BPMN in Healthcare Challenges and Best Practices

Information Systems 107 (2022) 102013
Contents lists available at ScienceDirect
Information Systems
journal homepage: www.elsevier.com/locate/is
BPMN in healthcare: Challenges and best practices

∗ ∗
Luise Pufahl a , , Francesca Zerbato b , , Barbara Weber b , Ingo Weber a
a
Software and Business Engineering, Technische Universitaet Berlin, Berlin, Germany
b
Institute of Computer Science (ICS-HSG), University of St. Gallen, St. Gallen, Switzerland
article info a b s t r a c t
Article history: The design and analysis of process models is a critical factor for organizational improvement across
Received 7 July 2020 various industries. Thanks to its potential to enable common understanding and foster automation,
Received in revised form 30 December 2021 process modeling is increasingly adopted in the healthcare sector. However, the complexity of the
Accepted 16 February 2022
healthcare domain makes process modeling a challenging task, potentially explaining the modest
Available online 23 February 2022
uptake of process modeling standards like the Business Process Model and Notation (BPMN). In this
Recommended by Manfred Reichert
paper, we identify common challenges of process modeling in healthcare, elicited from healthcare
Keywords: process modeling initiatives and supported by the literature. For each challenge, we present some
BPMN BPMN best practices in the form of ready-to-use process fragments that guide the standard modeling
Best practices of complex healthcare aspects. Also, we report the results of a first evaluation of the use and perceived
Healthcare usefulness of best practices conducted with junior experts in medicine and IT. We observed that the
Modeling challenges domain-specific process fragments help to capture healthcare aspects in detail and are perceived as
Process modeling
a source of learning, turning out to be especially useful for modelers with a basic understanding of
Standard-based design
BPMN and the healthcare domain.
© 2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
1. Introduction variability [3]. Finally, process models allow different kinds of

process analyses [6], and serve as a blueprint for the automation of
Within the industry, healthcare is one of the fastest-growing clinical and organizational activities and information flows [2,7].
sectors, driven by the enactment of complex and dynamic pro- The leading standard for process modeling is the Business
cesses targeting optimal patient outcomes and always seeking Process Model and Notation (BPMN), overseen by the Object
increased effectiveness and efficiency [1]. To face the growing de- Management Group (OMG), featuring a graphical notation aimed
mand for assistance and technological innovation, care providers to be ‘‘readily understandable by all business users’’ [8]. BPMN
are increasingly resorting to business process management (BPM) allows defining process diagrams at different levels of abstrac-
initiatives to systematically analyze and (re-)design their pro- tion [9], which can be used for documentation purposes and
support implementation efforts. Besides, BPMN is supported by a
cesses and streamline care delivery, reduce costs, and increase
wide range of modeling tools and benefits from the availability of
quality [2].
professional and academic training opportunities [10]. Moreover,
In particular, process modeling is more and more integrated
BPMN can be complemented by other OMG standards, such as
into healthcare management routines thanks to its potential to
the Decision Model and Notation (DMN) [11], which can be used
enable common understanding among different stakeholders, fos- to capture information and decision-making aspects typical of
ter digital transformation, and enhance care delivery [3,4]. The healthcare domains [12].
use of process models in healthcare brings manifold benefits. First Nevertheless, albeit the rich expressiveness of the standard,
of all, graphical representations of processes serve as an intuitive BPMN-based modeling approaches have seen a modest uptake
and more immediate reference for training and communicating in the healthcare domain, and the adoption of BPM approaches
with health professionals as they are easier to grasp and less still lags behind compared to other sectors [13]. This trend can be
ambiguous than textual documents [5]. Secondly, they support partially explained by the inherent complexity of healthcare pro-
the standardization of clinical procedures and decision-making, cesses [14], the highly regulated hospital environments, and the
thus fostering compliance with shared protocols and minimizing relatively slow adoption of IT technologies [13], which contribute
to increased modeling complexity [15]. A possible direction to
∗ Corresponding authors. improve the uptake of BPMN in healthcare contexts is to propose
E-mail addresses: [email protected] (L. Pufahl), domain-specific extensions, such as those in [4,16–19]. However,
[email protected] (F. Zerbato), [email protected] (B. Weber), such extensions need to come to terms with a lack of comprehen-
[email protected] (I. Weber). sive evaluation and little or no support by modeling tools [20]. For
https://doi.org/10.1016/j.is.2022.102013
0306-4379/© 2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
L. Pufahl, F. Zerbato, B. Weber et al. Information Systems 107 (2022) 102013
this reason, in this paper, we focus on the BPMN standard and good design practices and supporting a shared understand-
on unraveling its suitability for modeling relevant yet complex ing among different groups of stakeholders. BPMN and
healthcare aspects. DMN are rich and expressive modeling languages, such
In line with literature [21], we define a healthcare process as that the goal of this work is to provide process modeling
‘‘a set of medical and organizational activities that are performed support with the modeling concepts already included in
in coordination to provide medical care for one or more patients’’. the standards. The advantage of such an approach is that
Healthcare processes have some specific characteristics that pose existing BPMN and DMN modeling and execution tools
unique modeling challenges [14,22,23]. Among others, we focus already support these concepts, without the need for ex-
on the following specifics: tensions. The proposed best practices are meant to be used
by health workers with a basic understanding of BPMN
• Patient-centered. The key entity on which activities are exe- and process modelers and business analysts working for
cuted is usually not a document or an item. It is a patient –
healthcare organizations. Indeed, the best practices may
a human being. He or she needs to be actively involved in
also help experienced modelers to come up with ideas or
medical decision-making and therapeutic activities and may
explain a particular concept [34].
require individualized care [24,25].
3. First evaluation of the use and perceived usefulness of
• Time-critical. Diagnostic and treatment processes need to be
the BPMN best practices through an exploratory study
scheduled based on patient conditions, resource availability,
where junior experts in medicine and IT solve modeling
and hospital regulations, and need to respect several time
tasks with and without the help of the best practices. The
constraints [23,26,27].
results of our study showed that, with the help of the
• Decision-intensive. Decision-making builds upon medical
best practices, participants were able to capture (complex)
knowledge and requires health workers to consider clini-
medical aspects in detail and to obtain process models con-
cal evidence and available patient information [22]. Such
taining fewer errors. Besides, the best practices have been
information is stored not only in IT systems but also in
perceived as useful to increase the general understanding
paper documents and medical devices, thus complicating
of BPMN.
the understanding and automation of information exchange
and transformation tasks [3,12,13]. The rest of this paper is structured as follows. Section 2 pro-
• Multidisciplinary. For providing the best possible treatment vides a brief introduction to the BPMN and DMN standards,
for a patient, different medical disciplines and organiza- which the expert reader may skip. Section 3 discusses related
tional units need to be integrated and coordinated [16,22]. work. Section 4 describes the research method followed to iden-
• Resource-intensive. Medical specialists, equipment, and tify relevant challenges and the related best practices. Section 5
rooms are usually scarce and expensive resources, shared summarizes recurrent challenges in healthcare process models
among different departments and, thus, need to be properly and the suggested best practices. Section 6 describes the study
managed and optimized [22,23]. conducted to evaluate the best practices and reports our findings.
Finally, Section 7 discusses the results and limitations of this
For all these reasons, we advocate that healthcare process
work, and Section 8 summarizes our contribution and discusses
modeling tasks require systematic guidance in terms of mod-
future research directions.
eling tools and methods, especially to support shared under-
standing by different stakeholders. Indeed, process modelers of-
2. Introduction to the BPMN and DMN standards
ten lack advanced healthcare knowledge, while domain experts
may perceive the richness of the BPMN language as a draw-
The Business Process Model and Notation (BPMN) [8] and
back [28] and may benefit from a demonstration by concrete
the Decision Model and Notation (DMN) [11] are among the
examples [12,20]. Building upon existing literature [3,13,16,22,
most widespread process and decision modeling languages. They
29–31] and our experience in the field, in this paper, we aim
aim to provide highly understandable models for business users,
to support the modeling of healthcare processes in BPMN by
which can also be used for implementation purposes. Based on
proposing best practices to model common yet challenging as-
the modeling goal, process models exhibit different levels of
pects of healthcare processes. Supporting process modeling ini-
abstraction with regard to the underlying business process [9]. On
tiatives with reusable process fragments is a common practice, as
an organizational level, the high-level textual description of a pro-
shown in a literature overview by Haddar et al. [32]. Furthermore,
cess including the process inputs, outputs, and responsibilities are
domain-specific modeling patterns have already been proposed
given. An operational process model captures activities, their rela-
in the BPM field, e.g., patterns for ordering and manufacturing
tion as well as organizational information. The operational level
processes [33]. However, to the best of our knowledge, no such
is sub-divided by BPMN, in descriptive models, for a high-level
proposal has focused on healthcare.
documentation, and analytical models, to analyze the process in
Following this research direction, the goal of the presented
detail. Based on this, an implemented process model extends and
work is to increase the applicability of BPMN for healthcare pro-
adapts it with technical aspects needed for the implementation.
cess modeling, considering the digital transformation of health-
This section aims to quickly introduce both standards with a
care as a long-term goal. In line with this goal, this work provides
clinical example. The focus is more on BPMN because our focus
the following main contributions:
is mostly on process modeling, and DMN includes a smaller set
1. Structured elicitation of relevant modeling needs and of modeling constructs. Although DMN will be only applied in
challenges in healthcare processes. These were observed one of the best practices in this research work, an introduction is
in a set of healthcare process modeling initiatives con- included below due to its relevance and to make the paper self-
ducted in different health institutions, and their practical contained. More details about the semantics of specific BPMN
relevance was confirmed by literature. and DMN constructs will be given in Section 5 when introduc-
2. Definition of a set of one or more BPMN best practices ing the best practices and can be easily found in the standard
for each identified challenge in the form of ready-to-use specifications [8,11].
process fragments. These fragments should guide process Fig. 1 shows a simplified version of an emergency care process,
designers in capturing (challenging) healthcare process as- which is triggered by a worsening of the patient’s conditions. As
pects at different levels of process abstraction, encouraging a first task, the physician assesses the conditions of the patient.
2
Fig. 1. Simplified emergency care process shown as a BPMN process diagram.
Then, while he or she performs the physical examination, a nurse

draws blood from the patient. The blood samples are then sent
to the clinical laboratory for analysis and, once ready, results are
sent back to the emergency department. With the notes from the
examination and the blood analysis results, the physician has to
make a diagnosis. Based on the results, a surgical intervention is
either immediately performed by the physician or it is planned
for the following days by the nurse, keeping in mind that the pa-
tient may also refuse to undergo surgery. In all cases, the patient
is discharged while being provided with discharge documents.
BPMN diagrams can represent the control flow, data flow, and Fig. 2. DMN Decision Requirement Graph representing the decision Make
diagnosis made in the corresponding business rule task of Fig. 1.
organizational aspects of a process with the following concepts.
• BPMN process diagrams can represent the control flow of a

process with activities (e.g., Perform physical examination), For describing in more detail how the Make diagnosis decision
relevant events (e.g. Receive results), and sequence flows is made, a DMN decision model such as the one shown in Fig. 2
connecting them. Gateways, which depicted as diamonds, can be associated with the BPMN process model. A DMN model
are used to control sequence flows and to define parallel and includes a decision requirements graph (DRG), which allows one
alternative behaviors. to describe the decision-making at a higher abstraction level. The
• Activities can be atomic (tasks) or compound (sub- DRG in the provided example shows that for the Make diagnosis-
processes). Tasks (e.g., Perform physical examination) repre- decision, the Blood analysis results and the Medical notes are
sent atomic units of work, while sub-processes represent needed input data, and the ER guideline supports the decision-
compound activities, which can be collapsed to hide details making as a knowledge source. For each decision in a DRG, the
(e.g., Surgical intervention). BPMN features different types underlying decision logic can be specified either by a decision
of tasks having different inherent behavior. For example, table or with a Friendly Enough Expression Language expres-
business rule tasks are used to represent decision-making sion [11], a DMN-specific expression language for decision logic.
tasks (e.g., task Make diagnosis). Such a task can be linked to We will focus on decision tables in this work.
a DMN model, which captures the decision logic, i.e., how
to make the decision. 3. Related work
• Relevant external events are shown as BPMN events (e.g.
Worsening of patient’s conditions), which capture process In this section, we discuss related work in the context of
triggers to which the process reacts such as messages, sig- healthcare process modeling and touch upon related work about
nals, conditions (e.g., Worsening of patient’s conditions, ex- the use of BPMN best practices in general. In healthcare working
ceptions (e.g., Patient refused surgery), and outputs produced environments, a wide range of processes with different features
by the process (e.g., Blood analysis requested). and requirements is managed and executed daily [31]. Such pro-
• Relevant data and documents used or produced by process cesses usually require the structuring and logical coordination of
activities can be represented with the help of data objects interlinked clinical and administrative tasks [22], posing different
(e.g., Medical notes) or data stores when the considered challenges for IT support. To improve the understanding of such
data is persistent. Text annotations can be attached to con- challenges, Lenz and Reichert [22] distinguish between organi-
trol flow elements to enhance their description, such as zational processes, which focus on the coordination of different
the annotation ‘‘Prefer laparoscopic [...]’’ associated with the healthcare professionals and organizational units, and medical
sub-process Surgical intervention. treatment processes having the patient as their main focus. This
• Interactions between organizations can be shown through categorization is often recalled in the literature to classify health-
the concept of pools (each one owning a process) (e.g., the care process modeling languages and approaches proposed over
Emergency department and the clinical laboratory) and a mes- the years by the BPM and medical informatics communities [7,
sage flow between them. 31,35].
• Lanes in pools (e.g., the Nurse and Physician) can be used to In Section 3.1, we introduce research on computer-inter-
represent different roles or systems executing the included pretable languages and the modeling of clinical practice guide-
activities. lines (CPGs) [36,37]. Then, in Section 3.2, we review approaches
3
for modeling processes enacted within healthcare organizations, includes an overview of the main challenges posed by health-
with a focus on BPMN. care processes. The authors give high-level insights into typical
challenges in modeling healthcare processes, such as the need to
3.1. Healthcare processes and computer-interpretable guidelines support shared activities and multidisciplinary tasks. Müller and
Rogge-Solti [16] have selected one specific challenge and discuss
Over the past decades, research in medical informatics has put in their work different options to facilitate the shared working
forward approaches for computer-interpretable guidelines (CIGs) behavior with BPMN (and conservative extension of BPMN) in
to improve the accessibility and maintainability of narrative CPGs healthcare processes. Zerbato et al. [51] use BPMN diagrams to
and to support the implementation of (guideline-based) decision model clinical pathways for catheter-related bloodstream infec-
support systems [37]. CIGs allow to represent and execute clinical tions and focus thereby on representing temporal constraints,
guidelines over patient-specific clinical data [37], being particu- another relevant aspect of healthcare processes that is difficult
larly suitable for modeling medical treatment processes. Among to capture. Evidence-based decision-making, also a challenging
them, approaches based on task-network models (e.g., Asbru [38], aspect, is studied in [29]. This work lays the groundwork for the
EON [39], GLARE [40], GLIF3 [41], and ProForma [42]) allow one methodology in [30] which suggests combining BPMN and DMN
to formally represent and execute CPGs through an execution to represent decision-intensive healthcare processes coming from
engine, and explicitly support the management of temporal as- chronic care.
pects. From a control-flow perspective, CIG-based approaches Overall, the introduced studies remark the need to model
based on task-network models share similarities with impera- healthcare processes in a graphical and standard way to pro-
tive process modeling languages [31,43]. However, in contrast mote common understanding among the different stakeholders
with general-purpose, standard BPM techniques, they are mainly involved in the collaborative process design and support model
based on specialized modeling formalisms and execution en- sharing and reuse [52]. However, although some of these works
vironments [31], which, together with the complexity deriving partially pick up challenges of using BPMN in healthcare pro-
from their local adaption, seems to have limited their broad cesses and provide initial solutions, a comprehensive overview of
adoption [37]. It is beyond the scope of this paper to provide challenges and related solutions is still missing.
a complete overview of such approaches. The interested reader (2) BPMN extensions for healthcare. Some domain-specific BPMN
may consult the systematic comparison by Peleg [37] and the extensions have been introduced to improve the modeling rele-
recent work by Gatta et al. [44] for more details on CIGs and their vant aspects of clinical domains [4,17–19]. Braun et al. [18] in-
synergy with BPM methods. troduce BPMN4CP, a BPMN extension incorporating concepts re-
trieved from the context of clinical pathways to enhance
3.2. Application of BPM techniques in healthcare evidence-based decision activities related to the considered do-
main. New task types are added to model medical diagnoses and
In the BPM field, standard-based methods have been devel- therapies, while different data object types are used for repre-
oped to support the modeling and enactment of healthcare pro- senting documents related to medical and administrative proce-
cesses [3,30], mainly focusing on pre-specified and repetitive dures. Neumann et al. [17] introduce BPMNSIX as an enhanced
routines requiring standardization. For example, the UML-based modeling language for the design and execution of surgical pro-
methodology by Ferrante et al. [3] is driven by the need to inte- cesses in an integrated operating room. BPMNSIX incorporates
grate medical processes and health information systems. Instead, extension elements taken from BPMN4CP and includes new con-
the work by Combi et al. [30] addresses the integrated design cepts for representing surgical activities at different levels of
and enactment of decision-intensive care pathways in BPMN. granularity, as well as anatomical structures and medical instru-
Vanwersch et al. [45] build upon best practice process models mentation. In [4], BPMNSIX is extended with the IEEE 11073 SDC
to introduce a systematic redesign technique for care processes. family of standards for the interoperability of networked med-
Chiao et al. [46] propose an object-based approach based on the ical devices to enable rule-based medical device orchestration.
PHILharmonicFlows framework to coordinate interactions among Finally, Onggo et al. [19] propose BPMN4SIM an extension of
healthcare processes based on business objects. the BPMN analytic conformance subclass supporting the explicit
Although BPM methods for the use in healthcare exist and representation of tasks shared by multiple process instances,
are helpful to represent complex and multidisciplinary proce- queues, and attributes and data-driven decision points.
dures, their application for healthcare process modeling is still The introduced extensions remark that BPMN cannot straight-
modest [13]. In the following, we review studies supporting the forwardly represent specific aspects of healthcare processes.
process modeling in healthcare considering approaches (1) using BPMN extensions are one possible approach to model challeng-
BPMN and other process modeling languages for healthcare, and ing aspects of healthcare processes through the definition of
those (2) defining healthcare-specific extensions of BPMN. healthcare-specific solutions. Still, BPMN extensions have the
disadvantage of lacking support from existing modeling tools,
(1) BPMN for healthcare process modeling. Several studies focus which adds to the lack of precise execution semantics needed
on analyzing the benefits of BPMN for healthcare by applying it for process implementation [20], making them difficult to use in
to specific medical contexts, as discussed next. Rolon et al. [47] practice.
use BPMN to capture programmed surgical patient processes to Building upon the introduced research and observed health-
visualize, understand, and improve them. Rojo et al. [48] ana- care practice, this paper aims to pick up challenges common to
lyze the use of BPMN for the modeling of anatomic pathology various healthcare settings and provide a set of native BPMN
processes. Scheuerlein et al. [5] present a pilot project aimed process fragments serving as best practices to support healthcare
to identify the advantages of combining BPMN with tangible process modeling. Thus, our work is also related to approaches
business process modeling [49] for the design of clinical pathways providing modeling support for process design, e.g., through the
for colon and rectum carcinomas. In [50], BPMN is used to model reuse of modeling patterns and best practices derived from em-
standard operating pathways, to analyze the challenges related to pirical evidence [32]. In the BPM field, there has been extensive
operating room planning and scheduling. research on the definition of patterns to ease the comparison
One of the first works studying the use of BPMN for healthcare of process-aware modeling languages and systems [27,53–55].
process modeling in general is presented by Ruiz et al. [24], which The most notable example is the workflow patterns [53], which
4
Fig. 3. Design science process adapted from Peffers et al. [58] and its application in this research work.
put a strong emphasis on the syntactical structure of process practice framework, was evaluated in the fifth step through a
models. Thom et al. [55] proposed the Activity Patterns, a set modeling experiment with junior medical and IT professionals.
of seven recurring patterns describing business functions de- The exploratory study compares the treatment against a set-
rived from real-world process models. However, as opposed to ting without any modeling support, and it analyzes its use and
the best practices proposed in this paper, these patterns are perceived usefulness (cf. Section 6).
language- and domain-independent, i.e., they do not focus on In Section 4.1, we describe how we derived an initial set of
the content of business activities within a particular applica- challenges and the corresponding solutions from process elicita-
tion domain [33]. Another stream of research has emphasized tion and modeling projects. Then, in Section 4.2, we report on the
the support of domain-specific process modeling. For example, process we followed to select the final set of challenges and best
Koschmider and Reijers [33] have proposed Domain Process Pat- practices based on their relevance in the literature.
terns to foster reuse in the ordering and manufacturing domain,
while Scholta et al. [56] have put forward the RefPA method for 4.1. Identification of challenges and best practices from real-world
detecting and representing best practices in the public adminis- projects
tration domain.
While our work also aims to support process modelers with Since we are interested in challenges having practical implica-
best practices, similar to [33], we specifically start from recur- tions for healthcare process modeling, we considered real process
ring modeling challenges and related solutions in the healthcare modeling initiatives as our primary data source. In detail, we
domain. We propose to use BPMN as modeling language given the considered different healthcare process elicitation and model-
significant uptake of the standard in business domains and the ing projects in which we were involved as modeling experts or
availability of training opportunities and tool support that could consultants. The projects, which are summarized in Table 1, had
also benefit healthcare settings. different goals and were conducted from 2015 to 2018 in various
healthcare organizations, mostly university hospitals located in
Germany, Italy, Spain, and the Netherlands. As a result of these
4. Research method
projects, 38 process models were designed, each one comprising
an average of 15.4 activities and 5.5 roles.
This research work complies with the principles of design sci-
The challenges were identified as follows. First, we created
ence research [57,58] and follows the steps of the design science
a list of 42 individual challenges considering all the challenges
process as described by Peffers at al. [58], i.e., a procedure to
mentioned in the projects. Our focus was on modeling challenges
identify real-world problems and create useful research artifacts
concerning internal healthcare processes to understand, analyze,
for solving them and, consequently, contributing to the body of compare, or redesign the healthcare process as a first step of dig-
scientific knowledge. Fig. 3 outlines the application of the design ital transformation. Thus, we did not consider challenges related
science process in this work. to inter-organizational clinical processes or model-driven pro-
In line with the design science guidelines of problem relevance cess implementation. In detail, for each challenge, we collected
by Hevner et al. [57], we observed in the first step in several information about the main process, the medical department
process modeling initiatives, in which we were involved, as well involved, the challenge description, and the details about the
as in the literature [16,18] that even advanced process modelers proposed solutions, if available. Then, since we observed that
face challenges while modeling healthcare processes due to their challenges were recurring, we grouped challenges addressing
specific characteristics. This is described in Section 1 as well as similar modeling aspects into more abstract categories. For ex-
in Section 3. After problem identification and motivation, in the ample, challenges described as ‘‘Aftercare needs to last at least 30
second step, we defined the solution objectives. As presented in min’’ and ‘‘Examination shall be prescribed within 30 days from
this section, Section 4, we identified typical modeling challenges the first appointment’’ were both categorized as challenge ‘‘tem-
for healthcare processes with the help of a structured analysis poral constraints’’. After this grouping, multiple solutions existed
of a broad set of healthcare process modeling projects. Then, from different projects for each challenge. Thus, we grouped so-
we checked their relevance based on literature identified by a lutions addressing the same issue with a similar level of modeling
structured literature review. Based on the identified challenges, abstraction and transformed them into general applicable BPMN
we collected in the third step – for the treatment design [59] – a fragments. For example, if a clear decision logic for a decision-
set of solutions from the modeling projects and the literature, and making task in a healthcare process exists, we used a business
synthesized from them ready-to-use process fragments struc- rule task referencing a DMN model. If a medical expert mainly
tured into different types of process abstractions. The design and takes the decision, then we chose a user task. Additionally, we
development step resulted in a best practice framework for mod- re-checked that the developed fragments conformed to the BPMN
eling healthcare processes that is presented and demonstrated standard. In the end, we obtained a final list of 10 candidate
with examples in Section 5. Our resulting design artifact, the best challenges, each one having one or multiple solutions.
5
Table 1
Summary of healthcare process elicitation and modeling projects used as a source for the challenges along with information concerning the total number of process
models (column #PM) and activities (column #Act) designed during the project.
Main process Department Country #PM #Act Project goal
Admission of patients Center for DE 1 17 Process design and
for an elective musculoskeletal analysis
surgery surgery
Management of Intensive care ES 14 217 Process analysis for
catheter-related evaluating adherence
bloodstream to clinical pathways
infections and supporting
decision-making
[51,60]
Emergency care Trauma emergency DE 3 42 Process evaluation
Endoscopic procedure Gastroenterology DE 4 68 Process comparison
of different sites
Pre-surgical Trauma center DE 4 29 Process design and
consultation analysis
Pre-surgical Center for DE 4 41 Process design and
consultation musculoskeletal analysis
surgery
Management of Internal medicine, IT 3 104 Process design and
chronic obstructive Primary care, ER analysis for
pulmonary disease standardization and
(COPD) improvement [29,30]
Elective surgery Surgery NL 5 67 Process redesign
4.2. Checking the relevance of the candidate challenges in literature and searched for document titles referring to BPMN-based ap-
proaches. Our search was conducted in September 2019. Since the
As shown in Table 2, some candidate challenges were observed first official version of the BPMN standard was released in 2006,
more often in the projects (up to eight times), than others (only we considered results published starting from January 2006 for
two times). Our goal is to provide a set of challenges and related BPMN, and starting from January 2005 for UML. At the end of
solutions of general interest and relevant for different kinds of this search, we obtained 229 articles, which we narrowed down
internal healthcare modeling projects. For avoiding a potential to 1284 by manually removing duplicates.
bias due to our personal experience, we conducted a literature
Study selection. Then, we examined the title, abstract, and
search for studies addressing the application of standard graph-
scanned the content of each article looking for articles with a
ical process modeling languages in healthcare to find evidence
strong focus on the use of BPMN or UML for modeling health-
supporting the practical relevance of each candidate challenge.
care processes. For the selection, we considered the following
In the following, we first describe how we searched and selected
inclusion and exclusion criteria:
the studies for our relevance check (cf. Fig. 4). Then, we explain
how we established that a candidate challenge from the analyzed I1 The article describes the use of standard modeling lan-
real-world projects was also considered relevant in the literature. guages, i.e., BPMN or UML, for healthcare process model-
ing potentially discussing benefits, challenges and domain-
Literature search. Since we are interested in healthcare process
specific extensions;
modeling challenges related to the practical use of standard mod-
I2 The article describes a study about healthcare process mod-
eling languages, we focused on approaches supporting process
eling with a standard language, i.e., BPMN or UML;
modeling in healthcare and describing modeling initiatives using
BPMN and UML.1 In detail, we began with searching the DBLP2 E1 The article has a modeling focus but does not use BPMN or
and IEEE Xplore3 databases for conference and journal articles UML;
explicitly focusing on approaches based on BPMN or UML activ- E2 The article focuses only on the use of BPMN and UML to
ity or sequence diagrams [61] for healthcare process modeling. support the development of healthcare applications.
For retrieving work about BPMN in healthcare we constructed E3 The article focuses on process mining, analysis, and simu-
search queries combining the term BPMN with the terms clin- lation rather than process modeling.
ical, medical, and health for the article title and metadata. E4 The article focuses on the BPMN ontology instead of the
For retrieving work about healthcare process modeling in UML, BPMN notation.
constructed search queries combining the terms ‘UML’ and pro- E5 The article refers to the healthcare domain as a motivat-
cess with clinical, medical and health. We quoted ‘UML’ ing/application example but does not provide an approach
to avoid retrieving results about the Unified Medical Language tailored to healthcare-specific settings.
System, while the search term process was added to narrow E6 The article is not written in English or its full-text or rele-
down results to process modeling approaches. In addition, we vant parts related to process modeling are not available.
combined the keywords process modeling and healthcare E7 The article is an application in another scenario of concepts
presented in another publication by the same authors.
E8 The article is related to one of the projects used to identify
1 Although the notation of UML is different from BPMN, the basic modeling
the challenges (cf. Table 1).
concepts are similar and, thus, we assume that similar challenges can be
observed and only the modeling solutions differ in their notation details.
2 https://dblp.uni-trier.de. 4 The list of the identified papers as well as inclusion and exclusion criteria
3 https://ieeexplore.ieee.org/Xplore/home.jsp. can be found here: https://tinyurl.com/r47gl8r.
6
Fig. 4. Process followed for the literature search and study selection.
Table 2
List of candidate challenges identified in modeling projects and relevance in literature.
Candidate BPMN Found in Projects Evidence in literature Related work on challenges Process modeling case studies in healthcare
modeling challenge
Centralized 3 Yes [47,50,68]
scheduling of patients
Evidence-based 5 Yes [18,19,24] [63–65,67]
decision-making
Involvement of 5 Yes [48,65,68,69]
diagnostic or
treatment services
Involvement of the 8 Yes [25,67,70,71]
patient
Limited resources 2 Minor [47]
Monitoring of the 4 Yes [5,63,69]
patient
Patient-specific needs 2 Yes [18,24] [5,47,50,71]
Scheduled vs. 2 No
emergency patients
Shared working 7 Yes [16,18,24,62] [50]
behavior
Temporal constraints 4 Yes [18] [64,66]
As a result of our selection, we obtained 8 articles. Then, we in healthcare organizations. Whereas the first group of articles
conducted another search for modeling studies starting from the focuses on challenges in general, the second group does not
systematic literature review authored by Mincarone et al. [52] explicitly report on challenges and provides insights into a spe-
and the overview of process modeling initiatives presented in cific healthcare process. Therefore, we required the support of at
[24], considering the same inclusion and exclusion criteria. As least one article of the first category, because their results are
a result of this search, we found 12 additional articles, includ- generalizable, and of at least three studies of the second category,
ing [24], which also discusses challenges of using BPM in health- as their results cannot be universalized and are specific to one
care. process. By applying such relevance criteria, from the 10 initially
Then, we distinguished the selected articles into (1) works
identified challenges, we ended up with 8 main challenges.
explicitly summarizing and discussing the challenges with stan-
Table 2 also shows that existing related work on process
dard process modeling languages in the healthcare domain [16,
modeling challenges in healthcare has focused so far only on four
18,19,24,62] and (2) studies reporting on process modeling ini-
tiatives enacted in a specific healthcare organization [5,25,47, of our challenges: evidenced-based decision-making, patient-
48,50,63–71]. The first group consists of 5 studies focusing on specific needs, shared working behavior, and temporal constraints.
the same topic as this article, i.e., mentioning challenges and Nevertheless, through the analysis of the projects, we found that
possible solutions explicitly, sometimes proposing extensions. also other challenges are present in practice.
The other 14 articles report on case studies, often mention- In the selected studies, we found mention of a few other
ing challenges only implicitly but providing modeling solutions challenges, such as the collaboration with external care providers,
for them. The latter ones cover hospital processes (e.g., surgi- the media break (scanning and printing out documents at several
cal processes [47,50]), prevention and rehabilitation processes points of the healthcare process), the education of medical stu-
(e.g., falling prevention [70], stroke rehabilitation processes [63]), dents, and the visualization of critical clinical documents, which,
and diagnostic processes (e.g., anatomic pathology [47] and radi- as concluded by Braun et al. [62], could be easily realized with
ology processes [68]). One study did not fall in any of these two BPMN data objects. Whereas the latter three were reported only
categories because it proposes a domain-specific BPMN extension once, the first one is relevant to consider. Cross-organizational
for healthcare but does not discuss any challenge [4]. patterns need additional modeling concepts and will also en-
Relevance evaluation. The candidate challenges are provided in compass a number of patterns, such that we decided to leave
alphabetical order in Table 2, together with the evaluation of cross-organizational challenges out of scope and, thus, we did
their relevance. A candidate challenge is evaluated as relevant not include collaboration with external care providers among our
if it is mentioned by at least one peer-reviewed work focusing candidate challenges. Based on this analysis, we can conclude that
on challenges related to the use of standard process modeling we identified a set of relevant modeling challenges for internal
languages in healthcare, or by at least three modeling case studies healthcare processes.
7
Fig. 5. Overview of common challenges arising when modeling healthcare processes in BPMN.
5. BPMN best practices in healthcare 5.1. Patient-related challenges
This section presents the challenges and best practices for In this paragraph, we describe patient-related challenges and
modeling healthcare processes in BPMN. For giving an overview illustrate the proposed BPMN best practices. As already men-
of the selected challenges and related best practices, we start tioned, the key entity of a healthcare process is a human being,
from the general structure of a healthcare process [31] and con- who usually moves independently through the healthcare organi-
sider the six high-level activities shown in Fig. 5. In general, zation and interacts with different health workers. Representing
a healthcare process modeling initiative might focus only on the involvement of a patient in care activities (C1) is already one
selected aspects of a specific treatment or on the complete care of the most prominent challenges. Besides, healthcare processes
are flexible enough to react to different and changing patient-
process for a specific patient group. However, regardless of the
specific needs, among which we distinguish foreseen (C2) and
scope of the modeled processes, there exist some common as-
unforeseen (C3) needs.
pects of care practice that contribute to increasing the complexity
of modeling a healthcare process.
5.1.1. C1 – involvement of the patient
Fig. 5 shows the modeling challenges identified in Section 4
In contrast to traditional processes acting on a product or
categorized in (1) patient-related, (2) medical practice-specific,
documents, activities of a healthcare process are executed to-
and (3) medical resource-related challenges. For them, we aim gether with a patient – a human being with wishes, expectations,
to provide some BPMN best practices to support process mod- fears, etc. [25]. Patients bring documents, they may move on their
elers in capturing complex aspects of healthcare processes. As own through a healthcare organization, they are usually actively
previously mentioned, in this paper, we focus on processes that involved in therapeutic decisions, as exemplified in [71], and they
are enacted within the boundaries of a single medical orga- have to give consent to treatment or to the sharing of personal
nization. That is, we do not consider best practices for cross- data. In healthcare processes, patients may have an active role
organizational healthcare processes describing the interactions (e.g., if a patient has to agree to therapy) and a passive role
between different healthcare organizations. (e.g., when a patient undergoes blood testing).
In the remainder, we discuss the salient features of the chal- BPMN assumes that a business case is usually executed based
lenges introduced above; for each challenge, we discuss possible on its documents, which are read or written by activities [8].
BPMN best practices that address the respective challenge en- Certain interactions with a customer, e.g., the update of the
tirely or partially. Thereby, we structure the description of each customer’s address, are shown as message flow between the main
challenge as follows: process and the customer, who is represented in a separate pool.
Although some healthcare activities can be executed based on
• An overview of a representative clinical situation; patient documents or patient samples (e.g., blood analyses make
• The challenging aspects of modeling it in BPMN; use of blood samples), the majority of the activities are actually
• Best practice modeling solution as ready-to-use BPMN frag- happening together with the patient, be it actively or passively.
ments. When possible, we discuss different solutions (iden- Thus, the challenge is to understand how and at which abstrac-
tified by a numerical enumeration) for capturing the same tion level the patient and his/her involvement in the healthcare
aspect. Further, we propose, where possible, best practices process should be represented.
at different levels of process abstraction (c.f. Section 2): Table 3 shows the different abstraction levels that we identi-
descriptive process models (D) – i.e., suitable for high- fied to represent the patient’s involvement in BPMN. We suggest
level documentation purposes, analytic process models (A) using option (D) for most of the activities in a healthcare process
– i.e., suitable for analysis purposes, and implementable to have a more readable model. However, if the patient plays an
process models (I) – i.e., suitable to prepare an implemen- active role in the process that cannot be disregarded, e.g., he/she
tation. These abstraction levels should be understood as has to deliver certain medical documents, needs to give consent
proposals. to treatment, or has to make a decision, then we suggest to
8
Table 3
C1 involvement of the patient.
represent this explicitly through option (A). Option (I) is a rele- patient shows adverse reactions to a certain drug. Such changes
vant alternative if the process model is used for implementation in the way certain activities are enacted lead to the definition of
purposes. Indeed, if the patient should have a distinct role in the different process variants. The fact that patients require slightly
healthcare system, then it is essential to represent the patient as different executions of a healthcare process based on their char-
a lane to consider him/her when implementing user interfaces. acteristics and needs, is known as process variability [73] in the
BPM domain and different approaches exist to support the mod-
5.1.2. C2, C3 patient-specific needs eling of process variants [74]. However, capturing different pro-
Being driven by the individual needs of a patient, healthcare cess variants leads to more complex process diagrams, especially
processes tend to be highly dynamic and subject to changes. when using imperative process modeling languages. To abstract
Clinical practice guidelines [37] and clinical pathways [72] are from such complexity, foreseen changes may be represented at
powerful methods for the standardization and streamlining of different levels of abstraction and with different solutions, as
care based on medical knowledge, typically aiming to increase shown in Table 4.
quality of care while decreasing costs [24]. However, the health- Option (D) should be used mainly for documentation purposes
care processes implementing such guidance need to be flexible because of its compactness. However, the annotations have no
to meet individual patient needs. Indeed, although the main execution semantics, such that option (AI1 ), which explicitly out-
therapeutic activities may be predefined, their course of action lines specific process variants, is more suitable for analytical and
may change based on clinical evidence and organizational factors, implementation purposes. It considers different alternatives for
such as resource availability. Thereby, we distinguish between the same activity or process path. Option (AI2 ) should be used
foreseen and unforeseen patient needs and situations. to react with alternative/additional activities to the occurrence of
relevant events. Finally, option (AI3 ) allows even more flexibility
C2 - patient-specific needs (foreseen). Core diagnostic and treat-
by having a set of individual, executable activities.
ment activities are usually described based on clinical guidelines,
which serve as a standard reference for the care process. How- C3 – patient-specific needs (unforeseen). In this paragraph, we
ever, specific categories of patients may require the enactment focus on unforeseen patient needs, i.e., exceptional situations
of additional or slightly different diagnostic and treatment ac- that may cause changes or interruptions in the healthcare pro-
tivities [18,24], e.g., if diagnostic results are inconclusive or the cess [18,24]. In the literature, this is discussed at a generic level
9
Table 4
C2 patient-specific needs (foreseen).
in the sense that all kinds of exceptions can occur [18]. In this also inspire solutions for other similar exceptional situations that
work, we focus on two common unforeseen needs: (1) patient require changes to the planned healthcare flow.
conditions change suddenly and the planned treatment cannot be
continued [21], and (2) the initial diagnosis is disproved and the 5.2. Medical practice-specific challenges
patient needs to enter a different healthcare process.
Table 5 shows the best practices for (1), which rely on BPMN
boundary events, and for (2), which is based on the concept In this section, we describe the medical-related challenges and
of event-sub-process [8]. For suddenly changing conditions, we illustrate the related BPMN best practices.
propose two options based on different levels of process abstrac- Medical activities have some specifics which are less observed
tion. Option (D) is useful for documentation purposes and shows in other more structured process domains — we refer to them
that the process is interrupted because of such changes. Option as medical practice-specific challenges. During the patient as-
(AI) is suitable for analytical and implementation purposes, and sessment, treatment plan definition, or the review of treatment,
captures in detail that the changing conditions can be managed medical experts have to make essential decisions for patient
concurrently to the main healthcare process. These best practices care. These are usually made by relying on medical evidence
10
Table 5
C3 patient-specific needs (unforeseen).
systematically collected in CPGs, the state of the art in medical to find one or several decision-making activities that trigger
knowledge, and by considering the information about the patient. (patient-specific) diagnostic or treatment activities.
As a result, healthcare processes are characterized by frequent Evidence-based decision-making is not explicitly supported
evidence-based decision-making (C4) tasks. Furthermore, diag- by BPMN [18]. Thus, Braun et al. [18] extended BPMN in or-
nostic and treatment activities are time-sensitive and therefore der to properly manage evidence-based decision activities, the
require advanced temporal constraints (C5) to be observed during information referenced by CPGs and the relevant decision logic
care delivery. Last but not least, monitoring activities (C6) play a in a CPG to diagnose or treat a certain patient group. How-
crucial role in the observation of the patient’s health conditions. ever, this is not supported by any existing BPMN modeling tools.
Combi et al. [29] employ user tasks to model evidence-based
5.2.1. C4 – evidence-based decision-making
decision-making activities. Additionally, BPMN provides business
Medical activities, such as diagnosis and treatment, are driven
rule activities that can be used to represent operational decisions
by observations, experiential evidence, and the decision-making
that are made according to particular decision logic, which can be
of a medical expert. Lenz and Reichert [22] describe them as a
diagnostic–therapeutic cycle in which patient-related informa- directly linked to a DMN model [11]. This combination of BPMN
tion and medical knowledge are used as input for the decision- and DMN is applied also in [30] to design decision-intensive care
making by a medical expert. Evidence-based decision-making pathways. In clinical practice, it is challenging to select the best
requires evidence for a specific decision based on existing medical fitting CPG to the clinical environment as usually a broad range
knowledge, usually captured in CPGs, which provide recommen- of CPGs are published for a specific disease. Another challenge
dations for decision-making based on literature reviews and the is how to proceed if multi-morbid patients have to be handled.
best available evidence [31]. The results of the decision-making We assume that the selection of the CPGs and the possible ways
trigger the execution of certain diagnostic or treatment activi- to take care of a multi-morbid patient are determined by the
ties for a patient. Thus, in healthcare processes, it is common clinical experts before the healthcare process modeling. Thus, the
11
Table 6
C4 evidenced-based decision-making.
modeling challenge is how to represent evidence-based decision- made available to the medical expert to conduct decision-making
making activities in a BPMN process diagram, and further, if activity.
decision models are used, how to apply them to semi-formalized
medical guidelines. Different abstraction levels can be employed 5.2.2. C5 – temporal constraints
to represent a decision-making activity, as shown in Table 6. Temporal phenomena, and especially temporal constraints,
Option (D) should be used for cases where the decision- play a significant role in healthcare domains [18,27], encom-
making is mainly driven by the medical expert and not supported passing both administrative and medical activities. Proper pro-
by any guideline/standard. Option (DA) is suitable if a decision- cess design entails the need for managing temporal constraints
making activity is supported by a CPG or internal standard. Option and coordinating the constraint-based interactions among the
(AI) can be used if the decision rules are given in detail (e.g., in processes contributing to the overall patient treatment [23].
a CPG or in an internal document) so that the decision logic However, BPMN offers limited (direct) support to the mod-
can be specified in a decision table to assist the medical experts eling of the temporal perspective [27] as witnessed by recent
in applying it or to use it as implementation blueprint. The works aimed to improve such support [23,26,75]. Indeed, al-
advantage of explicitly using these different abstraction levels is though BPMN includes timer events, these cannot be used di-
that the existing support of decision-making is explicitly visible. rectly to capture all the above-introduced time constraints. In
Process analysis and redesign initiatives can analyze whether particular, BPMN supports three kinds of temporal constraints,
(D) decision user activities require increased support in order namely (1) maximum activity duration, (2) waiting time between
to streamline patient care. DMN models, as proposed in (DA) subsequent activities, and (3) iteratively performed activities. All
and (AI), can also support digital transformation initiatives as these constraints can be represented with timer events as shown
they show which data, guidelines, and decision logic has to be in Table 7.
12
Table 7
C5 temporal constraints.
In our projects, we often observed more complex time con-

straints, e.g., a minimum activity duration (e.g., ‘‘aftercare needs
to last at least 30 min’’). To capture more complex temporal con-
straints, such as time-lags, minimum activity duration, schedule-
restricted elements, and maximum patient waiting time, we sug-
gest relying on BPMN extensions proposed in the literature. In
particular, the proposal presented in [75] allows representing
temporal constraints in a user-friendly manner, by defining con- Fig. 6. Example of process with temporal constraints represented through the
straints as activity attributes and by representing them on the extension proposed in [75].
process model through graphical icons, text annotations, and
different kinds of edges, as shown in Fig. 6. Duration constraints
and time-lags can be easily specified, whereas constraints such his/her physiological function, and the assessment of life support
as ‘‘Start As Soon As Possible’’, ‘‘Start No Later Than’’, can be equipment. Usually, monitoring runs in parallel with other diag-
used to capture the maximum patient waiting time and to define nostic or therapeutic activities to check their efficacy and possibly
schedule-restricted elements. Fig. 6 shows the graphical notation adjusting them [63].
for a time lag of 12 to 24 h between the start of Activity 1 and Continuous monitoring is common practice in intensive care
the end of Activity 3, the minimum duration of 1 h and maximum settings and is realized with the help of (electronic) sensors
duration of 3 h on Activity 1 and a constraint ‘‘Start No Later Than and data processing devices (e.g., continuous EEG, transducers,
10 am’’. on Activity 2. glucose monitoring systems) that monitor the vital signs of the
An approach fully based on standard BPMN elements for cap- patient in real-time. These systems are able to detect significant
turing duration constraints is presented in [26]. Being completely changes in the patient’s conditions and to accordingly prompt
based on BPMN semantics, this approach may be useful for im- alarms to alert care professionals. Instead, repeated monitoring
plementation purposes, but can be complex for documentation activities consist of clinical observations carried out regularly
purposes. by care professionals, typically several times per day. Although
clinicians may rely on devices, the human role in this kind of
5.2.3. C6 – monitoring activity monitoring activity is predominant. Examples of repeated mon-
Monitoring activities refer to continuous or repeated activities itoring are drug therapy assessment, blood pressure measure-
involving the observation of the patient, the measurement of ment, and catheter surveillance. Changes of patient conditions
13
Table 8
C6 monitoring activity.
detected during monitoring can be seen as events that need to be focus is on representing continuous execution and data elements,
interpreted and managed in parallel to the diagnostic/treatment but the details on how the monitoring takes place are hidden
activities — these events may also prompt process adaptation in the sub-process. Instead, options (I1 ) and (I2 ) are useful for
(as also described in the challenges C2 and C3). As an example, implementation, as they show the details of the monitoring sub-
consider the self-monitoring of glucose blood levels in patients process. We can distinguish between (I1 ) manual monitoring
affected by diabetes. Glucose monitoring is crucial to prevent (manual task), which is realized periodically by the medical staff
hypo- and hyperglycemic excursions and can be realized in real- who takes care of the monitoring, and (I2 ) automatic or semi-
time, e.g., with the help of a wearable sensor device, or manually, automatic monitoring (service task), which is conducted with the
with the help of a glycemic reader [76]. Patients can adjust their help of electronic medical devices and requires the clinician to
diet or insulin intake based on the measured glycemic level. interpret data.
The challenge of representing monitoring activities in BPMN
is the complexity, consisting of three aspects which need to be
considered: capture the continuous or repeated character of mon- 5.3. Medical resource-related challenges
itoring activities, the generation and processing of monitoring
data, and the detection of significant changes in such data and Medical staff and machines are often highly specialized and,
the possible adaptation of parallel running diagnostic or treat- thus, scarce and cost-intensive, thus giving rise to some medical
ment activities [69]. Monitoring activities can be represented resource-related challenges. Here, we focus on the involvement
at different abstraction levels in a BPMN process diagram, as of multidisciplinary teams in clinical activities (C7), which is tra-
shown in Table 8. In detail, option (A) should be used when the ditionally not considered in BPMN, on the centralized scheduling
14
Table 9
C7 shared working behavior.
of patients (C8) motivated by the scarce and cost-intensive re- Rogge-Solti in [16], who identified the following scenarios for
sources, and on the need of centralized treatment and diagnostic shared working behavior in healthcare processes: (1) several
services at the service of different healthcare processes (C9). specialists work together on a shared task, (2) different roles
can alternatively perform an activity, and (3) an activity can
5.3.1. C7 – shared working behavior optionally involve additional roles. In BPMN process diagrams,
Healthcare processes are often executed by multidisciplinary the concept of lanes, which subdivide a pool, can be used to assign
teams in order to provide the best healthcare service to a patient. persons, roles, or units to activities [8]. All activities added to a
This phenomenon has been extensively studied by Müller and specific lane are in the responsibility of the person, role, or unit
15
Table 10
C8 centralized scheduling of patients.
assigned to this lane. Lanes make the representation of shared instances). Thereby, it is assumed that these process cases have
work and flexibility in healthcare processes more challenging and an independent existence and are executed without reference
the flexibility in healthcare processes. to each other [78]. When patients need to be ordered based
Müller and Rogge-Solti in [16] discuss several solutions on on their medical priority or scheduled for appointments, several
how shared tasks can be presented in BPMN process diagrams. process cases running in parallel need to be enacted. However,
Based on our experience gained through the conducted health- running cases cannot be represented in a single BPMN process
care modeling initiatives, we suggest different best practices to diagram.
capture the different kinds of shared working behavior intro- We propose to represent the scheduling in an additional pool,
duced above. The best practices are shown in Table 9. as shown in Table 10.
The proposed best practices use existing elements of BPMN
process diagrams. Other proposals based on newly introduced,
non-standard elements can be found in [16]. Our suggestions 5.3.3. C9 – involvement of diagnostic and treatment services
focus on the usage of the swimlane concept delineated in BPMN, In hospitals, certain diagnostic or treatment activities are or-
which may help to understand the role of different participants ganized centrally, such as radiologic tests, laboratory analysis,
and the activities they perform for documentation or analysis etc. [79]. They usually require certain specialists and equipment
purposes. If the process model is designed for implementation and are used within different healthcare processes running in
purposes, swimlanes are often not used, as the information about a hospital. For example, blood analyses or X-ray scans can be
resources is hidden in the properties of the activities. This solu- requested during emergency care as well as during the pre-
tion can be considered if the involvement of the medical staff is surgical consultation. Often, the service is requested via phone
not relevant for the process. or over an IT system with all patient-related information [68,69].
Subsequently, if the patient is needed for the diagnostic or treat-
5.3.2. C8 – centralized scheduling of patients ment service, the patient is sent to the responsible department,
Healthcare processes are from time to time confronted with e.g. radiology. Instead, in the case of blood or laboratory analyses,
the situation that the number of patients requesting a medical only the sample has to be sent to the laboratory. The healthcare
service is higher than the number of resources available at a process can continue as soon as the results are received and, if
specific moment. Each patient has a different level of clinical the patient is needed for the diagnostic service, the patient is
priority based on which the care plans are scheduled. For ex-
back. The single healthcare process usually does not influence the
ample, for emergency care, patients’ priority and their maximum
design and execution of the service process. In most cases, diag-
waiting times are often defined based on the Manchester Triage
nostic and treatment services are individually running processes
system [77]. Based on this information, patients are dynamically
that collaborate with the to-be modeled healthcare processes [48,
scheduled, and treatment priorities are defined.
65,68]. Thus, in a BPMN diagram, the service process needs to be
On the other hand, resources such as operating rooms need
presented as an own pool which communicates via message flow.
to be used efficiently. Similarly to the manufacturing area, the
right order of surgeries can play a role in avoiding changeover Here, the challenge is to represent this interaction efficiently, also
times [50]. Thus, the scheduling of patients is also needed in involving, in many cases, the patient.
healthcare processes with scarce and cost-intensive resources We propose to use a collapsed pool to represent the diagnostic
(e.g., operating rooms, medical imaging systems [68]). Although, or treatment service to abstract from the details of the service
this is especially a challenge of the process implementation, execution, as shown in Table 11. In case that the details of the
where specific resource planning or allocation mechanisms need diagnostic/treatment service are relevant, an expanded pool can
to be used, we want to discuss in this part the related modeling be used to capture the service process. We distinguish between
challenge. two versions, i.e., (1) where only a sample from the patient is
A BPMN process diagram traditionally represents the possi- needed (e.g., a blood sample) and (2) where also the patient is
ble execution sequences of a specific set of process cases (or needed (e.g., a radiologic test).
16
Table 11
C9 involvement of diagnostic and treatment services.
6. A first evaluation of the best practices Section 5. In this section, we describe the planning and execution
of our study and report the findings.
To evaluate the usefulness of the proposed best practices, 6.1. Exploratory study design and execution
we followed an exploratory approach to understand how the
proposed best practices support process designers in modeling In this study, we investigate how process modelers use and
the challenging aspects of healthcare processes introduced in perceive the best practices in the context of a process modeling
17
Fig. 7. Main steps of the conducted exploratory study, represented as a BPMN process. The colored bands and lines are used only for presentation purposes.
task. Specifically, we are interested in exploring how the use of Design. Fig. 7 shows the design of our study, which we organized
the best practices is reflected in the obtained process models into two sessions, each one centered around a modeling task, and
and how modelers perceive the usefulness of the best practices. a retrospective interview. The first session aims to investigate
Accordingly, we define the following two research questions. how participants deal with modeling challenges in absence of
the best practices. The session begins with an introduction to
(RQ1) How does the use of the best practices support the task the modeling tasks and a review of the main BPMN concepts.
of healthcare process modeling? With RQ1, we investigate Then, we administer the initial questionnaire to gather informa-
how the best practices are used in the modeled processes tion about the participants’ background in process modeling and
and whether they have an impact on the syntactic and their experience as healthcare workers. For the modeling task,
semantic quality of the obtained models. we follow a within-subjects design [81] and randomly divide
(RQ2) How do process modelers perceive the usefulness of the best participants into two groups, G1 and G2. For the first session,
practices? With RQ2, we aim to gain insights into the we ask participants in G1 to model P1 starting from its textual
perceived usefulness of the proposed best practices. description and participants in G2 to do the same with P2 (cf.
Fig. 5). Then, we administer the first post-task questionnaire to
Participants. The proposed best practices are meant to support
both groups. The second session focuses on investigating how
process modeling in healthcare organizations. Specifically, they participants deal with the modeling challenges with the best
are addressed to healthcare stakeholders experienced in process practices at hand. Initially, we provide a 30-minute tutorial to
modeling with BPMN and business process modelers and analysts introduce the challenges and best practices. Then, we switch
with experience in healthcare. Thus, for our evaluation, we target the process descriptions between groups and ask participants to
both healthcare and IT experts who are to some degree familiar model the processes with the help of the best practices, which
with process modeling in BPMN. We want to avoid that partici- remain at their disposal for the whole duration of the task.
pants encounter problems due to their lack of BPMN knowledge Then, we administer the two post-task questionnaires. Finally, we
but also want to observe if our best practices can be used by conduct semi-structured interviews with selected participants to
participants that are moderately familiar with BPMN. learn how they perceive the process modeling task without and
with the best practices and why they made specific modeling
Materials. For our evaluation, we developed different materi-
choices.
als. First, we prepared textual descriptions of two healthcare
processes, P1 and P2. Each process included six of the model- Execution. The study was conducted between November and De-
ing challenges introduced in Section 5, such that all challenges cember 2019 at the University of Potsdam, Germany. Participants
were covered. The process descriptions were derived from clin- were 28 graduate students attending the course ‘‘business process
ical guidelines and pathways related to real-world healthcare analysis in healthcare’’ which is part of the master’s program in
processes: P1 was taken from the clinical pathway for COPD Digital Health. Our participants have varied backgrounds: 13 of
presented in [29,30], while P2 was designed to include main di- them hold either a degree in medicine or medical and pharma-
agnostic and treatment steps of acute appendicitis in emergency ceutical sciences, 8 in IT-related subjects, three in biomedical or
care [80]. We prepared the descriptions selecting relevant parts biotechnology engineering, two in healthcare management, and
of the guidelines and balancing them in terms of the number two in neuroscience and psychology. Regarding the requirements
and kind of BPMN elements needed in the expected solutions. for participation, we would like to note that our participants are
representatives of junior experts in IT and healthcare. Indeed,
We also prepared the best practice solutions for P1 and P2 to be
most of them have enrolled in the study program to obtain
able to compare the process models designed by the participants.
an additional qualification and have already gained significant
Besides, we prepared a handout summarizing all the challenges
experience in their field of study, as reported in the initial ques-
and best practices, formatted as in Tables 3–11. However, options
tionnaire. In detail, 18 participants have advanced or professional
(AI3 ) for C2 and (1) for C9 were not included in the handout
healthcare expertise, while 8 of them have advanced or profes-
since they were added after our study, as discussed in Sections 7
sional IT expertise. Moreover, 13 participants had already gained
and 6.2. Moreover, we designed an initial questionnaire to collect working experience in a hospital at the time of the study, either as
demographics and learn about the participants’ process model- medical personnel or as quality consultants and hospital adminis-
ing and healthcare expertise and two post-task questionnaires. trators. Speaking of process modeling experience, our participants
The first post-task questionnaire included multiple choice and indicated having designed a varying number of process models,
open questions about the modeling task; the second included with 6 participants reporting zero process models, 18 of them
multiple-choice questions about the perceived usefulness of the reporting 1–5 process models and four of them indicating 10–20
best practices. Finally, we prepared a semi-structured protocol for process models. However, since all participants had followed the
the retrospective interviews, including general questions about introductory lectures about process modeling in BPMN (approx-
the process modeling task, modeling choices, difficulties faced imately 13 h covering teaching and guided studying), we could
during the modeling task, and the perceived usefulness of the best ensure that they have the BPMN modeling skills necessary for the
practices. study. Each session (cf. Fig. 7) lasted about 1.5 h, with 45 min
18
dedicated to the modeling task. After both modeling tasks were process models shows that the best practices were not used
completed, we reviewed the process models designed by the as much as the best practices addressing the other challenges.
participants in both sessions, focusing on the modeling challenges Without best practices, the alternative therapy options (C2) were
and evaluating their syntactic and semantic quality. To follow up shown successfully in some process diagrams with an exclusive
on our evaluation, we interviewed nine participants, whom we gateway. In the remaining diagrams, this was either captured
selected considering both their different backgrounds and the low erroneously, as a simple activity hiding the relevant details, or
or high quality of the designed process models. not modeled. With the best practices, the challenging aspects of
foreseen needs were almost always represented. However, many
Data analysis. For each participant, we collected two process
models, one designed without and one with the help of best diagrams still included syntactic issues (e.g., missing conditions
practices. We also collected data in the post-task questionnaires after the exclusive gateway) or used a simple activity. Thus, we
and complementing interviews. For addressing RQ1, we focused assumed that the best practices were not used carefully. One
on the obtained process models. In detail, for the models designed student with a background in medicine used an ad-hoc sub-
during the first session (i.e., without the best practices), we re- process, which we considered as a valid and relevant alternative,
viewed how participants tried to capture the challenging aspects and thus we added it into the final version of best practices for C2.
and whether their modeling solution was similar to any of the The challenging aspect of C3 was either not modeled when best
best practices. For the second session, we investigated whether practices were not available or included as an alternative path
best practices were used and, specifically, for each challenge, at the end of the model, such that it was not allowed to react
we checked whether the described part was missing, incorrect to changing patient conditions during the whole process. With
(i.e., not compliant with the BPMN syntax or semantics), cor- the best practices, we could observe in four process diagrams the
responding to a best practice solution, or corresponding to an correct application of the best practices, whereas the rest still not
alternative solution. To ensure consistency, two authors checked (correctly) captured the reaction to changing patient conditions.
the process models designed with and without the best practices A possible explanation for these findings could be that the
independently and, then, discussed differences in the results to best practices for C2 and C3 include advanced BPMN constructs
reach consensus. For addressing RQ2, we relied on the answers (e.g., ad-hoc and event sub-processes), which require good BPMN
to the second questionnaire and the interviews with selected knowledge and abstraction capabilities to be used properly. Still,
participants. In the questionnaire, for each challenge, participants some of the interviewed participants reported that the best prac-
reported if they had used the corresponding best practice it in tices for C2 and C3 were useful to show how advanced BPMN
their models and, in general, if they find it helpful, easy to un- constructs could be used ‘‘I like that the best practices show how
derstand, and if they would use it in the future. In the interviews, comments [text annotations] can be used on activities, I used them
we collected qualitative feedback about the perceived usefulness in several parts of my process model’’.. Another one reported ‘‘I liked
of the best practices. to see how events and event-sub-processes can be used to interrupt
an activity based on the patient’s wishes or sudden changes in the
6.2. Findings health conditions’’..
The collected data provided insights into how best practices Medical practice-related challenges. The observations related to
are used, especially when considering the different backgrounds the medical-related challenges are summarized in Fig. 9. Without
of the participants and the influence of their prior experience best practices, most of the participants were not able to capture
on the modeling task. For each challenge, in Figs. 8–10, we decision-making activities (C4), temporal constraints (C5), and
summarize how the corresponding best practices were used in monitoring activities (C6), and those who tried using timer events
the process models (cf. RQ1). Besides, we report the results of the were often not able to use them properly. As a result, these pro-
final questionnaire and the interviews related to the perceived cess diagrams included syntactical errors and did not represent
usefulness of the best practices (cf. RQ2). the aspects described in the textual process description correctly.
The best practices were improving the way medical practice-
Patient-related challenges. Overall, the best practices for patient-
specific challenges are modeled. Mainly, the best practices for
related challenges were deemed relevant by most of the partic-
the time constraints (C5) were widely applied and contributed
ipants. However, when observing the process models designed
to increasing the syntactic and semantic quality of the obtained
during the study (cf. Fig. 8), it is clear that patient-specific needs
process models. The representation of C5 was most appreciated
are hard to capture in BPMN, particularly without the help of best
for its compactness ‘‘I was very helpful to see how timer events
practices.
can be used for iteration and to replace loops in the process flow’’.
The involvement of the patient (C1) was not modeled explic-
Decision-making activities were mostly modeled as simple tasks
itly by most of the participants without best practices. In the
and only four subjects used the DMN DRG.5 A couple of people
resulting process models, relevant interactions with the patients,
such as the patient has to give consent, were abstracted away and reported that ‘‘I didn’t find C4 useful, as it is not clear to me how
not shown, having a negative influence on the semantic model DMN can be used’’ and ‘‘It is not clear to me which abstraction level
quality. is the best from a process perspective’’.
Four people, three of which with a background in IT, modeled Without the best practices, the monitoring activity (C6) was
the patient as a lane or an independent pool. not modeled, or the parallel and repetitive character of the mon-
With the help of the best practices, more participants were itoring activity was not correctly depicted. However, one stu-
able to model the patient correctly. Most subjects chose option dent with a medical background modeled it as proposed in the
(I) because it allows to highlight the interaction (‘‘C1 was clear, best practice. With the best practice, the parallel and repetitive
the interaction was obvious, that is why I applied option (I)’’) and character of the monitoring activity was better captured by the
is compact (‘‘I choose option (I) as I find it stylistically better than participants. C6 was often compared to the ones for patient-
having the patient mentioned in labels [option (D)], and takes less specific needs (C2 and C3), as it was perceived as a useful example
space than using a lane [option (A)]’’).
Patient-specific needs, foreseen (C2) and unforeseen (C3), were 5 Apart from the details needed for C4, DMN was not explained in the
indicated as the most difficult ones to understand in the post- BPMN course. Therefore, most people lacked the knowledge to use its advanced
task questionnaires. Accordingly, the analysis of the resulting constructs properly.
19
Fig. 8. Summary of the observations for the patient-related challenges. The filling of the boxes shows the proportion of how often the best practice was missing,
incorrect, etc..
Fig. 9. Summary of the observations for the medical practice-related challenges. The filling of the boxes shows the proportion of how often the best practice was
missing, incorrect, etc.
of connecting different BPMN constructs and to capture activity in this group also applied the best practice correctly when mod-
interruption, e.g., ‘‘I perceived the best practices very useful to model eling P1. Instead, the group starting with P1, had more challenges
and understand how different BPMN construct can be connected, with representing an interaction: the diagnostic service was often
such as in the monitoring’’ or ‘‘I took inspiration from the monitoring represented as a simple lane, or the interaction with the main
activity and used boundary events for capturing interruptions in process was modeled incorrectly. When modeling P2, participants
other parts of the process, e.g. C3’’. seemed not to have used this best practice and their process
models contained several issues. From the initial questionnaire,
Medical resource-related challenges. The observations related to we learned that this group included more participants with less
C7–C9 are reported in Fig. 10. As regards shared working behav- BPMN experience, which can explain the presence of modeling
ior (C7), we observed a difference between the models requiring issues. Another reason could lie in the differences between the
shared activities (option 1) and alternatively performed ones processes. Indeed, in P1 the patient was needed explicitly for the
(option 2). Shared activities were modeled by half of the people diagnostic service whereas in P2 only a sample was needed. In
already without the best practices, and almost everyone used the version of the challenges used for the evaluation, we provided
the best practices correctly in the second session, such that the only the best practice for the case in which the patient was
responsibilities are correctly shown in these process models. In- needed explicitly (c.f., C9 option (2)) and only discussed the need
stead, the best practice for alternatively performed activities was for samples in the text. However, based on the results of the
less applied. For both kinds of shared working behavior, partici- evaluation, we decided to extend C9 by including the current
pants also provided correct solutions already without the help of option (1) for the situation where only a sample is necessary.
the best practices.
The centralized scheduling of patients (C8) was mostly cap- 7. Discussion
tured as a simple activity without best practices and only one
participant was applying the best practices, with the effect that In this work, we have developed a set of best practices to sup-
the interrelation to other patients is missing in these process port the modeling of healthcare processes. Modeling challenges
models. According to one of the interviewed participants, the and solutions were derived from real-world healthcare projects,
centralized scheduling of patients ‘‘makes a lot of sense from an and a comparison with literature helped us to generate a relevant
organizational perspective, but I focused more on the patient-flow’’.. set. The use and impact of the best practices on healthcare process
The modeling of C9, the involvement of the diagnostic service, modeling and their perceived usefulness were evaluated through
requires capturing the interaction with the help of another pool an exploratory study with graduate students having working
and message flow. The group that started with P2 captured the in- experience in healthcare and IT.
teraction with the diagnostic service successfully in many process In this section, we discuss the main results and limitations of
diagrams without the best practices. The majority of participants our work.
20
Fig. 10. Summary of the observations for the resource-related challenges. The filling of the boxes shows the proportion of how often the best practice was missing,
incorrect, etc.
7.1. Results of the exploratory study Thus, we learned that it is essential to have enough time available
to use the best practices effectively.
Our findings revealed that the process models obtained with Involving graduate students with different backgrounds in the
the best practices contained fewer errors and captured healthcare study allowed us to gain insights into how domain knowledge
aspects in greater detail. Results also showed that complex con- is used during process modeling, bringing both advantages and
cepts, such as shared working behavior or patient needs, were not disadvantages. Participants familiar with the healthcare domain
abstracted away so often. Moreover, the post-task questionnaires could relate more to certain clinical aspects described in the
and interviews revealed that the best practices were perceived as text and considered their representation in the process model
‘‘useful’’ by all the participants. Six best practices out of nine were essential: For example, a participant with a medicine background
also considered as ‘‘easy to understand’’, whereas foreseen (C2) modeled the monitoring activity without the best practices in a
and unforeseen (C3) patient-specific needs, and the monitoring way similar to our solution. However, domain knowledge also
activity (C6) were pointed out as more challenging to understand led to models with unnecessary details (e.g., differentiating all
by some participants because they included advanced BPMN con- kinds of samples collected for the analysis), which often made the
structs. This suggests that more BPMN training is needed to be process complicated even with the help of the best practices. This
able to use such best practices properly. happened more for participants with a healthcare background but
Having ‘‘ready-to-use’’ process fragments was perceived as a less experience with BPMN.
way to speed up the modeling and improve the appearance of From our first evaluation, we can evince that the process mod-
the graphical model, as reported by some of the interviewed elers with relatively good experience of the used BPMN modeling
participants, e.g., ‘‘My goal was to make the process as concise and constructs and domain-knowledge can benefit most from the
easy to understand as possible and the small examples helped me presented best practices.
very much to achieve this’’ or ‘‘I find the best practices very helpful
as they show how to apply BPMN for representing healthcare aspects 7.2. Limitations and threads to validity
and how to connect the different constructs correctly, for example,
to represent the monitoring activity’’. The results of the exploratory The main limitation of this work is that the identified list
study also helped us to improve and enhance the best practices of challenges and best practices might not be comprehensive.
by adding two new options for challenges C2 and C9. Nevertheless, we identified them from a diverse set of health-
In general, we observed that participants with prior BPMN care modeling initiatives from different countries and compared
or software diagram modeling experience benefited more from them to existing related work to confirm our findings. No addi-
the domain-specific best practices, which were used to find in- tional relevant challenges were identified besides communication
spiration for modeling ideas, e.g., ‘‘I used the best practices as an challenges arising in cross-organizational healthcare processes.
inspiration to solve general issues, such as the shared working behav- However, since in this paper we focused on healthcare processes
ior’’. Also, the best practices for evidence-based decision-making within one organization, we did not discuss challenges related
(C4), which included DMN concepts that were not introduced, to cross-organizational healthcare processes. Besides, we built
made evident that the modeling concepts have to be clear to a upon our experience and the reviewed literature to derive the
process designer to apply the proposed best practices success- proposed best practices, but we do not exclude that there may
fully. Still, several participants with basic BPMN understanding be other BPMN solutions equivalent to those we have presented.
used the best practices as a way to improve their understanding Last but not least, the focus of this study was on supporting pro-
of BPMN, both in general (e.g., for understanding the semantics cess modeling challenges by providing a set of BPMN fragments,
of boundary events) and in the context of healthcare. Some of the industry-standard. The available concepts of BPMN and the
the interviewees claimed that the best practices helped them notation’s limitation might have led to a bias in the creation of
to become more acquainted with some BPMN constructs and to the fragments. We tried to minimize the impact of such bias by
learn how to use them to capture similar concepts (e.g., ‘‘I focused taking solutions from literature and from the modeling projects,
on time constraints and I found it helpful to see that timer events and by discussing them again in the group of co-authors. The frag-
can be used to avoid loops’’ or ‘‘the modeling was easier with the ments could be additionally complemented with other types of
best practices, as I knew which constructs were needed for capturing models for capturing additional process-related aspects, e.g., UML
which aspects’’.). A few participants complained that they found it models [61] for representing the interaction with healthcare IT
hard to go through all the best practices during the modeling task systems or Asbru [38] for specifying and handling advanced tem-
and preferred not to use them as they felt under time pressure. poral aspects. However, since the presented best practices are
21
grounded on practical evidence and include several concrete ex- be complete but provides a set of best practices grounded in
amples taken from clinical practice, they have the potential of empirical evidence that can encourage the uptake of BPMN for
reaching out to a broad audience, including academics but also healthcare process modeling.
healthcare practitioners and digital health students. A first evaluation of the best practices through a process
The results of our exploratory study should also be seen in modeling study with graduate students in Digital Health showed
light of some limitations. First, the final results could be influ- that these domain-specific best practices could help to streamline
enced by the different expertise of the participants in BPMN and the task of healthcare process modeling and improve the general
their domain knowledge. Indeed, although all the participants understanding of BPMN. Besides, the participants in our study
were attending the same BPMN course and were randomly di- perceived them as useful to speed up the modeling and improve
vided into two groups, we could not control the differences in the visual appearance of the BPMN diagram. We observed that
healthcare and process modeling expertise acquired during their participants with a basic knowledge of BPMN and an understand-
undergraduate studies. Indeed, prior BPMN and DMN knowledge ing of healthcare domains benefited more from the best practices
positively impacted the quality of the obtained process mod- than those with limited process modeling expertise and with
els, including those modeled without the help of best practices. no expertise in healthcare. Thus, we assume that the proposed
Speaking of domain knowledge, none of the participants reported BPMN best practices for healthcare are most beneficial for process
working with the healthcare processes chosen for the modeling designers with a basic understanding of BPMN and the healthcare
task. We noticed that a few participants with a clinical back- domain.
ground relied on their general clinical knowledge and modeled In the future, we plan to test the best practices in real-world
details not specified in the process description. Nevertheless, healthcare modeling initiatives, also involving (senior) expert
we could not notice significant differences in the quality of the process modelers. Furthermore, we want to evaluate their use-
results based on the background of the participants. Second, we fulness for the representation of clinical practice guidelines and
acknowledge that time constraints imposed on the modeling task aim to extend the list of best practices by addressing the typical
may have affected the quality of the obtained models. Indeed, a challenges of inter-organizational healthcare processes. The best
few participants reported not having enough time to complete practices presented in this work focus specifically on modeling
the task and did not accurately model some of the challenges at challenges observed in the healthcare environment. Some of the
the end of the process description. However, in the final question- challenges and best practices might also be relevant to other
naire, we asked participants to report if they have had enough domains; however, this needs to be more systematically explored
time for modeling and, if not, to specify the aspects they could in the future.
not model. Thus, in our findings, we could distinguish actual
modeling challenges from issues caused by the lack of time. Declaration of competing interest
Third, our study was conducted with a rather limited sample
size of 28 participants. Thus, we acknowledge that additional The authors declare that they have no known competing finan-
studies are needed to generalize our results to a broader group cial interests or personal relationships that could have appeared
of subjects. Still, the obtained process models combined with to influence the work reported in this paper.
the qualitative feedback from our participants provided insights
into the practical use of the best practices, including suggestions Acknowledgments
for improving them. Finally, we chose to evaluate best practices
with graduate students instead of healthcare and IT profession-
We would like to thank Jan-Philipp Sachs and Erwin Böttinger
als. The participants in our study were representative of junior
of the Digital Health Center, HPI, at the University of Potsdam for
healthcare and IT experts since most of them also had prior
making contact to different hospitals and arranging five of the
professional experience in their respective fields of study. Thus,
presented process modeling initiatives.
we acknowledge that more studies are needed to generalize our
results to experienced process modelers and real clinical contexts.
References
However, the healthcare concepts captured by the best practices
are quite complex, and seeing that students with a background in [1] S.C. Buttigieg, P.K. Dey, D. Gauci, Business process management in health
medicine were able to understand and use them reasonably well care: current challenges and future prospects, Innov. Entrepreneurship
is promising for future developments, especially since health- Health 3 (1) (2016) 1–13, http://dx.doi.org/10.2147/IEH.S68183.
care process modeling initiatives are becoming more and more [2] M. Reichert, What BPM technology can do for healthcare process support,
multidisciplinary. in: Conference on Artificial Intelligence in Medicine in Europe, Springer,
2011, pp. 2–13, http://dx.doi.org/10.1007/978-3-642-22218-4_2.
[3] S. Ferrante, S. Bonacina, G. Pozzi, F. Pinciroli, S. Marceglia, A design
8. Conclusion and outlook methodology for medical processes, Appl. Clin. Inf. 7 (01) (2016) 191–210,
http://dx.doi.org/10.4338/ACI-2015-08-RA-0111.
In this paper, we have presented a set of BPMN best practices [4] J. Neumann, S. Franke, M. Rockstroh, M. Kasparick, T. Neumuth, Extending
BPMN 2.0 for intraoperative workflow modeling with IEEE 11073 SDC
for supporting process modeling in healthcare, tackling some
for description and orchestration of interoperable, networked medical
critical modeling challenges typical of this domain. We iden- devices, Int. J. Comput. Assist. Radiol. Surg. 14 (8) (2019) 1403–1413,
tified nine challenges which we categorized into three main http://dx.doi.org/10.1007/s11548-019-01982-6.
categories, namely (1) patient-related, i.e., challenges considering [5] H. Scheuerlein, F. Rauchfuss, Y. Dittmar, R. Molle, T. Lehmann, N. Pienkos,
the active role and needs of the patient throughout the health- U. Settmacher, New methods for clinical pathways—Business process mod-
eling notation (BPMN) and tangible business process modeling (t.bpm),
care process, (2) medical practice-specific, i.e., related to the
Langenbeck’s Arch. Surg. 397 (5) (2012) 755–761, http://dx.doi.org/10.
knowledge-intensive character of healthcare activities, and (3) 1007/s00423-012-0914-z.
medical-resource specific, dealing with dynamic and multidisci- [6] E. Rojas, J. Munoz-Gama, M. Sepúlveda, D. Capurro, Process mining in
plinary working behavior of health workers. For each challenge, healthcare: A literature review, J. Biomed. Inform. 61 (2016) 224–236,
we developed a list of best practices with different modeling http://dx.doi.org/10.1016/j.jbi.2016.04.007.
[7] M. Poulymenopoulou, F. Malamateniou, G. Vassilacopoulos, Emergency
options and abstraction levels and provided concrete examples healthcare process automation using mobile computing and cloud services,
showing the suitability of a particular option over others given J. Med. Syst. 36 (5) (2012) 3233–3241, http://dx.doi.org/10.1007/s10916-
specific modeling goals. The presented list might not necessarily 011-9814-y.
22
[8] Object Management Group, Business process model and notation (BPMN), [31] A. Russo, M. Mecella, On the evolution of process-oriented approaches
v2.0.2, 2013, URL http://www.omg.org/spec/BPMN/2.0.2/. for healthcare workflows, Int. J. Bus. Process Integr. Manag. 6 (3) (2013)
[9] M. Weske, Business Process Management - Concepts, Languages, Archi- 224–246, http://dx.doi.org/10.1504/IJBPIM.2013.056962.
tectures, third ed., Springer, 2019, http://dx.doi.org/10.1007/978-3-662- [32] N.Z. Haddar, L. Makni, H.B. Abdallah, Literature review of reuse in business
59432-2. process modeling, Softw. Syst. Model. 13 (3) (2014) 975–989, http://dx.doi.
[10] M. Kunze, P. Berger, M. Weske, N. Lohmann, S. Moser, BPM aca- org/10.1007/s10270-012-0286-4.
demic initiative-fostering empirical research, in: BPM (Demos), in: CEUR [33] A. Koschmider, H.A. Reijers, Improving the process of process modelling by
Workshop Proceedings, vol. 940, CEUR-WS.org, 2012, pp. 1–5. the use of domain process patterns, Enterp. Inf. Syst. 9 (1) (2015) 29–57,
[11] Object Management Group, Decision model and notation (DMN), v1.1, http://dx.doi.org/10.1080/17517575.2013.857792.
2016, Available at: http://www.omg.org/spec/DMN/1.1/. [34] E.S. Chung, J.I. Hong, J. Lin, M.K. Prabaker, J.A. Landay, A.L. Liu, Development
[12] OMG Healthcare Domain Taskforce, Field guide to shareable clinical path- and evaluation of emerging design patterns for ubiquitous computing,
ways, v. 2.0, 2019, online, Available at: https://www.omg.org/registration/ in: Proceedings of the 5th Conference on Designing Interactive Systems:
registration-field-guide.htm. Processes, Practices, Methods, and Techniques, 2004, pp. 233–242, http:
[13] E. Laurenza, M. Quintano, F. Schiavone, D. Vrontis, The effect of digital //dx.doi.org/10.1145/1013115.1013148.
technologies adoption in healthcare industry: a case based analysis, Bus. [35] E. Mathisen, J. Krogstie, Modeling of processes and decisions in healthcare-
Process Manag. J. 24 (5) (2018) 1124–1144, http://dx.doi.org/10.1108/ state of the art and research directions, in: IFIP Working Conference on
BPMJ-04-2017-0084. the Practice of Enterprise Modeling, in: LNBIP, vol. 134, Springer, 2012, pp.
[14] P. Dadam, M. Reichert, K. Kuhn, Clinical workflows — The killer application 101–116, http://dx.doi.org/10.1007/978-3-642-34549-4_8.
for process-oriented information systems? in: 4th International Conference [36] K. Kaiser, M. Marcos, Leveraging workflow control patterns in the domain
on Business Information Systems, BIS 2000, Springer London, London, of clinical practice guidelines, BMC Med. Inf. Decis. Mak. 16 (1) (2015) 20,
2000, pp. 36–59, http://dx.doi.org/10.1007/978-1-4471-0761-3_3. http://dx.doi.org/10.1186/s12911-016-0253-z.
[15] M. Rosemann, Potential pitfalls of process modeling: part a, Bus. [37] M. Peleg, Computer-interpretable clinical guidelines: A methodological
Process Manag. J. 12 (2) (2006) 249–254, http://dx.doi.org/10.1108/ review, J. Biomed. Inform. 46 (4) (2013) 744–763, http://dx.doi.org/10.
14637150610657567. 1016/j.jbi.2013.06.009.
[16] R. Müller, A. Rogge-Solti, BPMN for healthcare processes, in: Proceedings [38] Y. Shahar, S. Miksch, P. Johnson, The Asgaard project: a task-specific
of the 3rd Central-European Workshop on Services and their Composition, framework for the application and critiquing of time-oriented clinical
ZEUS 2011, in: CEUR Workshop Proceedings, vol. 705, CEUR-WS.org, 2011, guidelines, Artif. Intell. Med. 14 (1–2) (1998) 29–51, http://dx.doi.org/10.
pp. 65–72. 1016/S0933-3657(98)00015-3.
[17] J. Neumann, M. Rockstroh, S. Franke, T. Neumuth, BPMNSIX–A BPMN [39] M.A. Musen, S.W. Tu, A.K. Das, Y. Shahar, EON: A component-based ap-
2.0 Surgical intervention extension, in: 7th Workshop on Modeling and proach to automation of protocol-directed therapy, J. Am. Med. Inf. Assoc.
Monitoring of Computer Assisted Interventions (M2CAI), At MICCAI 2016, 3 (6) (1996) 367–388, http://dx.doi.org/10.1136/jamia.1996.97084511.
Athens, Greece, 2016. [40] P. Terenziani, S. Montani, A. Bottrighi, M. Torchio, G. Molino, G. Correndo,
[18] R. Braun, H. Schlieter, M. Burwitz, W. Esswein, BPMN4CP: Design and The GLARE approach to clinical guidelines: main features, Stud. Health
implementation of a BPMN extension for clinical pathways, in: IEEE Technol. Inf. 101 (6) (2004) 162–166, http://dx.doi.org/10.3233/978-1-
International Conference on Bioinformatics and Biomedicine, BIBM, IEEE, 60750-944-8-162.
2014, pp. 9–16, http://dx.doi.org/10.1109/BIBM.2014.6999261. [41] A.A. Boxwala, M. Peleg, S. Tu, O. Ogunyemi, Q.T. Zeng, D. Wang, V.L. Patel,
[19] B.S.S. Onggo, N.C. Proudlove, S.A. D’Ambrogio, A. Calabrese, S. Bisogno, R.A. Greenes, E.H. Shortliffe, GLIF3: a representation format for sharable
N.L. Ghiron, A BPMN extension to support discrete-event simulation for computer-interpretable clinical practice guidelines, J. Biomed. Inform. 37
healthcare applications: an explicit representation of queues, attributes (3) (2004) 147–161, http://dx.doi.org/10.1016/j.jbi.2004.04.002.
and data-driven decision points, J. Oper. Res. Soc. 69 (5) (2018) 788–802, [42] D.R. Sutton, J. Fox, The syntax and semantics of the PROforma guideline
http://dx.doi.org/10.1057/s41274-017-0267-7. modeling language, J. Am. Med. Inf. Assoc. 10 (5) (2003) 433–443, http:
[20] K. Zarour, D. Benmerzoug, N. Guermouche, K. Drira, A systematic liter- //dx.doi.org/10.1197/jamia.M1264.
ature review on BPMN extensions, Bus. Process Manag. J. 26 (6) (2019) [43] N. Mulyar, W.M. Van der Aalst, M. Peleg, A pattern-based analysis of
1473–1503, http://dx.doi.org/10.1108/BPMJ-01-2019-0040. clinical computer-interpretable guideline modeling languages, J. Am. Med.
[21] C. Combi, G. Pozzi, P. Veltri, Process Modeling and Management for Inf. Assoc. 14 (6) (2007) 781–787, http://dx.doi.org/10.1197/jamia.M2389.
Healthcare, CRC Press, 2017. [44] R. Gatta, M. Vallati, C. Fernandez-Llatas, A. Martinez-Millana, S. Orini, L.
[22] R. Lenz, M. Reichert, IT Support for healthcare processes - premises, Sacchi, J. Lenkowicz, M. Marcos, J. Munoz-Gama, M. Cuendet, et al., Clinical
challenges, perspectives, Data Knowl. Eng. 61 (1) (2007) 39–58, http: guidelines: A crossroad of Many Research Areas. Challenges and opportu-
//dx.doi.org/10.1016/j.datak.2006.04.007. nities in process mining for healthcare, in: International Conference on
[23] A. Jiménez-Ramírez, I. Barba, M. Reichert, B. Weber, C. Del Valle, Clinical Business Process Management, in: LNBIP, vol. 362, Springer, 2019, pp.
processes - the killer application for constraint-based process interactions, 545–556, http://dx.doi.org/10.1007/978-3-030-37453-2_44.
in: International Conference on Advanced Information Systems Engineer- [45] R. Vanwersch, L. Pufahl, I. Vanderfeesten, H. Reijers, The RePro technique:
ing, in: LNCS, Springer International Publishing, Cham, 2018, pp. 374–390, a new, systematic technique for rethinking care processes, in: BETA
http://dx.doi.org/10.1007/978-3-319-91563-0_23. Publicatie: Working Papers, Vol. 465, Citeseer, 2014.
[24] F. Ruiz, F. Garcia, L. Calahorra, C. Llorente, L. Gonçalves, C. Daniel, B. Blobel, [46] C.M. Chiao, V. Künzle, M. Reichert, Object-aware process support in
Business process modeling in healthcare, Stud. Health Technol. Inf. 179 healthcare information systems: Requirements, conceptual framework and
(2012) 75–87, http://dx.doi.org/10.3233/978-1-61499-086-4-75. examples, Int. J. Adv. Life Sci. 5 (1 & 2) (2013) 11–26.
[25] I. Svagård, B.A. Farshchian, Using business process modelling to model [47] E. Rolón, E. Aguilar, F. Garcia, F. Ruiz, M. Piattini, L. Calahorra, et al., Process
integrated care processes: Experiences from a european project, in: In- modeling of the health sector using BPMN: a case study, in: Proceedings
ternational Work-Conference on Artificial Neural Networks, in: LNCS, vol. of the First International Conference on Health Informatics, HEALTHINF,
5518, Springer, 2009, pp. 922–925, http://dx.doi.org/10.1007/978-3-642- 2008, pp. 173–178.
02481-8_140. [48] M.G. Rojo, E. Rolón, L. Calahorra, F.O. García, R.P. Sánchez, F. Ruiz, N.
[26] C. Combi, B. Oliboni, F. Zerbato, A modular approach to the specification Ballester, M. Armenteros, T. Rodríguez, R.M. Espartero, Implementation
and management of time duration constraints in BPMN, Inf. Syst. 84 (2019) of the business process modelling notation (BPMN) in the modelling
111–144, http://dx.doi.org/10.1016/j.is.2019.04.010. of anatomic pathology processes, Diagn. Pathol. 3 (1) (2008) S22, http:
//dx.doi.org/10.1186/1746-1596-3-S1-S22.
[27] A. Lanz, B. Weber, M. Reichert, Time patterns for process-aware infor-
[49] A. Grosskopf, J. Edelman, M. Weske, Tangible business process modeling–
mation systems, Requir. Eng. 19 (2) (2014) 113–141, http://dx.doi.org/10.
methodology and experiment design, in: International Conference on
1007/s00766-012-0162-3.
Business Process Management, Springer, 2009, pp. 489–500, http://dx.doi.
[28] N. Genon, P. Heymans, D. Amyot, Analysing the cognitive effectiveness
org/10.1007/978-3-642-12186-9_46.
of the BPMN 2.0 visual notation, in: Software Language Engineering, in:
[50] S. Barbagallo, L. Corradi, J.d.V. de Goyet, M. Iannucci, I. Porro, N. Rosso, E.
LNCS, vol. 6563, Springer Berlin Heidelberg, Berlin, Heidelberg, 2011, pp.
Tanfani, A. Testi, Optimization and planning of operating theatre activities:
377–396, http://dx.doi.org/10.1007/978-3-642-19440-5_25.
an original definition of pathways and process modeling, BMC Med. Inf.
[29] C. Combi, B. Oliboni, A. Zardini, F. Zerbato, Seamless design of decision-
Decis. Mak. 15 (1) (2015) 38, http://dx.doi.org/10.1186/s12911-015-0161-
intensive care pathways, in: IEEE International Conference on Healthcare
7.
Informatics, ICHI 2016, IEEE, 2016, pp. 35–45, http://dx.doi.org/10.1109/
ICHI.2016.9. [51] F. Zerbato, B. Oliboni, C. Combi, M. Campos, J.M. Juarez, BPMN-based
[30] C. Combi, B. Oliboni, A. Zardini, F. Zerbato, A methodological framework for representation and comparison of clinical pathways for catheter-related
the integrated design of decision-intensive care pathways—an application bloodstream infections, in: 2015 International Conference on Healthcare
to the management of COPD patients, J. Healthc. Inform. Res. 1 (2) (2017) Informatics, IEEE, 2015, pp. 346–355, http://dx.doi.org/10.1109/ICHI.2015.
157–217, http://dx.doi.org/10.1007/s41666-017-0007-4. 49.
23
[52] P. Mincarone, C.G. Leo, M.d.M. Trujillo-Martín, J. Manson, R. Guarino, [66] M. Ramos-Merino, L.M. Álvarez-Sabucedo, J.M. Santos-Gago, J. Sanz-Valero,
G. Ponzini, S. Sabina, Standardized languages and notations for graph- A BPMN based notation for the representation of workflows in hospital
ical modelling of patient care processes: a systematic review, Int. J. protocols, J. Med. Syst. 42 (10) (2018) 181, http://dx.doi.org/10.1007/
Qual. Health Care 30 (3) (2018) 169–177, http://dx.doi.org/10.1093/intqhc/ s10916-018-1034-2.
mzx197. [67] W. Yao, J. Rolia, S. Basu, S. Singhal, A. Kumar, A context-aware framework
[53] W.M. van Der Aalst, A.H. Ter Hofstede, B. Kiepuszewski, A.P. Barros, for patient navigation and engagement (CANE), in: 8th International
Workflow patterns, Distrib. Parallel Databases 14 (1) (2003) 5–51, http: Conference on Collaborative Computing: Networking, Applications and
//dx.doi.org/10.1023/A:1022883727209. Worksharing, CollaborateCom, 2012, pp. 316–325, http://dx.doi.org/10.
[54] B. Weber, M. Reichert, S. Rinderle-Ma, Change patterns and change support 4108/icst.collaboratecom.2012.250454.
features–enhancing flexibility in process-aware information systems, Data [68] R. Noumeir, Radiology interpretation process modeling, J. Biomed. Inform.
Knowl. Eng. 66 (3) (2008) 438–466, http://dx.doi.org/10.1016/j.datak.2008. 39 (2) (2006) 103–114, http://dx.doi.org/10.1016/j.jbi.2005.07.001.
05.001. [69] A. Alahmadi, B. Soh, A. Ullah, Automated health business process modelling
[55] L.H. Thom, M. Reichert, C. Iochpe, Activity patterns in process-aware and analysis for E-health system requirements elicitation, in: Pacific Asia
information systems: Basic concepts and empirical evidence, Int. J. Bus. Conference on Information Systems, Vol. 1, 2014, pp. 1–13.
Process Integr. Manag. 4 (2) (2009) 93–110, http://dx.doi.org/10.1504/ [70] M. Askari, Improving Quality of Fall Prevention and Management in El-
IJBPIM.2009.027778. derly Patients Using Information Technology: The Impact of Computerized
[56] H. Scholta, M. Niemann, P. Delfmann, M. Räckers, J. Becker, Semi-automatic Decision Support (Ph.D. thesis), University of Amsterdam, 2014.
inductive construction of reference process models that represent best [71] M. Martin, R. Champion, L. Kinsman, K. Masman, Mapping patient flow in a
practices in public administrations: A method, Inf. Syst. 84 (2019) 63–87, regional Australian emergency department: a model driven approach, Int.
http://dx.doi.org/10.1016/j.is.2019.03.001. Emerg. Nurs. 19 (2) (2011) 75–85, http://dx.doi.org/10.1016/j.ienj.2010.03.
[57] A.R. Hevner, S.T. March, J. Park, S. Ram, Design science in informa- 003.
tion systems research, MIS Q. (2004) 75–105, http://dx.doi.org/10.2307/ [72] H. Campbell, R. Hotchkiss, N. Bradshaw, M. Porteous, Integrated care
25148625. pathways, Br. Med. J. 316 (7125) (1998) 133–137, http://dx.doi.org/10.
[58] K. Peffers, T. Tuunanen, M.A. Rothenberger, S. Chatterjee, A design science 1136/bmj.316.7125.133.
research methodology for information systems research, J. Manage. Inf. [73] M. Reichert, B. Weber, Enabling Flexibility in Process-Aware Informa-
Syst. 24 (3) (2007) 45–77, http://dx.doi.org/10.2753/MIS0742-1222240302. tion Systems: Challenges, Methods, Technologies, Springer-Verlag Berlin
[59] R.J. Wieringa, Design Science Methodology for Information Systems and Heidelberg, 2012, http://dx.doi.org/10.1007/978-3-642-30409-5.
Software Engineering, Springer, 2014, http://dx.doi.org/10.1007/978-3- [74] M.L. Rosa, W.M. Van Der Aalst, M. Dumas, F.P. Milani, Business process
662-43839-8. variability modeling: A survey, ACM Comput. Surv. 50 (1) (2017) 2, http:
[60] B. Canovas-Segura, F. Zerbato, B. Oliboni, C. Combi, M. Campos, A. Morales, //dx.doi.org/10.1145/3041957.
J.M. Juarez, R. Marin, F. Palacios, A process-oriented approach for support- [75] S. Cheikhrouhou, S. Kallel, N. Guermouche, M. Jmaiel, Toward a time-
ing clinical decisions for infection management, in: 2017 IEEE International centric modeling of business processes in BPMN 2.0, in: International
Conference on Healthcare Informatics, ICHI, 2017, pp. 91–100, http://dx. Conference on Information Integration and Web-Based Applications &
doi.org/10.1109/ICHI.2017.73. Services, in: IIWAS ’13, ACM, 2013, pp. 154–163, http://dx.doi.org/10.1145/
2539150.2539182.
[61] Object Management Group, Unified modeling language, v2.5, 2015,
[76] P. Adolfsson, C.G. Parkin, A. Thomas, L.G. Krinelke, Selecting the appro-
Available at: http://www.omg.org/spec/UML/2.5/.
priate continuous glucose monitoring system–a practical approach, Eur.
[62] R. Braun, H. Schlieter, M. Burwitz, W. Esswein, BPMN4CP Revised–
Endocrinol. 14 (1) (2018) 24, http://dx.doi.org/10.17925/EE.2018.14.1.24.
extending BPMN for multi-perspective modeling of clinical pathways, in:
[77] K. Mackway-Jones, J. Marsden, J. Windle (Eds.), Emergency Triage, BMJ
2016 49th Hawaii International Conference on System Sciences, HICSS,
Publishing Group, 1997.
IEEE, 2016, pp. 3249–3258, http://dx.doi.org/10.1109/HICSS.2016.407.
[78] N. Russell, W.M. van der Aalst, A.H. Ter Hofstede, D. Edmond, Workflow
[63] S. Ferrante, S. Bonacina, F. Pinciroli, Modeling stroke rehabilitation pro-
resource patterns: Identification, representation and tool support, in:
cesses using the unified modeling language (UML), Comput. Biol. Med.
International Conference on Advanced Information Systems Engineering,
43 (10) (2013) 1390–1401, http://dx.doi.org/10.1016/j.compbiomed.2013.
Springer, 2005, pp. 216–232, http://dx.doi.org/10.1007/11431855_16.
07.012.
[79] R. Hall, D. Belson, P. Murali, M. Dessouky, Modeling patient flows through
[64] F. Cossu, A. Marrella, M. Mecella, A. Russo, G. Bertazzoni, M. Suppa, F.
the healthcare system, in: Patient Flow: Reducing Delay in Healthcare
Grasso, Improving operational support in hospital wards through vocal
Delivery, Springer, 2006, pp. 1–44, http://dx.doi.org/10.1007/978-0-387-
interfaces and process-awareness, in: IEEE International Symposium on
33636-7_1.
Computer-Based Medical Systems, CBMS, IEEE, 2012, pp. 1–6, http://dx.
[80] S. Di Saverio, A. Birindelli, M.D. Kelly, F. Catena, D.G. Weber, M. Sartelli,
doi.org/10.1109/CBMS.2012.6266329.
M. Sugrue, M. De Moya, C.A. Gomes, A. Bhangu, et al., WSES jerusalem
[65] E.M. Cutting, C.L. Overby, M. Banchero, T. Pollin, M. Kelemen, A.R.
guidelines for diagnosis and treatment of acute appendicitis, World J.
Shuldiner, A.L. Beitelshees, Using workflow modeling to identify areas to
Emerg. Surg. 11 (1) (2016) 34, http://dx.doi.org/10.1186/s13017-016-0090-
improve genetic test processes in the university of maryland translational
5.
pharmacogenomics project, in: AMIA Annual Symposium Proceedings
[81] C. Wohlin, P. Runeson, M. Host, M. Ohlsson, B. Regnell, A. Wesslén, Ex-
Archive, Vol. 2015, American Medical Informatics Association, 2015,
perimentation in Software Engineering, Springer-Verlag Berlin Heidelberg,
p. 466.
2012, http://dx.doi.org/10.1007/978-3-642-29044-2.
24

BPMN in Healthcare Challenges and Best Practices

Uploaded by

Copyright:

Available Formats

BPMN in Healthcare Challenges and Best Practices

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

BPMN in Healthcare Challenges and Best Practices

Uploaded by

Copyright:

Available Formats

Information Systems 107 (2022) 102013

Contents lists available at ScienceDirect

BPMN in healthcare: Challenges and best practices

1. Introduction variability [3]. Finally, process models allow different kinds of

Fig. 1. Simplified emergency care process shown as a BPMN process diagram.

Then, while he or she performs the physical examination, a nurse

• BPMN process diagrams can represent the control flow of a

5. BPMN best practices in healthcare 5.1. Patient-related challenges

In our projects, we often observed more complex time con-

You might also like