Research Paper
Mining context-aware resource profiles in the presence of multitasking

https://doi.org/10.1016/j.artmed.2022.102434Get rights and content

Highlights

  • Healthcare managers have to allocate their (human) resources efficiently.

  • Context-aware resource profiles group similar resources by considering context.

  • Our method can mine context-aware resource profiles in the presence of multitasking.

  • Our method is demonstrated by a real-life case study at a Dutch hospital.

  • Domain experts acknowledge the practical value of the obtained results.

Abstract

Healthcare organisations are becoming increasingly aware of the need to improve their care processes and to manage their scarce resources efficiently to secure high-quality care standards. As these processes are knowledge-intensive and heavily depend on human resources, a comprehensive understanding of the complex relationship between processes and resources is indispensable for efficient resource management. Organisational mining, a subfield of Process Mining, reveals insights into how (human) resources organise their work based on analysing process execution data recorded in Health Information Systems (HIS). This can be used to, e.g., discover resource profiles which are groups of resources performing similar activity instances, providing an extensive overview of resource behaviour within healthcare organisations. Healthcare managers can employ these insights to allocate their resources efficiently, e.g., by improving the scheduling and staffing of nurses. Existing resource profiling algorithms are limited in their ability to apprehend the complex relationship between processes and resources because they do not take into account the context in which activities were executed, particularly in the context of multitasking. Therefore, this paper introduces ResProMin–MT to discover context-aware resource profiles in the presence of multitasking. In contrast to the state-of-the-art, ResProMin–MT is capable of taking into account more complex contextual activity dimensions, such as activity durations and the degree of multitasking by resources. We demonstrate the feasibility of our method within a real-life healthcare context, validated by medical domain experts.

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 ϑˆ 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)

  • PlacidiL. et al.

    Process mining to optimize palliative patient flow in a high-volume radiotherapy department

    Tech Innov Patient Support Radiat Oncol

    (2021)
  • KovalchukS.V. et al.

    Simulation of patient flow in multiple healthcare units using process and data mining techniques for model identification

    J Biomed Inform

    (2018)
  • NajjarA. et al.

    A two-step approach for mining patient treatment pathways in administrative healthcare databases

    Artif Intell Med

    (2018)
  • Kempa-LiehrA.W. et al.

    Healthcare pathway discovery and probabilistic machine learning

    Int J Med Inform

    (2020)
  • RojasE. et al.

    Process mining in healthcare: a literature review

    J Biomed Inform

    (2016)
  • AhnH. et al.

    Formal approach for discovering work transference networks from workflow logs

    Inform Sci

    (2020)
  • DustdarS. et al.

    Interaction pattern detection in process oriented information systems

    Data Knowl Eng

    (2007)
  • SchönigS. et al.

    A framework for efficiently mining the organisational perspective of business processes

    Decis Support Syst

    (2016)
  • HuangZ. et al.

    Mining association rules to support resource allocation in business process management

    Expert Syst Appl

    (2011)
  • LiuY. et al.

    A semi-automatic approach for workflow staff assignment

    Comput Ind

    (2008)
  • SuriadiS. et al.

    Discovering work prioritisation patterns from event logs

    Decis Support Syst

    (2017)
  • MartinN. et al.

    Detection of batch activities from event logs

    Inf Syst

    (2021)
  • MartinN. et al.

    Retrieving batch organisation of work insights from event logs

    Decis Support Syst

    (2017)
  • MartinN. et al.

    DaQAPO: supporting flexible and fine-grained event log quality assessment

    Expert Syst Appl

    (2022)
  • JanssenswillenG. et al.

    bupaR: enabling reproducible business process analysis

    Knowl-Based Syst

    (2019)
  • McLachlanG.J. et al.

    Model-based clustering

  • McLachlanG.J.

    Model-based clustering

  • MansR.S. et al.
  • McCaugheyD. et al.

    Improving capacity management in the emergency department: a review of the literature, 2000–2012

    J Healthc Manage

    (2015)
  • Smith-DanielsV.L. et al.

    Capacity management in health care services: review and future research directions

    Decis Sci

    (1988)
  • DumasM. et al.

    Fundamentals of business process management

    (2018)
  • RuleA. et al.

    Using electronic health record audit logs to study clinical activity: a systematic review of aims, measures, and methods

    J Am Med Inform Assoc

    (2020)
  • van der AalstW.M.P.

    Process mining: data science in action

    (2016)
  • AbdalkareemZ.A. et al.

    Healthcare scheduling in optimization context: a review

    Health Technol

    (2021)
  • DrakeR.G.

    The nurse rostering problem: from operational research to organizational reality?

    J Adv Nurs

    (2014)
  • AppiceA.

    Towards mining the organizational structure of a dynamic event scenario

    J Intell Inf Syst

    (2018)
  • BurattinA. et al.

    Business models enhancement through discovery of roles

  • FerreiraD.M.R. et al.

    Discovering user communities in large event logs

  • JinT. et al.

    Organizational modeling from event logs

  • NiZ. et al.

    Mining organizational structure from workflow logs

  • PikaA. et al.

    Mining resource profiles from event logs

    ACM Trans Manage Inf Syst

    (2017)
  • van der AalstW.M.P. et al.

    Discovering social networks from event logs

    Comput Support Coop Work

    (2005)
  • YeJ. et al.

    Mining resource community and resource role network from event logs

    IEEE Access

    (2018)
  • View full text