Clinical workflows are known to be often complex and have to be handled very flexible due to the patients individual anamnesis and state of health. Certain situations require urgent changes of the previously planned process at run time. Some choices to be made in this context depend very much on the data from clinical backend systems. Thus, data and processes cannot be treated independently of each other.
We present an approach for flexible, data centric workflows. It extends the control-flow perspective of a workflow management system with new concepts for handling process adaption at run-time. The approach combines the method of late modeling with declarative concepts and under-specification. Due to constraints on data from clinical backend systems, process adjustment is triggered at certain points of the process and is then performed at runtime.