The increasing focus on patient safety in the field of health policy is accompanied by research programs that articulate the role of the social sciences as one of contributing to enhancing safety in healthcare. Through these programs, new approaches to studying safety are facing a narrow definition of 'usefulness' in which researchers are to discover the factors that support or hamper the implementation of existing policy agendas. This is unfortunate since such claims for useful involvement in predefined policy agendas may undo one of the strongest assets of good social science research: the capacity to complexify the taken-for-granted conceptualizations of the object of study. As an alternative to this definition of 'usefulness', this article proposes a focus on multiple ontologies in the making when studying patient safety. Through such a focus, the role of social scientists becomes the involvement in refiguring the problem space of patient safety, the relations between research subjects and objects, and the existing policy agendas. This role gives medical sociologists the opportunity to focus on the question of which practices of 'effective care' are being enacted through different approaches for dealing with patient safety and what their consequences are for the care practices under study. In order to explore these questions, this article draws on empirical material from an ongoing evaluation of a large quality improvement collaborative for the care sectors in the Netherlands. It addresses how issues like 'effectiveness' and 'client participation' are at present articulated in this collaborative and shows that alternative figurations of these notions dissolve many 'implementation problems' presently experienced. Further it analyzes how such a focus of medical sociology on multiple ontologies engenders new potential for exploring particular spaces for 'acting with' quality improvement agents.