BACKGROUND: In October 1995 the University of Michigan Healthcare System initiated a program to develop and implement guidelines for primary care in an effort to improve the quality and cost-effectiveness of care for common conditions associated with wide variations in clinical practice. One of these conditions was Group A beta-hemolytic streptococcus (GABHS), present in 5% to 20% of adults complaining of sore throat. METHODS: A draft guideline was developed on the basis of a theoretical model of sore throat management, local data, and research evidence. The guideline was revised to reflect physicians' beliefs and practices regarding sore throat management. Guideline recommendations depended only on the number of clinical signs experienced by the patient and included testing only if it was likely to provide additional information about the probability of GABHS. Data on pre- and postdissemination data on patients presenting with sore throat were collected. RESULTS: When physicians believed testing or antibiotics were unnecessary, only 7% of patients demanded screening and only 6% of patients wanted antibiotics. Physician beliefs about a patient's need for testing agreed with guideline recommendations in 63% of patients both before and after guideline dissemination. DISCUSSION: Disseminating locally modified, evidence-based guidelines may not be sufficient to produce practice changes. If the guideline had been followed, the amount of testing would have been reduced by 17% and the appropriateness of testing improved for 32% of sore throat patients. The results indicate the need for implementation efforts that go beyond presenting evidence, even when that evidence is from both the literature and the local practice setting.