Authors' objectives:

The objective of this Special Report is to evaluate the evidence on interventions to change physician prescribing behavior.

Authors' results and conclusions: Twenty-three controlled trials met the inclusion criteria for this review, testing a total of 34 different interventions. The majority of interventions (24/34) resulted in significant changes in prescribing behavior; however, some programs (10/34) were not successful. It was difficult to determine any clear patterns of intervention type, setting, or study design that distinguished successful and unsuccessful programs. It was also difficult to estimate the magnitude of effect for these interventions. Some programs demonstrated relatively large effects, while others demonstrated little or no effect. Estimates of the attributable effect had wide variability within and across groups, thus limiting the utility of this summary measure. There is evidence from this Special Report that the class of medication targeted may be a determinant of program success. Interventions targeting antibiotics demonstrated consistent success and often had a relatively large magnitude of effect. There may be numerous reasons for this pattern. Reducing unnecessary antibiotic use is a compelling and unambiguous message. Physicians are likely to be receptive to changes in antibiotic prescribing, given the recent attention on problems of overprescribing and recognition of the problem by most physicians. Also, these were primarily recent studies, which incorporated findings from earlier research into the design of targeted, multidimensional interventions that were more likely to succeed. Patterns for other classes of medications were less pronounced. There was more variability in the intent of the programs and the specific messages given than with the antibiotic studies. The more diffuse nature of the messages in these interventions may have contributed to the lesser success. Most programs for antihypertensive medications were successful, but all had a small percent attributable change. Interventions that were intended to reduce polypharmacy had both a low success rate and a small percent attributable change. This evidence offers some support that specific features of interventions are associated with greater impact. For the majority of the 9 individual elements tested, the success rate was higher for programs including that element. However, this pattern was not universal, and the differences in success rates do not appear to be large when grouped this way. There was also a higher success rate for programs that included 3 or more recommended elements, compared to those that did not. There are other limitations to these data. This report was limited to prospective, controlled trials, which are not the most prevalent type of study for program evaluation. There is a relative lack of high-quality, randomized, controlled trials. On the other hand, this is a difficult area in which to perform high-quality trials. Researchers have to balance the need for high internal validity with pragmatic issues such as constrained resources and generalizability to the real world setting. The individual trials have low generalizability, and the reproducibility of successful programs is uncertain. The full range of possible interventions is not covered by the controlled trials included in this review. For example, no studies examined the impact of financial incentives on prescribing behavior, in spite of the fact that this is becoming a popular method used by health care organizations. There is also a possibility for publication bias, which may result in an overestimate of the proportion of programs that are successful. The durability of these educational improvements cannot be evaluated with the current evidence, as the longest trials do not extend beyond 12 years.

Authors' recommendations: The evidence from this review demonstrates that interventions intended to change provider behavior can be successful, in spite of the numerous challenges inherent in achieving this goal. This review also suggests that certain factors can contribute to program success, but that no single factor, or factors, is overwhelmingly a determinant of success. It is likely that numerous factors, including clinical factors, demographics, and features of the intervention, interact to determine overall impact. This review highlights the need to take a broad view in considering all potential contributing factors to maximize the potential benefit of an intervention.

Authors' methods: Review

Project Status: Completed

URL for project: http://www.bcbs.com/blueresources/tec/contact-tec.html

Year Published: 2004

English language abstract: An English language summary is available

Publication Type: Not Assigned

Country: United States

Organisation Name: BlueCross BlueShield Association

Contact Address: BlueCross BlueShield Association, Technology Evaluation Center, 225 North Michigan Ave, Chicago, Illinois, USA. Tel: 888 832 4321

Contact Name: tec@bcbsa.com

Contact Email: tec@bcbsa.com

Copyright: BlueCross BlueShield Association (BCBS)