Investigating the application of Declarative Modelling techniques to district nurse workflow processes

Abstract

District nurses perform numerous ad hoc activities, meaning the sequence of their activities depends on the particular situation they are faced with at that moment. As a result, the processes these nurses are involved in are largely unstructured and that leaves a challenge to understand the performed work. To bridge this gap, transcripts of client visits served as input to create different types of process models using event logs generated by Azure OpenAI. Several process models were created, and each model is a unique combination of process modelling notation and granularity level as described in a standardized nursing taxonomy. Next, these models were evaluated and standardized to investigate which model can encapsulate these processes given three quality metrics: Fitness, Precision, and Simplicity. This thesis illustrates the complexity of nursing activities, and describes how AI can play a role in discovering these processes. Additionally, the varying granularity levels of the event log serve as an example of how processes can be discovered on different levels of abstraction, facilitating communication with stakeholders on other levels of involvement. Based on the results, BPMN-D slightly outperforms BPMN on simplicity at the most detailed level, while BPMN performance increases as the abstraction level of activities increases. DECLARE was unable to express rich information about the process based on the event log used in this case study