Techniques to improve physicians' use of diagnostic tests: a new conceptual framework.

OBJECTIVES

To review the published literature on interventions aimed at improving physicians' testing practices and propose methodologic standards for these studies and to review selected studies using the PRECEDE framework, a behavioral model that helps categorize interventions based on which behavioral factors are being affected.

DATA SOURCES

MEDLINE, EMBASE, and HEALTHStar databases were searched for the years 1966 to January 1, 1998, for English-language articles pertaining to diagnostic testing behavior; bibliographies were scanned to identify articles of potential interest; and researchers in health services, health behavior, and behavior modification were contacted for proprietary and other unpublished articles.

STUDY SELECTION

A total of 102 articles were identified that described the results of interventions aimed at changing physicians' testing practices. We included the 49 studies that compared diagnostic testing practices in intervention and control groups.

DATA EXTRACTION

Two investigators independently reviewed each article in a blinded fashion using a standard data collection form to obtain a methodologic score and to abstract the key elements of each intervention.

DATA SYNTHESIS

On a 38-point methodologic criteria scale, the mean +/- SD score was 13+/-4.4. The desired behavior change was reported in the intervention group in 37 (76%) of 49 studies. Twenty-four (86%) of 28 interventions targeted at many behavioral factors were successful, while 13 (62%) of 21 studies aimed at a single behavioral factor were successful (P=.12).

CONCLUSIONS

A majority of interventions to improve physicians' testing practices reported in the literature claimed success, with interventions based on multiple behavioral factors trending toward being more successful. While methodologic flaws hamper drawing strong conclusions from this literature, application of a behavioral framework appears to be useful in explaining interventions that are successful and can facilitate interpretation of intervention results.