Background: Research into termination of long-term psychosocial treatment of mental disorders is scarce. Yearly 25% of people in Dutch mental health services receive long-term treatment. They account for many people, contacts, and costs. Although relevant in different health care systems, (dis)continuation is particularly problematic under universal health care coverage when secondary services lack a fixed (financially determined) endpoint. Substantial, unaccounted, differences in treatment duration exist between services. Understanding of underlying decisional processes may result in improved decision making, efficient allocation of scarce resources, and more personalized treatment. Methods: A qualitative study design, according to Grounded Theory principles, was used to understand the decision making process. In four teams in three large Dutch mental health services, 29 multidisciplinary case conferences were observed, and 12 semi-structured interviews were conducted. Results: We describe two constituent elements of decision making: the process through which decision making is prepared and executed, and the substantial factors guiding its outcomes. The first consists of: (1) steps towards a team discussion on treatment termination, (2) team-related factors that influence decisions, and (3) the actual team decision making process. The second consists of factors related to patients, professionals, organization, and wider environment. Our main finding was that discussions of treatment (dis)continuation are highly unstructured. Professionals find it difficult to discuss with patients and teams, team discussion are ad-hoc, and clear decisions are scarce. We offer four explanations: first, long-term treatment lacks golden rules on outcome and process to base decisions on. Second, in the absence of such rules professionals rely on experience but underappreciate their own biases. Third, consequently, professionals aim for decisional consensus, which however is scarce among professionals. Fourth, treatment environments are hardly in favour of changing default (continuation) settings. Conclusion: Clear decision making, and terminating treatment when appropriate, is systematically hampered within secondary mental health services. Since continuation is the 'easy' default option, discontinuation requires skillful and determined navigation of interpersonal negotiations. Given services' scarce means, people's large demands for help, and patients' unused potential autonomy, it is desirable to invest in decision making skills and procedures - both human and economic benefits may be substantial.