Research PaperMining context-aware resource profiles in the presence of multitasking
Introduction
Healthcare organisations, e.g. hospitals, are facing critical challenges, most notably increasing and ageing populations and, at the same time, tightening budgets [1], [2]. To cope with these challenges – while securing high-quality care standards – healthcare organisations are becoming increasingly aware of the need to improve their care processes and to manage their scarce resources efficiently [3], [4]. In order to determine suitable levels of resources (e.g. staff, equipment, and facilities [5]) and efficient resource allocation, healthcare managers need a comprehensive understanding of the complex relationship between processes and resources [6], [7].
To uncover the real behaviour of resources that perform activities in processes, the process execution data captured by Health Information Systems (HIS) and Electronic Health Records (EHR) can be used [2], [3], [8]. The events recorded by these systems can be compiled into an event log, which represents the real-life behaviour of a process [2]. Process Mining is a research domain focusing on the (semi-)automatic extraction of insights from event logs [9]. As most of these events are triggered by logging activities performed by human resources (e.g. nurses) on cases (e.g. patients) [3], we can exploit this information to gather extensive insights into the relationship between resources and the activities they perform. These insights provide a comprehensive and transparent overview on the behaviour of resources within a healthcare organisation and aid healthcare managers in more efficiently allocating their resources, e.g. improving the scheduling (i.e. rostering) and staffing (i.e. determining suitable levels) of nurses [10], [11], [12].
In Process Mining, the subfield of organisational mining is concerned with discovering organisational structures and social networks from event logs [13]. Several resource profiling techniques – i.e. finding groups of resources that perform similar activity instances – have been proposed [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23]. Nevertheless, most existing algorithms only consider a brief description of the activities that resources performed (e.g. “Create purchase requisition”, “Send invoice”, “Administer medication”, etc.) as a starting point. Thereby, the context is ignored, i.e. the circumstances in which the activity was executed [24]. In addition, compared to common business processes, such as the order-to-cash process, healthcare processes are generally more knowledge-intensive [3], [25], [26] and typically exhibit a higher degree of variability [26], [27], due to the involvement of knowledge workers, such as physicians and nurses, and complex ad hoc decision-making [3], [25]. For example, depending on the patient’s health condition, a more experienced or specialised senior nurse may be assigned take care of the patient. However, the activities as such (e.g. the activity labels recorded in the HIS) are the same regardless of the patient’s condition. Therefore, this limiting assumption can hide essential nuances on how resources conduct their tasks in a real-life healthcare setting, which highlights the importance of considering the context in which the activities were executed besides the activity labels. This context can be regarded as a multi-dimensional concept describing “who did what under which circumstances” [24]. Straightforward dimensions that can be considered include time-related attributes of the activity instance, e.g. the weekday, the time of the day (e.g. morning, afternoon, or evening), or the duration of the activity; case attributes, such as the case type or status; and a resource identifier [24]. However, additional dimensions can be added to capture more complex aspects which are not directly observable in the event log, such as the degree of multitasking by resources. It is not trivial to consider multitasking using existing algorithms because these cannot handle attributes of mixed types simultaneously, e.g. nominal (activity labels), discrete (number of concurrently performed activities), and continuous (duration of activities).
This paper extends our previous work on discovering context-aware resource profiles from event logs using ResProMin [24]. While ResProMin considers contextual variables (expressing the conditions under which activity instances were executed) when discovering resource profiles, the method cannot incorporate the multitasking behaviour of resources, which is particularly common in the healthcare sector [28]. Therefore, we introduce an extension to our previous work: ResProMin–MT (Resource Profile Miner–Multitasking). In addition to capturing the multitasking behaviour of resources, we also demonstrate how the context can be further defined by considering the activity duration. Both dimensions have not been considered before in resource profile identification. Moreover, whereas ResProMin was demonstrated using a public event log of a municipal service, we evaluated ResProMin–MT on a real-life case study in a healthcare context, more specifically, nursing. This also enabled us to present and discuss our findings with domain experts in nursing science at the hospital in order to validate the benefits of ResProMin–MT for healthcare managers in decision-making.
The remainder of this paper is structured as follows. Section 2 provides an overview of the related work. Section 3 introduces ResProMin–MT. Next, an introduction to the case study is provided in Section 4, an overview of the results in Section 5, and a discussion and evaluation of our method in Section 6. The paper ends with a conclusion and directions for future work in Section 7.
Section snippets
Related work
This work is related to Process Mining applied in healthcare on the one hand and the resource perspectives in Process Mining on the other hand. The following sections provide an overview of related work in these domains.
Method
In this section, we present ResProMin–MT1 as an extension of our previous work [24], which enables discovering context-aware resource profiles from event logs by taking into account additional contextual factors of activity executions, such as activity durations and multitasking. A general overview of ResProMin–MT is visualised in Fig. 1. Our method consists of three steps. In the first step, we
Case study
In order to demonstrate ResProMin–MT, a real-life case study is conducted at a large University Medical Centre in the Netherlands. This section provides an overview of the dataset that has been used, together with the context in which the data has been recorded and the preprocessing that has been applied. Next, we discuss how the method has been operationalised to derive context-aware resource profiles.
Results
The results of the estimated vector of model parameters of the FMM in Step 2 are shown in Table 7, Table 8, Table 9. In Table 7a, the intra-cluster multinomial distributions for Ward are displayed. For instance, 89.15% of the multitask sessions that belong to cluster 1 were performed in the Geriatrics ward, whereas in cluster 4, the sessions took place in Transplantations (42.40%), Neurology (22.77%), Internal med-system diseases (17.90%), and Geriatrics (11.55%) in decreasing probability.
Discussion
As the work and role of nurses are rather dynamic, there are no formalised nurse roles defined by the hospital. Nevertheless, healthcare organisations could benefit from having a transparent overview of the different “roles”, or more specifically, resource profiles of nurses working on similar activity instances. These in-depth insights could be used, e.g., to improve nurses’ scheduling and determine the suitable levels of required nurses. In Sections 4 Case study, 5 Results we demonstrated how
Conclusion
In this paper, we introduced a method to discover context-aware resource profiles from event logs in the presence of multitasking, i.e. ResProMin–MT. In addition, we demonstrated that our method is capable of taking into account more complex activity dimensions, such as durations. Despite the challenges that arose from the data, we demonstrated the feasibility of ResProMin–MT in a healthcare context. The output of our method was validated by domain experts in nursing science. The insights
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
The authors would like to thank Tim Korteland, Prof. Dr. Erwin Ista, and Prof. Dr. Monique van Dijk of the Erasmus University Medical Center Rotterdam, Department of Internal Medicine, division of Nursing Science for their time to assess and validate our findings.
This study was supported by the Special Research Fund (BOF) of Hasselt University under Grant No. BOF19OWB20, Belgium.
The resources and services used in this work were provided by the VSC (Flemish Supercomputer Center), funded by the
References (87)
- et al.
Applying lean principles to the design of healthcare facilities
Int J Prod Econ
(2015) - et al.
Recommendations for enhancing the usability and understandability of process mining in healthcare
Artif Intell Med
(2020) - et al.
Best practices in business process redesign: an overview and qualitative evaluation of successful redesign heuristics
Omega
(2005) - et al.
How many nurses do we need? A review and discussion of operational research techniques applied to nurse staffing
Int J Nurs Stud
(2019) - et al.
Towards comprehensive support for organizational mining
Decis Support Syst
(2008) - et al.
OrdinoR: a framework for discovering, evaluating, and analyzing organizational models using event logs
Decis Support Syst
(2022) - et al.
Business process analysis in healthcare environments: a methodology based on process mining
Inf Syst
(2012) - et al.
Improving our understanding of multi-tasking in healthcare: drawing together the cognitive psychology and healthcare literature
Applied Ergon
(2017) - et al.
Discovering role interaction models in the emergency room using process mining
J Biomed Inform
(2018) - et al.
Process mining in healthcare – an updated perspective on the state of the art
J Biomed Inform
(2022)
Process mining to optimize palliative patient flow in a high-volume radiotherapy department
Tech Innov Patient Support Radiat Oncol
Simulation of patient flow in multiple healthcare units using process and data mining techniques for model identification
J Biomed Inform
A two-step approach for mining patient treatment pathways in administrative healthcare databases
Artif Intell Med
Healthcare pathway discovery and probabilistic machine learning
Int J Med Inform
Process mining in healthcare: a literature review
J Biomed Inform
Formal approach for discovering work transference networks from workflow logs
Inform Sci
Interaction pattern detection in process oriented information systems
Data Knowl Eng
A framework for efficiently mining the organisational perspective of business processes
Decis Support Syst
Mining association rules to support resource allocation in business process management
Expert Syst Appl
A semi-automatic approach for workflow staff assignment
Comput Ind
Discovering work prioritisation patterns from event logs
Decis Support Syst
Detection of batch activities from event logs
Inf Syst
Retrieving batch organisation of work insights from event logs
Decis Support Syst
DaQAPO: supporting flexible and fine-grained event log quality assessment
Expert Syst Appl
bupaR: enabling reproducible business process analysis
Knowl-Based Syst
Model-based clustering
Model-based clustering
Improving capacity management in the emergency department: a review of the literature, 2000–2012
J Healthc Manage
Capacity management in health care services: review and future research directions
Decis Sci
Fundamentals of business process management
Using electronic health record audit logs to study clinical activity: a systematic review of aims, measures, and methods
J Am Med Inform Assoc
Process mining: data science in action
Healthcare scheduling in optimization context: a review
Health Technol
The nurse rostering problem: from operational research to organizational reality?
J Adv Nurs
Towards mining the organizational structure of a dynamic event scenario
J Intell Inf Syst
Business models enhancement through discovery of roles
Discovering user communities in large event logs
Organizational modeling from event logs
Mining organizational structure from workflow logs
Mining resource profiles from event logs
ACM Trans Manage Inf Syst
Discovering social networks from event logs
Comput Support Coop Work
Mining resource community and resource role network from event logs
IEEE Access
Cited by (4)
Preface: Special Issue on Knowledge Representation and Reasoning for Healthcare Processes
2023, Artificial Intelligence in MedicineThe Interplay Between High-Level Problems and the Process Instances that Give Rise to Them
2023, Lecture Notes in Business Information Processing