Multi-perspective process mining for emergency process

Pre-medical emergency team activations - Patient characteristics, outcomes and predictors of deterioration

BACKGROUND

Pre-medical emergency team (MET) calls are an increasingly common tier of Rapid Response Systems, but the epidemiology of patients who trigger a Pre-MET is not well understoof.

OBJECTIVES

This study aims to examine the epidemiology and outcomes of patients who trigger a pre-MET activation and identify risk factors for further deterioration.

METHODS

This is a retrospective cohort study of pre-MET activations in a university-affiliated metropolitan hospital in Australia, between 13 April 2021 and 4 October 2021. A multivariable regression model was used to identify variables associated with further deterioration, defined as a MET call or Code Blue within 24 h of pre-MET activation.

RESULTS

From a total of 39 664 admissions, there were 7823 pre-MET activations (197.2 per 1000 admissions). Compared to inpatients that did not trigger a pre-MET, the patients were older (68.8 vs 53.8 years, p < 0.001), were more likely to be male (51.0 vs 47.6%, p < 0.001), had an emergency admission (70.1% vs 53.3%, p < 0.001), and were under a medical specialty (63.7 vs 54.9%, p < 0.001). They had a longer hospital length of stay (5.6 vs 0.4 d, p < 0.001) and higher in-hospital mortality (3.4% vs 1.0%, p < 0.001). A pre-MET was more likely to progress to a MET call or Code Blue if it was activated for fever, cardiovascular, neurological, renal, or respiratory criteria (p < 0.001), if the patient was under a paediatric team (p = 0.018), or if there had been a MET call or Code Blue prior to the pre-MET activation (p < 0.001).

CONCLUSION

Pre-MET activations affect almost 20% of hospital admissions and are associated with a higher risk of mortality. Certain characteristics may predict further deterioration to a MET call or Code Blue, suggesting the potential for early intervention via clinical decision support systems.

A framework for multi-perspective process mining into a BPMN process model

Process mining is mainly focused on process discovery from control perspective. It is further applied to mine the other perspectives such as time, data, and resources by replaying the events in event logs over the initial process model. When process mining is extended far beyond discovering the control-flow models to capture additional perspectives; roles, bottlenecks, amounts of time passed, guards, and routing probabilities in the process can be identified. This is a such extensions are considered under the topic of multi-perspective process mining, which makes the discovered process model more understandable. In this study, a framework for applying multi-perspective process mining and creating a Business Process Modelling Notation (BPMN) process model as the output is introduced. The framework, which uses a recently developed application programming interface (API) for storing the BPMN Data Model which keeps what is produced from each perspective as an asset into a private blockchain in a secure and immutable way, has been developed as a plugin to the ProM tool. In doing so, it integrates a number of techniques for multi-perspective process mining in literature, for the perspectives of control-flow, data, and resource; and represents a holistic process model by combining the outputs of these in the BPMN Data Model. In this article, we explain technical details of the framework and also demonstrate its usage over a case in medical domain.