A population-based epidemiological study on visual and hearing impairment was planned in a random sample of 2100 clients, drawn from a base population of 9012 users of Dutch residential and day-care intellectual disability (ID) services with the whole range of IDs. Stratification was applied for age 50 years and over and Down syndrome. Visual and hearing functions were assessed according to a standardized protocol, in cooperation with regular ophthalmologists and regional audiological centres. Anticipated obstacles in sample collection, random inclusion, informed consent, expertise of investigators, time and costs were eliminated by a careful preparation. However, inclusion and participation were incomplete. In a descriptive retrospective design, we collected data from our study files on inclusion and participation as well as reasons for non-participation, to identify unanticipated obstacles for this kind of research. Consent was obtained for 1660 clients, and 1598 clients participated in the data collection (76% of intended sample of 2100). Inclusion and participation rates were especially lower in community-based care organizations, resulting in unintentional skewing of the sample towards more severe levels of ID. Complete and reliable data to diagnose visual impairment were obtained for 13581598 (85%) and to diagnose hearing impairment for 12371598 participants (77%). Unanticipated obstacles had to do with the quality of coordination within care organizations, with characteristics of screening methods, and with collaboration with the regular health care system. Assessments of visual function were more easy to organize than were those of hearing. Based on our current experience, practical recommendations are given for future multicentre research, especially in community-based settings.