnosis the
physician was asked to investigate.
Procedure.
In this repeated measures experiment, each
clinician was given one-half of the vignettes to investigate using the
conventional (noncomputer) approach, using paper requisitions and paper
laboratory reports. After the investigation of these cases was
completed, each clinician was given the other half of the vignettes to
investigate using the LAS. It is important to note that the advice
provided by the LAS can be completely or partially ignored and that
additional tests can be added. To emulate the usual office environment,
the initial laboratory orders were obtained on all the cases as though
they were consecutive patients in an office. Only when all the orders
had been placed were the results returned. The clinicians were then
interviewed to find out what diagnostic conclusions they had been able
to reach from the test results and whether there were any further tests
that they required. These steps were repeated until the clinicians had
completed the investigation. Each vignette appeared an equal number of
times in the two investigative modes. The presentation of cases
within a session was randomized.
The investigative approaches, conventional and LAS, were compared to determine whether LAS could influence physician behavior. The criteria studied were:
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
number of tests
A comparison of the mean number of tests showed that the six
clinicians ordered a mean of 32.7 tests with the conventional approach
vs 17.8 when they used the LAS (t = 5.4; P
<0.01). This is a highly significant difference, especially because
the physicians using the conventional approach referred 34% of the
cases to specialists with little or no laboratory investigation.
number of sample collections
All first-round testing was judged to involve one collection. If a
second round of testing was requested, a second collection was
attributed to the conventional approach, whereas a follow-up collection
was not required when the LAS was used, provided the original sample
was appropriate. Significantly more sample collections were required
with the conventional approach (mean, 7.5) vs the LAS (mean, 5.8;
t = 3.4; P <0.02). Again the difference is
striking given that so many of the cases investigated with the
conventional approach were referred to specialists rather than pursued
with laboratory tests. Consideration of the number of sample
collections is important for two reasons: First, 30% of the overall
cost of performing laboratory testing on an outpatient is attributable
to the collection process (20). Any methods that would
reduce patient encounters would potentially reduce overall costs.
Second, additional testing without requiring a second patient visit
will significantly reduce turnaround time (see below). These results
and the following observations are summarized in Table 1
.
|
cost of testing
The total cost of each investigation, as if it were a patient
workup, was determined by adding the internal costing estimates of the
tests ordered and the collections required. The results show
significantly higher costs for investigations pursued using the
conventional approach (mean, $232) vs the LAS (mean, $194;
t = 3.3; P <0.05). This is partially
attributable to fewer tests, but as Winkelman (21) points
out, the relationship between costs and number of tests performed is
nonlinear. The greatest reduction in costs is attributable to fewer
sample collections. As mentioned above, the sample collection process
accounts for a large proportion of the overall costs. This difference
is again striking because the cases that were referred to specialists
rather than pursued with laboratory testing (all in the conventional
approach) contributed zero costs. It should be noted that because the
vignettes do not represent the usual mixture of ambulatory clinical
problems, these results cannot be extrapolated to gauge the total
financial impact of the system.
diagnostic accuracy
Physicians using the LAS were more likely to arrive at correct
diagnosis (100% of the time vs 66%). In the investigations using the
conventional approach, physicians were correct in confirming or
rejecting the diagnosis 66% of the time because physicians
investigating autoimmune disorders referred many of the cases to
specialists before reaching a correct diagnosis. In addition, one
physician did not order a key test (Scl70 for scleroderma) and
incorrectly concluded the presence of a disorder "on clinical
grounds" when the result was negative.
turnaround time
In all cases using the LAS, the cases could be completely turned
around in 1 day. In contrast, when the conventional approach was used,
the following mean times could be estimated: thyroid cases, 4.5 days;
hepatitis cases, 1.4 days; autoimmune cases, 5.3 days (mean of means,
3.2 days). Reducing the turnaround time of a laboratory investigation
is beneficial because it reduces the number of "cases in progress",
thereby streamlining the physician's office. It also allows the
diagnosis to be reached sooner.
Given the significant differences in the number of tests ordered, the number of samples collected, and the cost of testing, it is possible to examine the impact of LAS under different circumstances: (a) in cases for which guidelines exist, e.g., thyroid testing; (b) in cases for which guidelines are not available, e.g., hepatitis and autoimmune; and (c) in cases for which the results are negative.
established guidelines
Given that making physicians aware of the guidelines even if they
agree with them is not sufficient to sustain behavior change
(8), it is of interest to investigate the influence of the
LAS in an area where there is a well-established, legislated protocol,
i.e., thyroid testing. In addition to being well established and
commonly used in our jurisdiction, the implementation of guidelines in
this area was documented in 1990 to save more than $1 million per year
in a region with a population of 2.3 million (20). In the
current study, a subset of four cases pertaining to thyroid testing was
examined to investigate the influence of LAS in a clinical area in
which established guidelines exist.
The mean number of tests ordered was the same for physicians using the conventional approach and the LAS approach, but fewer sample collections were required to complete the investigation using the LAS. We looked at the individual test cases and the physicians' interpretations and found that better compliance with the established guidelines was achieved when the LAS was used and that more specific diagnoses were made with the LAS. For example, a physician using the conventional approach ordered a test that is no longer recommended by the current guideline. Two physicians using the conventional approach interpreted results differently from the guidelines (they used different reference ranges), and in the two cases investigating thyroid monitoring, physicians used shorter retest periods than the guidelines suggest. Physicians using the LAS were more uniform in their test ordering, more consistent with the guidelines, and by choosing to clarify the presence of thyroiditis as offered by the LAS, achieved more specific diagnoses. Although the analysis of this subset of data is limited, there is indication that even with well-established, generally agreed upon, legislated guidelines, there are better outcomes when the LAS approach is used.
no established guidelines
A similar investigation was conducted on the cases in clinical
areas that did not have established guidelines, i.e., hepatitis and
autoimmune disease testing. In these cases, there were twice as many
tests ordered by physicians reviewing the cases using the conventional
approach than by physicians using LAS, 27% more sample collections,
and many more referrals to specialists. The differences in tests
ordered and samples collected are conservative given that clinicians in
the conventional mode referred 34% of the cases to specialists with
little or no investigation.
negative results
Given that the most common outcome of outpatient laboratory
testing is negative or normal results, it is of interest to investigate
the influence of LAS in the subset of cases for which the results are
returned to the physician as negative. There were four such cases (two
thyroid, one hepatitis, and one autoimmune). In these cases, there were
more than twice as many tests ordered by physicians using the
conventional approach than by physicians using the LAS and 55% more
sample collections.
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
One of the important features of the LAS is its ability to derive and reason with clinical information during the ordering process and to store this information in a database. This information is notoriously difficult for laboratories to obtain. Having such information allows several possibilities:
In conclusion, an expert system-based laboratory investigation tool (LAS) has been developed to assist physicians with ordering, interpreting, and understanding laboratory investigations. In this trial, the LAS enhanced the outcome of the investigation while reducing test utilization, turnaround time, and cost. As clinical practice becomes more electronically connected, such knowledge-based tools will be available for every physician to use. This study demonstrates that they may have a manifest effect on improving laboratory investigations.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  S. Shahangian and S. R. Snyder Laboratory Medicine Quality Indicators: A Review of the Literature Am J Clin Pathol, March 1, 2009; 131(3): 418 - 431. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. J. Poley, K. I. Edelenbos, M. Mosseveld, M. A.M. van Wijk, D. H. de Bakker, J. van der Lei, and M. P.M.H. Rutten-van Molken Cost Consequences of Implementing an Electronic Decision Support System for Ordering Laboratory Tests in Primary Care: Evidence from a Controlled Prospective Study in The Netherlands Clin. Chem., February 1, 2007; 53(2): 213 - 219. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 W G Simpson and P J Twomey Reflective testing J. Clin. Pathol., March 1, 2004; 57(3): 239 - 240. [Full Text] [PDF]
|
|
|
|
|
|
E. M. Lim, K. A. Sikaris, J. Gill, J. Calleja, P. E. Hickman, J. Beilby, and S. D. Vasikaran Quality Assessment of Interpretative Commenting in Clinical Chemistry Clin. Chem., March 1, 2004; 50(3): 632 - 637. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. M Kuilboer, M. A M van Wijk, M. Mosseveld, and J. van der Lei AsthmaCritic: Issues in Designing a Noninquisitive Critiquing System for Daily Practice JAMIA, September 1, 2003; 10(5): 419 - 424. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. M. Kuilboer, M. A. M. van Wijk, M. Mosseveld, and J. van der Lei AsthmaCritic: Issues in Designing a Noninquisitive Critiquing System for Daily Practice J. Am. Med. Inform. Assoc., September 1, 2003; 10(5): 419 - 424. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G Gopal Rao, M Crook, and M L Tillyer Pathology tests: is the time for demand management ripe at last? J. Clin. Pathol., April 1, 2003; 56(4): 243 - 248. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Regeniter, J. U. Steiger, A. Scholer, P. R. Huber, and W. H. Siede Windows to the Ward: Graphically Oriented Report Forms. Presentation of Complex, Interrelated Laboratory Data for Electrophoresis/Immunofixation, Cerebrospinal Fluid, and Urinary Protein Profiles Clin. Chem., January 1, 2003; 49(1): 41 - 50. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. Bonini, M. Plebani, F. Ceriotti, and F. Rubboli Errors in Laboratory Medicine Clin. Chem., May 1, 2002; 48(5): 691 - 698. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. A.M. van Wijk, J. van der Lei, M. Mosseveld, A. M. Bohnen, and J. H. van Bemmel Compliance of General Practitioners with a Guideline-based Decision Support System for Ordering Blood Tests Clin. Chem., January 1, 2002; 48(1): 55 - 60. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. Dube and B. Wu Supporting clinical laboratory test-ordering protocol specification, execution and management: an event-condition-action rule and database approach Health Informatics Journal, March 1, 2001; 7(1): 20 - 28. [Abstract] [PDF]
|
|
|
|
 |
|
 M. A.M. van Wijk, J. van der Lei, M. Mosseveld, A. M. Bohnen, and J. H. van Bemmel Assessment of Decision Support for Blood Test Ordering in Primary Care: A Randomized Trial Ann Intern Med, February 20, 2001; 134(4): 274 - 281. [Abstract] [Full Text] [PDF]
|
|
Read all eLetters
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
