Background: The multiple competing demands of the busy office visit have been shown to interfere with delivery of preventive services. In this study we used physician recommendations for screening mammography to examine the relative importance of physician, patient, and visit characteristics in determining on which patient visits this preventive service will be provided. Methods: Physicians in the Ambulatory Sentinel Practice Network (ASPN) completed a questionnaire describing their knowledge, attitudes, and beliefs about screening mammography. They also described the content of a series of nonacute care visits with women aged 40 to 75 years with regard to making a recommendation when the patient was due for screening mammography. The data were linked, and univariate and multivariate logistic regression methods were used to examine the relative importance of physician, patient, and visit characteristics on making a recommendation for mammography. Results: Ninety-three physicians reported making a recommendation for screening mammography on 53.1% of nonacute visits. When modeling physician, patient, and visit characteristics separately, 70% of the variability in the model is explained by physician characteristics only, 63% by patient characteristics only, and 73% by visit characteristics only. A combined model using all characteristics explained 85% of the variability. Conclusions: Although characteristics of physicians and patients can predict frequency of recommendations for mammography, the specific characteristics of the visit are equally important. Efforts to improve delivery of preventive services in primary care that emphasize physician education and performance feedback are unlikely to increase rates of mammography recommendation. Effective strategies must consider the multiple competing demands faced by patients and physicians during each office visit and seek ways for assisting them in setting rational priorities for services.