Decision-making laboratory computer systems as essential tools for achievement of total quality.
While many would argue that the time, effort, and expense involved in laboratory computerization is justified solely on the ability to do more with less, I suggest that the major motivation for laboratory computerization and automation should be to improve the total quality and predictability of the laboratory service. Quality can be defined as doing the right thing in a timely manner and doing it right the first time. Of course, what is "right" must be defined from the customer's perspective and expectations.
When comparing the efficiency of computers to humans, it is clear that computers can perform repetitive tasks much faster and much more accurately than humans can. On the other hand, if the task is less predictable (or nonroutine) and requires judgment and decisions, many older laboratory computer solutions are less effective than humans using a manual procedure. Indeed, many of the steps involved in performing laboratory tests are unpredictable in that they deviate from the routine because of unique patient conditions, special doctor orders, doctor notification, interpretive comments, atypical specimen conditions, special billing issues, and reporting issues. Some computer solutions require the user to recognize these deviations from the routine and develop "workarounds" to handle variances. The development and use of such "off-line" workarounds, especially for infrequent occurrences, form the basis of customer dissatisfaction because these nonroutine tasks rarely get handled in a predictable manner. Although older laboratory computer systems generally perform routine functions rather well but the nonroutine functions poorly, it follows that newer computer solutions for the laboratory must recognize and handle both routine and nonroutine situations with ease. Indeed, newer laboratory computer systems must handle these nonroutine "data" events by either (a) directly performing the nonroutine task or (b) alerting the user that a nonroutine variance has occurred so a manual solution can be used and documented. To date, only newer computer systems with expert programing capabilities coupled with a dynamic database can effectively handle most of the routine and nonroutine tasks involved in performing laboratory tests. Hence, expert computer systems and programs now allow us to handle those situations previously requiring judgment or special consideration by humans. What are the results? Expert computer solutions have the potential to improve overall quality by making even nonroutine laboratory services error-free and more predictable with essentially no human intervention. Although our expert system can perform 31 separate actions based on data events, I will focus on just a few examples of how an expert decision-making laboratory information system was adapted to our laboratory service.
Materials and Methods
EXPERT SYSTEM PROGRAMS
Decision-making programs were developed on our PATHLAB 3 (PL3) System (HBOC, Murray, UT). The PL3 System runs on our Data General (DG) 40,000 mini-computer, and for expert programing, we used the DG proprietary AOS/VS2 operating system, the DG MP/ AOS Slate text editor, and the DG INFOS relational database. Our various expert scripts or programs are invoked automatically by the system when a "data event" occurs such as a result entry, result verification, or patient admittance.
The expert script program shown in Fig. 1 monitors all results being entered into our system. In the example shown, when a TSH result code is detected, the program checks to see if the TSH result falls into a "borderline" range, 4.6-10 mU/L. (1) If so, the system then searches for the free [T.sub.4] results. If the free [T.sub.4] results fall into the normal range (i.e., 7.0-20.0 ng/L), three specific decisions or actions are automatically performed by the expert program. These are as follows.
DECN1. The technologist is alerted that a borderline TSH has been observed together with a normal free [T.sub.4], a result that many of our physicians regard as a discrepancy. The alert is displayed across the screen, and the technologist is required to enter a "Y" (or yes) to acknowledge the alert.
DECN2. This action appends a comment to the TSH result stating that increases of TSH are frequently seen together with normal free [T.sub.4] results and giving the reasons for such occurrences.
DECN3. This action appends a record to a file on the PL3 system that later can be processed and downloaded to a personal computer (PC) for further statistical evaluation of these detected data events. As in the example, the record layout for the log file includes accession number, user identification, patient name, medical record number, visit number, patient location, sex, TSH result, and free [T.sub.4] result.
A "macro" program was developed on the PL3 system to periodically extract the logged TSH-free [T.sub.4] data to a separate file, named TSHLOG, for downloading. Using the PL3 TEXTMENU program for downloading and a personal computer (IBM PS2 Model 77) attached as a terminal (ProComm Plus, Seattle, WA), we downloaded the TSHLOG file to the PC. The resulting PC ASCII file was then ported to DBASE IV (Borland, Scotts Valley, CA). Subsets of this PC database were then ported to a macro-driven spreadsheet the author developed in LOTUS (Lotus 123, Cambridge, MA). With these macros on the Lotus spreadsheet, correlations and other statistics were performed automatically along with printouts and x-y plots (see Fig. 2). Other examples of how expert system "actions" are used to perform routine functions and minimize errors in our laboratories are listed in Table 1.
[FIGURE 1 OMITTED]
[FIGURE 2 OMITTED]
Results and Discussion
I have concluded that true quality in the clinical laboratory cannot be achieved by focusing primarily on errors in the analytical process. Rather a laboratory should concentrate on trying to eliminate events that are clearly errors in the view of clients, i.e., lost specimens, poor specimen/ test inquiry and tracking, slow and unpredictable turnaround time, lost reports, and errors in billing. Clearly, the laboratory computer system must indeed address all issues relative to the testing process for us to approach the next degree of quality.
EXPERT SYSTEMS AND CODED COMMENTS
As shown in Fig. 1, an expert program activated on the PL3 can trap a data event such as an increased TSH observed with a normal free [T.sub.4] result. Three decision symbols are activated with DECN1 alerting the technologist of the data event, DECN2 "true" results in appending a comment to appear along with the TSH and free [T.sub.4] events, and DECN3 causing the data event to be logged in real time on the PL3 system for subsequent download. Data collected over a period of time were extracted from the system with a program macro on our PL3 system and then imported into a statistical spreadsheet for evaluation. The data gathered automatically by the system in a prospective manner show no correlation between borderline elevations of TSH and our associated free [T.sub.4] results. This example of the use of expert systems illustrates how the system can automatically perform multiple operations on data events, tasks that previously would have required manual intervention by the technologist.
EXPERT SYSTEMS AND THE TEST ORDER
Manual test request systems are the source of uncontrollable errors in the testing process. An adequate expert ordering system includes the handling of billing information, managed care group, appropriate codes including Current Procedure Terminology (CPT4) test codes and International Classification of Disease, 9th rev. (ICD9), diagnosis codes, previous encounter and pertinent clinical data, real time and electronic access to an online test catalog, and barcode label(s) identifying the specimen and patient (especially for remote clients). When the specimen arrives in the laboratory, the only requirement should be to scan the specimen barcode to update the electronic tracking system on the expert laboratory information system.
REFLEX TEST ORDER AND CANCELING
As part of the test-ordering process, the expert system must automatically handle reflex testing protocols preapproved by the medical staff for either ordering additional test(s) when preliminary screening test results indicate or canceling a test deemed useless by preliminary screening tests. Expert systems thus ensure that overutilization of valuable testing resources are minimized and monitored [1-4]. Clearly, the traditional laboratory performs an excessive amount of laboratory tests, and hence the reporting of this plethora of raw laboratory data in spreadsheet format frequently makes the meaningful information on the report more difficult to find .
On the other hand, expert systems and reflex testing allow testing to proceed along a more diagnostic track with the expert system supplying the diagnostic comment in some cases [4,6,7].
RESULTS RETURN ISSUES
The expert laboratory computer system handles results return issues including whether to print, fax, e-mail, or transmit electronically laboratory results for storage on the clients' electronic medical records systems. As described, manual reporting systems are replete with errors unless closely monitored . If the client requires all types of reporting, the expert system must handle the data event. For example, a client may wish to fax only critical results while printing others. Also, the expert system can handle issues like cumulative reporting for some clients or daily encounter summary reports for others. We have recently programmed our expert system to fax results based on entry of certain coded comments for results interpretation.
The expert computer system must support paperless or electronic reporting, where reports must be transmitted in a format compatible with the client's electronic medical records system. Here, the expert system is programmed and will make decisions relative to format headers and other requirements of the receiving electronic medical records system.
Physician inquiry programs must have expert programing to handle security issues and to ease the use of the computer system for physicians who are less frequent users of the system, for example, our expert system can reduce patient inquiry from a very complex, feature-rich, six-screen program to a single entry line program. Hence, a physician can simply enter the patient's medical record number, the number of days to include in the database search, and then the test codes, using commas for delimiters (e.g., 369056,9,CBC,CHEM7). The expert program will then display all of the specified laboratory results in 1-2 s. The expert script language required to perform this in-house-designed doctor inquiry program is called "expert path." Technologists also find such expert path programs helpful because multiple keystrokes are eliminated for commonly used tasks such as looking up the previous hemogram or urinalysis to help verify the accuracy of a current study.
LABORATORY SECTIONS REQUIRING EXPERT COMPUTER SUPPORT
Essentially all areas of the laboratory require expert computer support, especially chemistry, hematology, microbiology, and the blood bank. For adequate decision support, expert computer programing must be available for all aspects of the testing process including test request, phlebotomy (if required), specimen logging/processing/ tracking, specimen analysis, and reporting/ inquiry. Currently, blood bank expert systems must comply with Food and Drug Administration requirements, especially when "truth tables" are used to verify and automate procedures such as electronic crossmatch.
TYPES OF EXPERT "ACTIONS" BEING USED
On the basis of data events being monitored by the expert laboratory computer system in our laboratory, the system can (a) alert the analysts that a certain data event has occurred, (b) order a reflex test or cancel a test already ordered based on a data event, (c) handle critical results reporting issues, (d) monitor unacceptable specimens, (e) handle discrepant results reporting, (fl allow for standing orders, (g) append coded comments for interpretation, (h) log the occurrence of a data event into a downloadable ASCII file, and (i) based on a data event, e-mail or fax a laboratory report to a client.
Because many of the above data events occur infrequently (i.e., two to three times per year), we have found the only way to ensure predictable handling of the event is to program the expert system to handle these events automatically.
REQUIREMENTS OF AN EXPERT SYSTEM
An expert computer system must: (a) be easily modifiable by a noncomputer programmer; (b) be database-driven; (c) support multiple actions based on certain data events; and (d) not degrade the performance of the computer system. Generally, expert programs are written in a script language with programming activities that are readily understood. The database must be robust and allow for continual operation without older data elements being frequently purged from the "active" database. The expert script language program does not necessarily have to be changed when new or different data elements are added or existing data elements in the database are modified. In general, the expert language needs to be highly structured and have access to virtually any database element. The expert system must support decisions in data-rich areas such as chemistry and hematology and still handle textual data events generated in departments such as microbiology. The expert system must be adaptable "in-house" so that it can be readily modified to meet local laboratory needs.
FEATURES OF NEWER EXPERT SYSTEMS
In general, newer computer systems are more powerful and thus provide more rapid performance and have fewer problems with system degradation when expert system programing scripts are layered over traditional laboratory application software. Newer expert systems use intelligent terminals as opposed to entirely "dumb" workstations, a feature that facilitates downloading of data captured by expert script programs to PC spreadsheets, databases, documents, and statistical programs. In addition, application programing languages used are highly structured and screen-oriented, integrate seamlessly with the database and expert programing, and are portable to a variety of different platforms.
Expert computer systems for the laboratory together with automation provide the only path to elimination of recurring errors in the laboratory. In the past, much attention has been focused on quality of the analytical process, so much in fact that now 95% of errors occur in the nonanalytical steps of a laboratory test. Requirements of an expert laboratory system have been described  and are summarized as follows: (a) speed, expert programing must not degrade performance of the overall system; (b) reliable and supportable, the expert system script language must be highly structured and be relatively easy to program, debug, and support; 3) connectable, the expert scripts must integrate with existing application programing and databases; and 4) adaptable, the expert system must be easily adaptable to the changing laboratory environment.
Received December 11, 1996; revised and accepted February 24, 1997.
[1.] Blick KE. Medical laboratory systems. In: Stafford JEH, ed. London: Blackie Academic, 1995:108.
[2.] Travers EM, ed. Laboratory test appropriateness guidelines. Dunbar, NC: Regional Medical Education Center, 1995:1-70.
[3.] Speicher CA. The right test, 2nd ed. Philadelphia: WB Saunders, 1993:1-250.
[4.] Check WA. Changing how clinicians use the lab, decision-making system needed for laboratory test ordering followed by computer controlled reflex testing. CAP Today 1993;7:1, 27, 36-8.
[5.] Boyce N. Neural networks. Clin Lab News 1997;23:1-2.
[6.] Peters M. Managing test demand by clinicians: computer assisted guidelines. J Clin Pathol 1995;48:98-100.
[7.] Skjei E. Information systems follow the "rules." CAP Today 1992; 6:1, 31-6.
[8.] Banning J, Brown J, Hooper L, Hamilton J, Burnett J, Burnett L. Reduction of errors in laboratory test reports using continuous quality improvement techniques. Clin Lab Manage Rev 1994;7: 424-36.
(1) Nonstandard abbreviations: TSH, thyroid-stimulating hormone; [T.sub.4], thyroxine; PC, personal computer; ASCII, American Standard Code for Information Interchange.
KENNETH E. BLICK
Department of Pathology, University of Oklahoma Health Sciences Center, Box 26307, Oklahoma City, OK 73190. Fax 405-271-3620; e-mail KBLICK@REV.UOKHSC.EDU.
Table 1. Selected examples of how computerized expert decisions are routinely used. Action Data events Cancels CK For Intensive Care/Cardiac Care locations and isoenzyme test total CK <50 U/L. All other locations, cancellation of CK isoenzymes for total CK <100 U/L. Orders microscopic For patient <2 years and (or) urine specific examination of urine gravity >1.030, and (or) positive for reducing substances, and (or) glucose +3 or greater, and (or) protein trace or higher, and (or) nitrite positive, and/or occult blood positive, and (or) appearance hazy or turbid. Alerts technologist On verification of urinalysis, expert system to save urine scans orders to see if a pregnancy test is specimen for ordered on the specimen. If so, the system pregnancy test alerts the technologist to save the specimen for pregnancy testing. Otherwise, the specimen would be discarded. Logs results and On entry of lead results by atomic absorption data for billing spectrometry, the expert system logs and statistics on information into a file for downloading to a PC lead analysis for statistical purposes and billing. Logs cytogenetic On entry of cytogenetics, a results log file is results for billing updated and downloaded for use in statistics, and reporting billing, and reporting of final results. Adds interpretive On verification of various serology results, comments to interpretative report is automatically serology results generated based on the type of test. Corrects microbiology On results entry, expert system scans text of results microbiology results looking for "no growth to date" appended to specimens >5 days old. If found, alerts technologist and logs event for supervisor. Results are corrected to say "no growth at 5 days." Alerts technologist Expert system alerts technologist to call to call physician and attending physician and infection control infection control based on certain positive culture results. Expert system lists phone numbers etc. as part of a "help" expert path. Leukocyte esterase On entry of urine culture results, the expert positive results system scans for urinalysis leukocyte appended esterase-positive results on specimen collected the same day. If found, appends esterase results to microbiology results header and alerts technologist. Logs and prints fluid On entry of any fluid count and (or) fluid counts and differential, the system logs occurrence along differentials with results for a required review by supervisor and subsequent correlation with cytology results.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Clinical Chemistry Forum|
|Author:||Blick, Kenneth E.|
|Date:||May 1, 1997|
|Previous Article:||Quality: the next six months.|
|Next Article:||On the measurement of cholecystokinin.|