There is a general trend to work towards a more evidence based medical practice. An increasing number of guidelines are being developed to translate new evidence to medical practice. The implementation of guidelines, however, is a complex process. This thesis compares two methods of continuing medical education (CME). Reimbursement data were used to provide quantitative information to support the methods. We used the regionally organized, typically Dutch, peer review groups (PRGs) of primary care physicians and pharmacists to address professionals. In addition we included patient outcome measures to evaluate the effect of our program patients' well being. Chapter 1 describes the state of the art in the diffusion of innovations, the use of reimbursement data for research, and how to define -and how measure- the quality of prescribing and how to measure patient outcomes. Chapter 2 is about the assessment of the quality and suitability of our data. We describe a systematic tool to test the appropriateness of a given database for specific research questions. It gives insight into data collecting and data quality, relevant for both researchers and readers. We describe in particular the potency and the pitfalls of reimbursement data and express our belief that these data need to be protected and valued more. Chapter 3 describes how we used reimbursement data to assess the quality of prescribing. Chapter 4 compares the prevalence of complaints mentioned by users of anticholinergic antidepressants to a control group of former users in a population of ambulant elderly. In Chapter 5 we compared several rating scales to assess depression, since inclusion criteria are an essential element of good guidelines. Chapter 6 describes qualities of the PRGs in the area. We selected and evaluated characteristics of PRGs that were thought to be relevant for the effect of our outreach program. Chapter 7 describes the intervention we undertook to reduce the prescribing of highly anticholinergic antidepressants in the elderly and its effect on incident prescribing. The intervention was designed following insights usually referred to as academic detailing. We added a group versus individual approach design. Addressing groups may be an important tool to support acceptance of new guidelines. More research in group-learning processes is needed to improve our understanding of CME. Chapter 8 compares the questionnaires of users, living in the intervention and control areas. Chapter 9 includes some final considerations, in which we speculate on what CME might look like in the future. What will be the role of very large medical databases in diffusion of innovations, quality assessment and improvement? What actually is to be defined as optimal prescribing? How do we maintain large medical databases without jeopardizing privacy? It aims at stimulating to think ahead -and to take action- on employing ICT (Information and Communication Technology) for the benefit of health care. Please enjoy my unfinished thoughts.