Objective To describe how clinicians use decision aids. Background A 98-patient factorial-design randomized trial of the Statin Choice decision vs. standard educational pamphlet; each participant had a 1:4 chance of receiving the decision aid during the encounter with the clinician resulting in 22 eligible encounters. Design Two researchers working independently and in duplicate reviewed and coded the 22 encounter videos. Setting and participants Twenty-two patients with diabetes (57% of them on statins) and six endocrinologists working in a referral diabetes clinic randomly assigned to use the decision aid during the consultation. Main outcome measures Proportion and nature of unintended use of the Statin Choice decision aid. Results We found eight encounters involving six clinicians who did not use the decision aid as intended either by not using it at all (n = 5; one clinician did use the decision aid in three encounters), offering inaccurate quantitative and probabilistic information about the risks and benefits of statins (n = 2), or using the decision aid to advance the agenda that all patients with diabetes should take statin (n = 1). Clinicians used the decision aid as intended in all other encounters. Conclusions Unintended decision aid use in the context of videotaped encounters in a practical randomized trial was common. These instances offer insights to researchers seeking to design and implement effective decision aids for use during the clinical visit, particularly when clinicians may prefer to proceed in ways that the decision aid apparently contradicts. © 2009 Blackwell Publishing Ltd.