BloodLink - A decision support system for blood test ordering

Marc A.M. van Wijk, MD

Summary

In the Netherlands 3-4 percent of patients' encounter with the general practitioner result in ordering of blood tests. Dutch investigators report a lack of general practitioners’ knowledge concerning the indications for blood tests leading to inappropriate and inadequate use of diagnostic tests. Influencing test ordering behavior of Dutch general practitioners has been the objective of many studies. Three methods have proven to be effective in changing test-ordering behavior. Personal feedback, studied by Pop and Winkens, restricting the number of choices presented to the general practitioners on the order form, studied by Zaat, and the introduction of indication-oriented order forms based on guidelines, studied by Smithuis and van Gend all proved to be effective. Which of these methods is most effective remains unknown; randomized trials comparing these methods have never been conducted. The availability of guidelines of the Dutch College of general practitioners and the fact that almost all Dutch general practitioners replaced their paper patient records by computer based patient records creates new opportunities to introduce guideline based decision support in daily practice.

The objectives of this current thesis were to study (1) the consistency of the guidelines of the Dutch College of general practitioners with respect to recommendations for test ordering, (2) the requirements for a decision support system for test ordering, to be integrated in general practice, (3) the influence of a guideline based decision support system on test ordering behavior of general practitioners, and (4) the compliance of general practitioners to a decision support system for test ordering.

Consistency of the Guidelines

In Chapter 2, we analyze the practice guidelines of the Dutch College of General Practitioners, published until January 1^st 1998, with respect to the use of blood tests. We evaluated 64 practice guidelines of the Dutch College of General Practitioners. For each of the guidelines, we analyzed each sentence that contained a reference to a blood test (1) to determine the clinical situation in which the test should be performed (the indication), and (2) to determine the tests that should be performed in that situation (the advised test). Twenty-seven practice guidelines mentioned blood tests. Of these twenty-seven guidelines, three explicitly recommended not to request blood tests. Of the twenty-seven guidelines, twenty-three allowed us to identify indications and advised tests. We distinguished five different categories of indications: establishing a working diagnosis, investigating underlying pathology, monitoring the course of a disease, identifying therapy-influencing factors, and monitoring the side effects of drugs. Although some incomplete recommendations and inconsistencies were discovered, we conclude that the majority of the guidelines provide clear and unambiguous recommendations for blood test ordering in primary care.

Requirements for a Decision Support System for Test Ordering

After we concluded that the majority of the guidelines provide clear and unambiguous recommendations for blood test ordering in primary care, we commenced to build a decision support system for test ordering. Building a system to change test-ordering behavior requires us to select a method we will follow when providing support. We decided to use two methods: the restricted-order form method and the indication-oriented order form. To study these two methods for changing test-ordering behavior, we developed two versions of the decision support system BloodLink (described in Chapter 3). The first version, BloodLink-Restricted, is based on the notion of restricting the number of choices presented to the general practitioners. The second version, BloodLink-Guideline is based on the recommendations for test ordering as provided by the Dutch College of General Practitioners. BloodLink-Restricted is based on providing the general practitioner with a electronic version of a restricted order form that offers the general practitioner a list of 15 alphabetical ordered tests. These are the 15 tests that were judged by Zaat to be the most relevant in primary care. The only difference between the two versions of BloodLink is the method used to present the initial set of tests to the general practitioner. In all other respects, the two versions of the system are identical: the same integration with the computer-based medical record, the same layout of the screens, the same abbreviations of the tests, the same mechanism by which the general practitioner can add or remove tests, the same form printed, and the same notes left in the medical record. Implementation of both versions of BloodLink allowed us to study differences between the two methods.

The Influence of BloodLink-Guideline on Test Ordering Behavior

To compare the effect of two methods for changing blood test-ordering behavior, after stratification by single-handed practices and group practices, 44 primary care practices of 60 Dutch general practitioners in the region Delft were randomly assigned to BloodLink-Restricted, based on initially displaying a reduced list of tests, or to BloodLink-Guideline, based on the guidelines of the Dutch College of General Practitioners (described in Chapter 4). The intervention period was March 1996 through February 1997. The main outcome measure was the average number of blood tests ordered per order form per practice during intervention. Both approaches resulted in a decrease in the average number of tests ordered per order form when comparing the intervention period with the two years preceding the intervention: a 12 percent reduction in the BloodLink-Restricted group (p=0.001) and a 29 percent reduction in the BloodLink-Guideline group (p<0.001). General practitioners who had access to decision support based on guidelines requested on average 20 % fewer tests (5.5 tests versus 6.9 tests, respectively; Mann-Whitney test p=0.003, N=44) than general practitioners with access to decision support based on a form that initially displays a limited number of tests. We conclude that decision support based on guidelines is more effective in changing blood test-ordering behavior than merely reducing the number of test options. This study showed that BloodLink-Guideline was used for the majority of test ordering. We also conclude therefore that decision support systems are an effective method for introducing guidelines in primary care. In view of the little effort that was needed for the introduction of BloodLink in daily practice, the changes we found in test-ordering behavior could well be replicated elsewhere in primary care practices using computerized patient records

The Compliance of General Practitioners to BloodLink-Guideline

Although the effect of BloodLink-Guideline on test-ordering behavior was unequivocal, a clear impact of BloodLink-Guideline on the volume of tests ordered is not necessarily an indication of the degree of compliance to the guidelines. We studied, therefore, the compliance of the general practitioner to the recommendations of the test- ordering module BloodLink-Guideline (described in Chapter 5). Compliance was expressed as the percentage of order forms per practice and per indication that follow the recommendations for test ordering of the guidelines of the Dutch College of General practitioners. The study was conducted from March 1996 through February 1997. To assess if non-compliance was related to pending revision of guidelines, we analyzed which guidelines had been revised after the intervention period and compared the three most frequently added tests in the non-compliant order forms with the recommendations of the updated guideline. Thirty-nine percent of the indication-oriented order forms were compliant. Removing tests was rare compared to adding tests. The most striking finding was that many of the modifications that caused the order form to be non-compliant are supported by revisions of guidelines after the intervention period. We conclude that general practitioners rely on guideline-based test recommendations as a basic minimum for test ordering in daily practice but at the same time anticipate on pending revision of guidelines. The results of this study may encourage the use of computerized patient records to enter patient data during patient encounters.

Discussion

We conclude (in Chapter 6) that the guidelines of the Dutch College of General Practitioners contain concrete and specific recommendations for blood test ordering for specific indications. Based on this analysis, we developed the decision-support system BloodLink. By asking the general practitioner to identify the indication for test ordering, we limited the role of BloodLink to the selecting the test belonging to a specific indication. Blood-test ordering based on the guidelines of the Dutch College of General Practitioners results in a larger reduction of the number of tests ordered than test ordering based on an initially limited set of 15 blood tests relevant for general practice. BloodLink-Guideline implements the guidelines of the Dutch College in daily practice. General practitioners rely on guideline-based test recommendations as a basic minimum for test ordering in daily practice but at the same time anticipate on pending revision of guidelines. We showed that many of the modifications that caused the order forms to be non-compliant are supported by the revision of guidelines after the intervention period. We therefore recommend the Dutch College of General Practitioners to reconsider the method and frequency of revising guideline