Contamination of interventional research is possible through GP membership of more than one division

INTRODUCTION: General practice divisions were promoted in the 1990s to provide support for general practitioners. Membership patterns are not well understood and may have implications for research and health services development. METHODS: Within a postal questionnaire conducted in 1999, we determined self reported membership of divisions. RESULTS: We obtained a 60% response rate (n = 296) from a random sample drawn from all New South Wales GPs. The majority of GPs (n = 204, 69%) belonged to one division. Thirty respondents (10%) did not belong to any division. Fifty-nine GPs (20%) belonged to two or more divisions, women GPs (n = 27, 31%) significantly more than men (n = 32, 16%) (P = 0.002), and GPs with city or metropolitan area practices (n = 52, 24%) were significantly more likely than rural or remote GPs (n = 7, 9%) (P = 0.005) to belong to two or more divisions. DISCUSSION: If divisions are used as the unit of randomisation for interventional research, there is risk of contamination in study design. Articles reporting such trials should acknowledge this. (author abstract)