BackgroundMany health systems have implemented integrated care as an alternative approach to health care delivery that is more appropriate for patients with complex, long-term needs. The objective of this article was to analyse the implementation of integrated care at a German geriatric hospital and explore whether the use of a “context-mechanisms-outcomes”-based model provides insights into when and why beneficial outcomes can be achieved.MethodsWe conducted 15 semi-structured interviews with health professionals employed at the hospital. The data were qualitatively analysed using a “context-mechanisms-outcomes”-based model. Specifically, mechanisms were defined as the different components of the integrated care intervention and categorised according to Wagner’s Chronic Care Model (CCM). Context was understood as the setting in which the mechanisms are brought into practice and described by the barriers and facilitators encountered in the implementation process. These were categorised according to the six levels of Grol and Wensing’s Implementation Model (IM): innovation, individual professional, patient, social context, organisational context and economic and political context. Outcomes were defined as the effects triggered by mechanisms and context, and categorised according to the six dimensions of quality of care as defined by the World Health Organization, namely effectiveness, efficiency, accessibility, patient-centeredness, equity and safety.ResultsThe integrated care intervention consisted of three main components: a specific reimbursement system (“early complex geriatric rehabilitation”), multidisciplinary cooperation, and comprehensive geriatric assessments. The inflexibility of the reimbursement system regarding the obligatory number of treatment sessions contributed to over-, under- and misuse of services. Multidisciplinary cooperation was impeded by a high workload, which contributed to waste in workflows. The comprehensive geriatric assessments were complemented with information provided by family members, which contributed to decreased likelihood of adverse events.ConclusionsWe recommend an increased focus on trying to understand how intervention components interact with context factors and, combined, lead to positive and/or negative outcomes.Electronic supplementary materialThe online version of this article (doi:10.1186/s12913-017-2105-7) contains supplementary material, which is available to authorized users.