OBJECTIVES: Early evaluations of new technologies require knowledge about current care. Guidelines are sometimes used as a proxy. We examined the degree of compliance to guidelines for the care of patients with a recent TIA or minor ischemic stroke. We specifically focussed on the guidelines for carotid stenosis imaging and treatment criteria. METHODS: Semi-structured interviews are being conducted with neurologists and radiologists in 13 hospitals in different regions in the Netherlands. Questions were asked about the use of initial and confirmatory non-invasive imaging tests to assess carotid stenosis (emergency and outpatient clinic), criteria used to perform confirmatory tests and treatment criteria. We also examined the causes of variation and reasons for deviating from the guidelines. RESULTS: To date, we have conducted interviews with specialists at nine hospitals. The Dutch guidelines recommend duplex ultrasonography (DUS) as initial test. If carotid stenosis is >70% for females or 50-69% for males, a confirmatory computed tomography angiography (CTA) or magnetic resonance angiography is advised. We observed that only four of nine hospitals actually follow these guidelines in current practice; the other hospitals use differing test combinations (e.g. CTA only with no confirmatory test; or two DUS tests, followed by confirmatory CTA if DUS results differ). According to clinicians, these differences are caused by varying expertise in performing tests, capacity problems and preferences of vascular surgeons. We observed little variation in treatment criteria, with good correspondence to the guidelines. CONCLUSIONS: If guidelines are not followed, then early economic evaluations using guidelines as the basis to describe current care may lead to incorrect conclusions about the cost-effectiveness of a new technology. Moreover, important practice variation means that the use of just one comparator is unjustified. Assessments of health and economic impact of new imaging technology will therefore be hospital-dependent and require multiple comparisons and scenarios.