Exploring the Power of Triple Crown Process Modeling in Healthcare:

Sepsis Case

Camelia Maleki

and Frederik Gailly

Faculty of Economics and Business Administration, Ghent University, Tweekerkenstraat 2, Gent, Belgium

Keywords:

Patient Care Process, Triple Crown Modeling Approach, Process Improvement.

Abstract:

Effective process modeling plays a pivotal role in optimizing patient care processes within the continually

evolving healthcare landscape. This paper focuses on the application of the Triple Crown standard, which

encompasses the Business Process Model and Notation (BPMN), Case Management Model and Notation

(CMMN), and Decision Model and Notation (DMN), within the context of the sepsis diagnosis process.

Through an in-depth exploration of this case study, the paper uncovers the immense potential of these stan-

dards in empowering healthcare practitioners to streamline workﬂows, enhance decision-making at critical

junctures, and ensure the delivery of the highest quality care despite the diverse challenges inherent in patient

care processes. By dissecting key dimensions such as ﬂexibility, data and information ﬂow, complexity man-

agement, and decision points, this study provides valuable insights into how the Triple Crown approach can

signiﬁcantly enhance patient care process models.

1 INTRODUCTION

In today’s intricate healthcare landscape, healthcare

process modeling plays a pivotal role in efﬁciently

managing the ﬂexibility and complexity inherent in

these processes. It not only aids in identifying

and eliminating inefﬁciencies, and addressing bottle-

necks but also allows for dynamic adaptation to the

ever-evolving nature of healthcare procedures. Pro-

cess models empower healthcare providers and pol-

icymakers with the ability to make well-informed

decisions, thereby providing a framework for un-

derstanding and responding to ﬂuctuations caused

by emerging medical technologies, shifting patient

needs, evolving regulatory requirements, and other

factors. This enables healthcare professionals to

stay agile, predict potential challenges and opportuni-

ties, and make adjustments to ensure optimal patient

care(Pufahl et al., 2022).

Healthcare organizations are responsible for car-

rying out a variety of processes that vary in their

characteristics and needs, ranging from clinical pro-

cedures like diagnosis and treatment to organizational

and administrative tasks such as scheduling appoint-

ments and registering patients. These processes are

https://orcid.org/0000-0002-7818-2559

https://orcid.org/0000-0003-0481-9745

not isolated from one another, but rather intercon-

nected and dependent. Patient care processes have

become particularly important in recent years as they

consider both organizational requirements and clini-

cal tasks(Mans et al., 2015). In the patient care pro-

cesses, many sections require healthcare providers to

make decisions based on their knowledge and exper-

tise (Di Ciccio et al., 2015). These parts are dy-

namic, ﬂexible, and often referred to as ”knowledge-

intensive”. However, some tasks follow a speciﬁc

workﬂow or set of instructions, such as regular assess-

ments or performing a speciﬁc protocol(Rojo et al.,

2008).

To effectively model a process, it is imperative to

ﬁrst grasp its priorities and challenges. Lenz et al

(Lenz and Reichert, 2007) conducted an investigation

into the primary challenges associated with support-

ing and representing care processes, taking into ac-

count their unique characteristics. One major issue

arises from the restricted process ﬂexibility inherent

in existing traditional languages. This limitation en-

tails the ability of the implemented process to exe-

cute based on a loosely or partially speciﬁed model

that is completed at runtime. (Lenz et al., 2012).

Another concern relates to complexity management.

Since the primary goal of business process diagrams

is to enhance communication among process-related

stakeholders, overly complex diagrams can hinder

516

Maleki, C. and Gailly, F.

Exploring the Power of Triple Crown Process Modeling in Healthcare: Sepsis Case.

DOI: 10.5220/0012394000003657

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 17th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2024) - Volume 2, pages 516-528

ISBN: 978-989-758-688-0; ISSN: 2184-4305

their correctness and comprehension. Improper use

of a modeling language regardless of the character-

istics and requirements of a process can result in re-

duced clarity and increased complexity in model.(Jo

et al., 2019). Moreover, patient care processes de-

pend on medical knowledge and case-speciﬁc deci-

sions. This entails that healthcare professionals de-

pend on the amalgamation of the ﬁnest accessible ev-

idence, clinical expertise, and patient preferences to

arrive at well-informed decisions. To proﬁciently ma-

neuver through these processes, it becomes crucial

to establish streamlined information exchange and a

comprehensive depiction of the decision-making as-

pect within process models (Pufahl et al., 2022). In

light of these challenges and requirements, it becomes

evident that relying exclusively on a single model-

ing language may fall short of creating a compre-

hensive model, as different aspects of patient care

processes possess distinct characteristics and prior-

ities. Therefore, it may be essential to employ a

combination of modeling languages to comprehen-

sively capture the patient care process and represent

all its dimensions.(Pufahl et al., 2022; Mulyar et al.,

2008a). One choice for this is the Triple Crown of

OMG, which combines the BPMN (Activity-centric

language), CMMN (Artifact-centric language), and

DMN (Decision-centric language). The Triple Crown

approach is a popular choice due to its comprehensive

nature, adaptability, and standardization. It allows

healthcare providers to model both the workﬂow-

oriented and knowledge-intensive aspects of a pro-

cess and offers ﬂexibility by allowing organizations

to choose one or more of the standards based on their

speciﬁc needs(OMG Healthcare Domain Taskforce,

2020).

This paper aims to shed light on the substantial po-

tential inherent in the Triple Crown standard. By eval-

uating the synergistic capabilities of BPMN, CMMN,

and DMN in enhancing critical aspects of patient care

process models. In our pursuit of evaluating the mod-

eling languages within the healthcare landscape, we

deliberately chose to examine four critical dimen-

sions: ﬂexibility, complexity management, informa-

tion ﬂow, and decision-making due to their fundamen-

tal signiﬁcance in providing a comprehensive repre-

sentation of patient care processes. These four dimen-

sions are intricately interconnected and collectively

underpin the success of healthcare process modeling.

By focusing our research on these dimensions, we aim

to provide a comprehensive assessment of the Triple

Crown standard’s applicability and effectiveness in

the patient care processes. Ultimately, our goal is to

empower healthcare practitioners with the tools they

need to navigate the complex and ever-changing land-

scape of patient care, thereby enhancing the quality of

care delivered to patients.

2 BACKGROUND

The ”OMG Triple Crown” refers to the three ma-

jor standards developed by the Object Management

Group (OMG) for business process management

(BPM), decision management (DM), and case man-

agement (CM). These three standards are the Busi-

ness Process Model and Notation (BPMN)(OMG,

2011), the Decision Model and Notation (DMN), and

the Case Management Model and Notation (CMMN)

(OMG, 2016). Together, they provide a comprehen-

sive framework for modeling, analyzing, and manag-

ing complex business processes, decisions, and cases

(OMG Healthcare Domain Taskforce, 2020)

2.0.1 BPMN in Healthcare

Several studies have concentrated on the analysis of

BPMN’s beneﬁts within the healthcare sector, with

a focus on its application in speciﬁc medical scenar-

ios, as discussed below. Rolon et al. (Rol

on et al.,

2008) employed BPMN to visualize and enhance pro-

grammed surgical patient processes. Rojo et al. (Rojo

et al., 2008) conducted an examination of the utility

of BPMN in modeling anatomic pathology processes.

In a separate study (Barbagallo et al., 2015), BPMN

is used to model standard operating pathways, to ana-

lyze the challenges related to operating room planning

and scheduling.

2.0.2 CMMN in Healthcare

In the work of (Herzberg et al., 2015), they present

an approach aimed at addressing the requirements of

a ﬂexible healthcare process through the utilization of

CMMN. Additionally, they explore the establishment

of case monitoring and analysis by combining event

processing and case management. Another relevant

study, as proposed in (Mei et al., 2014), introduces a

CMMN-based model for care pathways in CHF (con-

gestive heart failure) processes.

2.0.3 Combination of BPMN, CMMN, and

DMN in Healthcare

In this section, we highlight studies that have lever-

aged a combination of OMG standards to model

healthcare processes. It’s important to note that

DMN cannot function independently and is typically

employed in conjunction with other modeling lan-

guages. For example, Wiemuth et al. (Wiemuth

Exploring the Power of Triple Crown Process Modeling in Healthcare: Sepsis Case

517

et al., 2017) demonstrated the utilization of a com-

bination of BPMN, CMMN, and DMN to model non-

deterministic medical processes. Likewise, Junger et

al. [2] conducted a comparative analysis of the ca-

pabilities of BPMN, CMMN, and a combination of

both notations in representing adaptable surgical pro-

cedures. The evidence-based decision-making is ex-

plored in (Combi et al., 2016). This study serves

as the foundation for the methodology discussed in

(Combi et al., 2017), which recommends the com-

bination of BPMN and DMN to depict decision-

intensive healthcare processes( particularly chronic

care processes).

Despite prior research on the utilization of BPMN,

CMMN, and DMN in the healthcare domain, a void

persists in the literature when it comes to applying all

three notations (the Triple Crown approach) within

healthcare processes and assessing its efﬁcacy. This

paper endeavors to bridge this gap by investigating

the Triple Crown approach within the realm of health-

care process modeling. Through a detailed analysis

of critical facets and an assessment of its capacity to

optimize workﬂows and elevate decision-making, this

study contributes to the growing body of literature on

healthcare process modeling.

3 METHODOLOGY

In this section, we have outlined a methodology that

facilitates the creation of a healthcare process model

utilizing the triple crown standard.

1. Deﬁne the scope of the process

(a) What is the goal of the process?

(b) What is the start and end point?

2. Deﬁne the key performance indicators. they

should reﬂect the critical success factors and the

value proposition of your process.

3. Deﬁne the process

(a) List of Activity groups, Artifacts, Events, and

Goals that are used in the Model

(b) Identify the relationship between elements

4. Choose the appropriate modeling language for

each part of the process, taking into consideration

the content captured and the key performance in-

dicators (KPIs)

5. Create the ﬁrst draft of the models

6. Continuously reﬁne and validate the model until

it is functional. Some techniques are listed below:

(a) Expert Review: Engage healthcare profession-

als and experts

(b) Comparative Analysis: Compare the model

with existing documentation and guidelines.

complies with relevant healthcare regulations

and standards

(d) Audit Trails and Documentation: Maintain

thorough documentation of the model’s devel-

opment and validation process

(e) Simulation: If possible use simulation software

to execute the process model under different

scenarios

Figure 1: Main steps of the proposed design methodology.

4 SEPSIS CASE STUDY

In many situations, including the sepsis diagnosing

process, there are no documented procedures that dic-

tate the way in which the care process is undertaken.

Consequently, there is a need to discover the process

from ﬁrst principles. During this phase of the method-

ology, the aim is to gain a thorough understanding of

the current healthcare process.

The study aimed to comprehensively understand

the sepsis diagnosing process in healthcare by gather-

ing data from various sources. Three primary meth-

ods were employed: event logs from a reputable

Dutch hospital, clinical guidelines, and interviews

with healthcare professionals. The event logs, con-

sisting of 15,214 recorded events related to 1,050 sep-

sis cases, were obtained from the hospital’s enterprise

resource planning (ERP) system. While valuable,

these event logs did not cover all aspects of the sepsis

process. To address this limitation and obtain a more

holistic perspective, interviews with healthcare pro-

fessionals and relevant clinical guidelines were uti-

lized. This combined approach provided nuanced de-

tails that enhanced the analysis, resulting in a rich

dataset for a thorough exploration of the sepsis pro-

cess. The diverse data sources enabled the researchers

to model and evaluate different aspects of the sepsis

process, highlighting the strengths and limitations of

various modeling languages.

HEALTHINF 2024 - 17th International Conference on Health Informatics

518

Figure 2: Patient administration process with BPMN.

4.1 Patient Administration

Patient administration is the functional, structural

consolidation of patient registration, admitting, and

other patient-generated activities in the hospital. It

is a simple and concise process (Table 2) requiring

no speciﬁc knowledge and low levels of ﬂexibility.

This part of the process begins when a patient arrives

through the acceptance process in the Entrance Room

and ends with sending the patient for sepsis triage

(Figure 2).

The choice to implement BPMN for the patient

registration process was based on its ability to effec-

tively capture structured, sequential, and task-centric

processes. BPMN aligns well with the distinct activ-

ity groups in patient registration, offering a straight-

forward representation of tasks. In contrast, CMMN

introduces additional elements like cases, stages, and

milestones, leading to unnecessary complexity for a

simple patient registration process. BPMN’s ﬂexi-

bility patterns, incorporating gateways and subpro-

cesses, make it more suitable for handling variations

in patient registration, whereas CMMN requires nav-

igating multiple paths and outcomes, making the pro-

cess more challenging to understand.

4.2 Patient Diagnostic Exams

Upon patient registration, the nurse prioritizes pa-

tients by severity, entering triage information into the

electronic health record (EHR). If sepsis is suspected,

the nurse initiates the formal sepsis pathway, docu-

menting the start in a clinical note in the EHR, and

promptly communicates with the ﬁrst available physi-

cian. This process primarily involves documenting

patient situations, categorized as a structured and rou-

tine process with low ﬂexibility and complexity. Con-

sequently, the decision was made to model this part

using BPMN.

In the second and main part of the diagnosis pro-

cess, the physician assesses the patient’s condition,

Figure 3: First part of diagnose process with BPMN.

drawing on clinical skills, patient information, and

diagnostic testing to ensure an accurate evaluation.

The diagnosis process involves gathering information

and employing clinical reasoning to identify the pa-

tient’s health problem. This process, as highlighted

by Rzepinski et al. (Rzepi

nski, 2007), is not entirely

predictable and demands ﬂexible execution. It com-

mences with initial examinations and patient inter-

views, concluding with the documentation of the di-

agnostic report.

The issue here is that the speciﬁc sequence and

set of activities in the patient examination process do

not follow a particular script and order, but rather are

based on a diagnostic procedure (e.g., information

discovered in one step can drastically alter the next

set of steps). In BPMN, we can represent this pro-

cess as an ”ad hoc subprocess” which is a group of

tasks designed for handling a speciﬁc case and can be

executed in any order (Ye et al., 2008). While both

BPMN ad hoc sub-processes and CMMN tasks can

involve tasks without a predeﬁned order, There are

some limitations of BPMN ad hoc sub-processes in

comparison with CMMN tasks. Here are some limi-

tations of BPMN ad hoc sub-processes:

• Unpredictable Flow: In an ad hoc sub-process,

the exact sequence of activities is not predeﬁned.

This can make it challenging to grasp the overall

process ﬂow, as the order of activities might vary

from instance to instance.

• Variability: With the possibility of repeated ac-

tivities and different ordering, the number of po-

tential paths through the process can grow signiﬁ-

Exploring the Power of Triple Crown Process Modeling in Healthcare: Sepsis Case

519

Table 1: Events and response activities after evaluating the results.

ID Event description Response

T0 10 minutes timer event notify the care staff to evaluate the patient’s status

T1 Alert for respiratory symptoms Doing Chest X-ray and/or CT scan

T2 Request culture Culture from a speciﬁc site

T3 Alert: Cardiovascular instability Echo exam

T4 High risk: Brain infection CFS analysis

T5 High risk: Request for Dialysis Dialysis

T6 surgical intervention is a consideration evaluation by a surgical team to assess the need for surgery

T7 Alert: presence of infection Prescribe Antibiotics

T8 Alert: have hypertension Administrator IVF

T9 Alert: renal failure BUN exam

T10 Alert: decreased oxygen saturation ABg, Pulse oximetry

Figure 4: Patient exam sub-process with CMMN.

cantly, making it more difﬁcult to understand and

analyze the process as a whole.

• Diagram Clutter: When activities are not con-

strained to a predeﬁned sequence, the diagram

can become cluttered with various arrows indicat-

ing conditional ﬂows and potential transitions be-

tween activities, making the visual representation

more complex.

• Loss of Structure: Traditional BPMN diagrams

have a clear structure with deﬁned start and end

points. Ad hoc sub-processes introduce a level of

non-linearity that might blur the structure, espe-

cially when used extensively.

• Cognitive Load: Readers of the diagram need

to mentally keep track of potential sequences and

conditional ﬂows, which increases the cognitive

load required to understand the process.

• Limited Standardization: Ad hoc sub-

processes can be interpreted and modeled differ-

ently by different people, leading to inconsisten-

cies in understanding

We will investigate how CMMN can address

BPMN limitations. As depicted in Figure 4, we mod-

eled investigations that are performed for sepsis ex-

ams in CMMN. The model includes four main stages:

Mandatory exams, patient initial exams, treatment,

and further investigation. All stages are initiated ac-

cording to the triggering event that represents the in-

cident escalation. The ﬁrst stage in the model is to

”perform an initial exam” which is depicted as 6 hu-

man tasks. The purpose of these tasks is to gather

real-time information about patient vital signs. Based

on the collected information from this stage and per-

forming the exams in the ”mandatory exams” stage,

the patient’s situation is evaluated through a Decision-

Task called ”evaluate exam results”. This task can

be modeled with a DMN decision table. This task

contains the events and their corresponding decisions

(response activity) based on the patient exam results

(T4, T5, T9, etc.) (Table 3). For example, the T7.

The presence of infection initiates the treatment stage,

which includes a Human Task called ”antibiotic se-

lection” and ”dosage determination”. The next stage

HEALTHINF 2024 - 17th International Conference on Health Informatics

520

Table 2: Comparing the data modeling capabilities languages.

Measures BPMN

CMMN

collections

Data Object with a ”Collection” property CaseFileItem used to represent collections

Relationships By Data Associations

Elements like association, dependency, re-

quired rule, and onPart condition

Meta data

Can be represented by annotations

CaseFileItemDeﬁnition can be used to rep-

resent metadata

Hierarchical data

BPMN can not represent hierarchical data By using Case File Items

Object oriented data

structure

BPMN is not designed to model object-

oriented data structures

using Case File object that is a container to

hold data and other objects related to a par-

ticular case instance

Complex data structures

BPMN is not well-suited for complex data

structures that involve multiple data types

using caseFileItem and PlanItem

lifecycle

BPMN does not have a speciﬁc notation to

represent lifecycles

CMMN can manage various states a case

ﬁle item goes through, like ”update”, ”re-

place”, ”add”

Figure 5: Simple example of hierarchical and nested data using caseﬁleitem.

of our investigation involves addressing various sce-

narios where patients may require additional exami-

nations triggered by different event escalations. For

instance, in the case of T4 (high risk: Brain infec-

tion), a critical Human Task must be executed: CPS

analysis.

The inherent ﬂexibility mechanisms of CMMN,

exempliﬁed in Figure 4, serve as valuable tools for ac-

commodating diverse scenarios and variations within

a process model. The escalation mechanism allows

the process to adapt by escalating tasks or cases when

predeﬁned conditions are met, ensuring timely res-

olution. Ordering and parallelism offer agility to

manage tasks sequentially or concurrently. Ad hoc

and human tasks permit the on-the-ﬂy inclusion of

tasks as needed, while dynamic task allocation en-

sures dynamic assignment of the right resources based

on changing requirements. Conditional expressions

in ”ifparts” provide dynamic control over case ele-

ment activation and completion. Milestones facili-

tate progress tracking and action triggering, and dis-

cretionary tasks offer optional activities that adapt

to case-speciﬁc requirements.(Routis et al., 2020;

Zensen and Kuster, 2018). These mechanisms col-

lectively eliminate the need to create multiple process

models, streamline processes, and enhance the adapt-

ability of a process(Andree et al., 2022; Routis et al.,

2023; Kurz et al., 2015).

This example demonstrates how CMMN success-

fully addresses the mentioned issues. for example,

CMMN leverages a combination of case context man-

agement, case stages, milestones, event listeners, dy-

namic task allocation, and ﬂexible dependencies, to

effectively manage maintenance challenges and ad-

dress the unpredictable ﬂow within case management

scenarios.

One more challenge during the triage and diag-

nostic process is ensuring the quality of the data and

information during the process(Albahri et al., 2018).

Representing complex data structures, such as hierar-

chical or nested data structures like trees or graphs,

with BPMN data objects can be difﬁcult. The patient

exam results, for example, contain essential details

about different exams, and other nested elements that

are critical for healthcare staff to access with the high-

est level of accuracy and completeness. We presented

a comparison of the data modeling capabilities of both

CMMN and BPMN This example demonstrates how

CMMN successfully addresses the mentioned issues.

for example, CMMN leverages a combination of case

Exploring the Power of Triple Crown Process Modeling in Healthcare: Sepsis Case

521

context management, case stages, milestones, event

listeners, dynamic task allocation, case plans, ad hoc

tasks, and ﬂexible dependencies, to effectively man-

age maintenance challenges and address the unpre-

dictable ﬂow within case management scenarios. or

for managing diagram clutter the notation includes

features like stages and plans that help organize and

structure tasks within a case.

An additional challenge during the triage and di-

agnostic process is ensuring the quality of data and

information (Albahri et al., 2018). Representing com-

plex data structures, such as hierarchical or nested

data structures like trees or graphs, with BPMN data

objects can be challenging. For instance, patient exam

results contain crucial details about various exams

and other nested elements critical for healthcare staff

to access with the highest level of accuracy and com-

pleteness. We provided a comparison of the data mod-

eling capabilities of both CMMN and BPMN.(Gagne,

2016; Von Rosing et al., 2014; Neskovic and Kirch-

ner, 2016). As shown in the table although both

BPMN and CMMN offer data modeling capabili-

ties, CMMN is particularly suited for handling intri-

cate data structures in case management procedures

and offers more sophisticated elements for model-

ing complex care process data. In addition, CMMN

comes with built-in access control capabilities and au-

dit trails including user roles and permissions, case

and task ownership, and user-deﬁned access control

rules. Consequently, CMMN offers a range of fea-

tures that can aid in preserving the integrity, and ac-

cessibility of patient data.

4.3 Diagnose Process

The ﬁnal diagnostic step involves determining sepsis

severity, a task heavily reliant on physician expertise

and intricate decision-making. While CMMN is ver-

satile for various processes, it may have limitations in

modeling complex decision scenarios. Though it sup-

ports simple decision-making with sentry and event

elements, CMMN might not be optimal for intricate

decision-making with multiple conditions and rules.

In this phase, models are created using standard

BPMN and DMN. Executable standard BPMN el-

ements are employed, with decision-making rules

translated into ﬁrst-order logic (FO(·)). The challenge

encountered during this part is the difﬁculty in mod-

eling certain rules using standard BPMN constructs

due to their complex logical conditions which can-

not be easily represented through a sequence of tasks

and gateways. While it may be possible to model

these rules using more advanced BPMN constructs,

such as nested subprocesses or complex gateways, it

would likely result in a highly complex and difﬁcult-

to-understand diagram that may not be practical for

a real-world scenario. As an example, consider the

following rule:







∀x[patient](Fever(x) ∧W BC(x) ∧Con f irmedin f ection(x)

∧(SBP(x) < 90 ∧ (HoT N(x) ∨ Lac > 4))∧ (∃y[labtest]

(ordered a f ter(y, x) ∧ ∆Lac(y, x) ≥ 0.5 ∧ ∆SBP(y, x) ≤ 20)∨

(∃z[medication](ordereda f ter(z, x) ∧Vasopressor(z)∧

∆Dose(z, x) ≥ 0.5 ∧ ∆SBP(z, x ) ≤ 20)))

→ Re f ractorysepticshock(x)







This rule states that if a patient has a fever, positive

blood cultures, and a conﬁrmed infection, and their

systolic blood pressure is below 90 mmHg, and either

they have evidence of tissue hypoperfusion (HoTN)

or lactate levels above 4 mmol/L, and they have ei-

ther experienced a signiﬁcant increase in lactate levels

or a decrease in systolic blood pressure by 20 mmHg

or more after receiving vasopressor medication, then

they should be diagnosed with refractory septic shock.

There are a few technical issues with modeling this

rule in BPMN:

• Expressing Temporal Relationships: BPMN

can express temporal relationships between activ-

ities using gateways and events, but it can be difﬁ-

cult to represent complex temporal relationships,

such as ”orderedafter” in the rule.

• Expressing Logical Conditions: BPMN can ex-

press simple logical conditions using gateways

and events, but more complex logical conditions

in the rule, can be difﬁcult to represent (e.g., the

change in lactate levels (∆Lac) between ”y” and

”x” is greater than or equal to 0.5, and the change

in systolic blood pressure (∆SBP) between ”y”

and ”x” is less than or equal to 20).

• Expressing Quantitative Conditions: BPMN is

not designed to represent quantitative conditions,

such as ∆Lac(y, x) ≥ 0.5. While it may be pos-

sible to use data objects or variables to represent

these values, it can be less intuitive than express-

ing them in a decision table.

• Handling Nested Expressions: The rule contains

nested expressions, such as the combination of

”OR” and ”AND” operators within parentheses.

While it is possible to represent nested expres-

sions in BPMN, it can make the diagram more

complex and harder to read.

However, this rule can be represented clearly and

concisely in a DMN decision table by breaking it

down into smaller logical expressions and assigning

them to decision inputs and outputs (Figure 6).

To compare BPMN and DMN in decision model-

ing, we have decided to limit our scope to a set of

rules for the sepsis severity diagnostic process that

HEALTHINF 2024 - 17th International Conference on Health Informatics

522

Figure 6: Decision table with the rule for refractory septic

shock.

can be modeled by both languages, excluding very

complex rules that BPMN may not be capable of

modeling. We have deﬁned speciﬁc rules for diag-

nosing the severity of sepsis based on clinical guide-

lines, sepsis protocols, and insights gathered from in-

terviews with healthcare experts (Dugar et al., 2020;

Taj et al., 2022). These rules serve as a framework

to assess the severity of sepsis cases and guide the

decision-making process in our model. Below is a list

of some examples of these rules:







∀x[patient](Fever(x) ∧W BC(x) ∧ SBP(x) > 90 → SIRS(x))

∀x[patient](Fever(x) ∧W BC(x) ∧Con f irmedin f ection(x)

∧SBP(x) > 90 ∧ Lac < 2 → Sepsis(x))

∀x[patient](Fever(x) ∧W BC(x) ∧Con f irmedin f ection(x)

∧SBP(x) < 90 ∧ Lac > 4 → Septicshock(x))

∀x[patient](Fever(x) ∧W BC(x) ∧Con f irmedin f ection(x)

∧SBP(x) < 90 ∧ HoT N(x) ∧ Lac > 4 → Septicshock(x))

∀x[patient](Fever(x) ∧W BC(x) ∧Con f irmedin f ection(x)

∧SBP(x) < 90 ∧ 2 < Lac < 4 → Seversepsis(x))

∀x[patient](Fever(x) ∧W BC(x) ∧Con f irmedin f ection(x)

∧SBP(x) < 90 ∧ 2 < Lac < 4 ∧ OD(x) → Septicshock(x))







BPMN allows representing decisions and their im-

pact or consequence respectively. However, BPMN

is not meant to represent the detailed decision logic

since modeling the decision logic often results in an

Intricate mess of nodes and connections (see Figure

8). According to integrated BPMN and DMN mod-

eling, decisions are not mapped to control ﬂow ele-

ments. Rather, decisions are externalized into a sepa-

rate DMN decision model that can be invoked by the

process.

To calculate decision table values, it is essen-

tial to outline the requirements for decision-making,

graphically represented in a Decision Requirements

Diagram (DRD). The DRD comprises four element

types: decision, input data, business knowledge

model, and knowledge source. Oval shapes like ”Pa-

tient history” represent input data, while ”evaluate pa-

tient result” and ”identify sepsis severeness” are de-

cision nodes. Process resources, such as physicians

and guidelines, are considered knowledge sources, di-

rectly inﬂuencing decision-making. Text annotations

can serve as knowledge sources or input data, depend-

ing on their information content.

Figure 7 provides an overview of decision model

execution. In the table’s upper left corner, ”unique”

signiﬁes the deﬁned hit policy for the decision table,

indicating that only one of the rows below can be true

when a decision is required.

5 DISCUSSION

In this section, we will explore our approaches in

terms of their capacity to handle complexity and ﬂex-

ibility.

5.1 Complexity Management

Analyzing BPMN, CMMN, and DMN reveals the po-

tential for a more comprehensive model through their

integration. However, our focus here is to assess

whether the combined model is more complex than

the standalone BPMN model. Currently, no widely

accepted uniﬁed complexity metric exists due to the

distinct purposes, elements, and scenarios these nota-

tions involve in business process management and de-

cision modeling. Each notation has unique elements

and objectives, making a universally applicable com-

plexity metric challenging. One potential solution

is adapting existing complexity metrics like nesting

depth and control ﬂow complexity (CFC) to consis-

tently assess the impact of using these languages to-

gether.(Kluza and Nalepa, 2012). Table 3 summarizes

metrics used for complexity measurement.

Table 3: Process model complexity metrics.

metrics Focus

NOA, NOAC Measure the activity complexity

CFC Measure the control-ﬂow complexity

Depth Evaluate the nesting of the process

The NOA metric calculates activity complexity

and was inspired by lines of code (LOC) metric.

NOAC metric considers both activities and control-

ﬂow elements for well-structured models. The for-

mula of CFC that captures the complexity of XOR-

split, OR-split, and AND-split constructs is as fol-

lows:

CFC

XOR−split

(a) = f an − out

(1)

XOR-split CFC is determined by the number of in-

duced states introduced with the split. For XOR splits,

the complexity corresponds to the fan-out of the split.

CFC

OR−split

(a) = 2

f an−out(a)

− 1 (2)

OR-split CFC is also determined by the number of

induced states introduced with the split.

CFC

AND−split

(a) = 1 (3)

For AND splits, the complexity is simply 1. Lastly,

Depth is deﬁned as the nesting of the process model.

If there is a split gateway, the depth is increased with

one. If there is a join gateway, the depth is decreased

with one. The cumulative sum is taken and the max-

imum of the cumulative sum is calculated for each

Exploring the Power of Triple Crown Process Modeling in Healthcare: Sepsis Case

523

Figure 7: Transform a BPMN model into DMN.

Table 4: Complexity metrics comparison.

Complexity metrics BPMN BPMN+DMN

NOA, NOAC 27, 56 14, 30

CFC XOR-Split 19 7

CFC OR-split 15 9

CFC And gateway 2 0

Total CFC 36 16

Depth 5 3

path. The nesting depth is the maximum of each path

value.

Table 4 shows that combining BPMN and DMN

reduces process model complexity. According to

anchez-Gonz

alez et al., 2011), a process model

with CFC OR-split decision nodes exceeding a value

of 14 can be incomprehensible; our BPMN process

had a value of 15. After integrating DMN, the value

decreased to 9, making it understandable. In sum-

mary, the BPMN-DMN combination streamlines pro-

cess model complexity by reducing gateways and

ﬂows. However, it introduces new intricacies related

to additional decision models. Despite this, the added

complexity shifts from the process model to the deci-

sion model. DMN decision tables facilitate the identi-

ﬁcation of overlapping and missing rules, allowing for

rule reordering and consolidation, leading to a more

concise decision-making process in the workﬂow, un-

like situations with complex sequential structures.

To perform a meaningful comparison of complex-

ity between BPMN and CMMN, it’s crucial to ensure

that we have an identical process model represented

in both notations. While we already have the patient

registration process represented in both languages as a

structured process, to assess the complexity of a ﬂex-

ible process model, we need to recreate what we’ve

modeled in CMMN using BPMN. However, model-

ing the entire CMMN model within BPMN poses a

formidable challenge, as certain elements are inher-

ently unsuitable for representation in BPMN. As a re-

sult, we have opted to narrow our focus to a speciﬁc

part of the CMMN model depicted in Figure 5, which

we will then represent using BPMN notation. (Figure

8).

Given these fundamental distinctions between

BPMN and CMMN, many established complexity

metrics may not be directly applicable to CMMN. For

example, one common metric is Activity Complex-

ity (AC), which quantiﬁes the nodes within a process

HEALTHINF 2024 - 17th International Conference on Health Informatics

524

Table 5: Complexity metrics comparision.

Flexible Structured

Complexity metrics BPMN CMMN BPMN CMMN

AC 66 5 16 14

CNC 1.28 0.83 1 1.07

Figure 8: Comparison of BPMN and CMMN complexity in a ﬂexible process.

graph. Assuming that all elements in a CMMN model

can be considered as nodes, AC can be calculated for

both BPMN and CMMN models (Marin, ). However,

certain complexity metrics, like Control Flow Com-

plexity (CFC), do not ﬁnd relevance in CMMN. This

is because CMMN lacks equivalents to the AND, OR,

or XOR nodes present in BPMN. Therefore, for com-

plexity measurement, we utilize Activity Complexity

(AC) to count the activities. Additionally, we em-

ploy a complexity metric known as the Coefﬁcient

of Network Connectivity (CNC) exploits the notion

of connectivity between elements to quantify struc-

tural complexity. Inspired by Graph theory, the given

metric explores the relation of the number of Arcs to

the number of Nodes. Higher values of it reveal a

dense model, which is more likely to contain errors

since the modeler has to perceive more connections

between nodes than in a model that is less dense. In

this metric, all activity elements are counted as nodes,

and all connectors are counted as arcs.

CNC = Arcs ÷ Nodes (4)

In CMMN, we assume elements that represent

various types of relationships in CMMN, including

Entry Criteria, Exit Criteria, On-Part Relationships,

and sentries, as connectors.

Figure 11 illustrates that even a small segment

of a CMMN model can become intricately complex

when translated into the context of BPMN, resulting

in 4! different execution sequences for each sequen-

tial ordering of activities. This unnecessary complex-

ity, combined with potential limitations in expressive-

ness, can hinder process model maintainability, re-

quiring multiple adjustments. While modeling ﬂex-

ible processes in CMMN may reduce the number of

activities and overall process model density, a com-

parison of CMMN and BPMN complexity in a struc-

tured process reveals no notable difference in AC and

CNC metrics between the two models. Consequently,

CMMN not only fails to reduce complexity in struc-

tured processes but also introduces additional com-

plexities to the model.

It is undeniably true that CMMN and DMN in-

troduce additional layers of complexity to the overall

model. Nevertheless, when dealing with knowledge-

intensive or decision-centric processes, attempting to

replicate the quality of CMMN or DMN using BPMN

invariably results in heightened complexity, and a di-

Exploring the Power of Triple Crown Process Modeling in Healthcare: Sepsis Case

525

Table 6: Catalog of ﬂexibility patterns.

Flexibility pattern CMMN BPMN Flexibility pattern CMMN BPMN

Flexible initiation Flexible reordering

Alternative entry-points + + Interleaving + +

Entrance skip + - Swap - -

Undeﬁned entry + - Momentary reordering - -

Momentary entry change - - Permanent reordering + \ - -

Permanent entry change - - Flexible elimination

Flexible termination Task skip + \ - -

Alternative exit points + + Foreseen bypass + +

Termination skip + + \ - Momentary task elimination - -

Undeﬁned exit + - Permanent task elimination - -

Momentary exit change - - Flexible extension

Permanent exit change - - Task invocation + -

Flexible selection Late creation + \ - -

Choice + + Momentary task insertion + \- -

Task substitution - - Permanent task insertion - -

Late selection + - Flexible Repetition

Permanent choice insertion + \ - - Redo + \ - -

Momentary choice insertion + \ - - Momentary loop insertion + \ - -

Flexible Concurrency Iteration + +

Parallelism + + Permanent loop insertion - -

Momentary Task Parallelization - -

Parmanent Task Parallelization - -

minished level of overall quality. In some cases, it

may even fall short of the comprehensiveness offered

by CMMN or DMN models. Therefore, it is impera-

tive that we exercise caution and judiciously employ

these modeling approaches only in situations where

they can yield optimal efﬁciency and performance.

5.2 Flexibility

Flexibility, particularly in the context of processes,

has undergone extensive exploration in academic re-

search, leading to the introduction of numerous tax-

onomies. One of the most comprehensive taxonomies

is presented by Mulyar et al. (Mulyar et al., 2008b).

This taxonomy delineates two types of process ﬂexi-

bility: ﬂexibility during design time, which addresses

foreseeable changes accommodated in modeled pro-

cess schemas, and ﬂexibility during runtime, allowing

alterations to the process instance after initiation. Or-

ganized based on this taxonomy, patterns are grouped

into eight distinct categories (Table 6). These pat-

terns facilitate high-level structural changes in pro-

cess models, encompassing actions such as adding,

deleting, selecting, or relocating activities and process

fragments.

BPMN, in its traditional form, is a static model-

ing language that deﬁnes processes at design time. It

is not inherently designed to handle dynamic or run-

time changes to process structures. So it can just han-

dle patterns like ”choice” by using gateways to al-

low the selection of one of several alternative tasks

at design time. CMMN offers a robust framework

for modeling and adapting knowledge-intensive pro-

cesses, providing ﬂexibility patterns for managing dy-

namic processes. It allows for adding, deleting, mov-

ing, and adapting activities and process fragments at

a high level of abstraction. However, there are sce-

narios where CMMN may not fully capture the dy-

namic and unpredictable nature of certain processes.

It is not designed to provide a high level of ﬂexibility

in modifying the case model at runtime. CMMN in-

troduces discretionary elements, allowing tasks and

stages to be manually activated during case execu-

tion, enabling execution-time planning for handling

unpredictability. These mechanisms offer some run-

time adaptability like ”late selection”, ”late creation”

or ”momentary choice insertion”. However, their ap-

plication relies on predictions made during the de-

sign phase, making CMMN most suitable for scenar-

ios requiring ﬂexibility within the originally designed

case structure. For extensive runtime changes or sit-

uations involving unpredictable events, extreme vari-

ability, emergent behaviors, and rapid changes, alter-

native workﬂow or process management systems may

be more appropriate

The sepsis diagnosis process, unlike emergency

room processes, is not entirely unstructured. Many

potential variations and changes in these processes

can be predicted during the design phase, based on

clinical guidelines and insights from domain experts.

Here are key ﬂexibility patterns crucial to this pro-

cess:

• Late Selection: Essential when test results or pa-

HEALTHINF 2024 - 17th International Conference on Health Informatics

526

tient conditions evolve over time, late selection is

crucial for adapting the care plan accordingly.

• Late Creation: When there’s a need to create new

tasks in response to unexpected patient develop-

ments or additional requirements during the diag-

nosis.

• Parallelism: Sepsis exams often involve multiple

concurrent tasks, such as laboratory tests, patient

assessment, and treatment initiation. The paral-

lelism pattern allows for handling these tasks si-

multaneously.

• Iteration and Momentary Loop Insertion: Sepsis

exams might involve iterative processes, such as

repeated assessments or treatments at speciﬁc in-

tervals. Using these patterns allows modeling of

recurring activities.

As shown in Table 6 CMMN is a well-suited model-

ing approach to efﬁciently accommodate the ﬂexibil-

ity patterns mentioned. Therefore, CMMN can be a

suitable choice for modeling and managing the ﬂexi-

ble parts of these processes.

6 CONCLUSION

In conclusion, this paper has illuminated the poten-

tial of the Triple Crown standard, comprising BPMN,

CMMN, and DMN, in effectively representing patient

care processes. Our exploration has delved into crit-

ical dimensions, including ﬂexibility, data and infor-

mation ﬂow, complexity management, and decision

points, vividly demonstrating how the Triple Crown

approach can substantially enhance patient care pro-

cess models. We have shown that, based on selected

KPIs, we can harness the combination of these lan-

guages to unlock their potential for creating more

comprehensive models. In this paper, we tackled

three challenges in patient care process modeling:

• Patients exhibit a variety of conditions and unique

needs that must be taken into account when mod-

eling a care process: Utilized CMMN to establish

distinct abstraction levels for representing pro-

cess variants and diverse paths, effectively captur-

ing various aspects of ﬂexibility while mitigating

complexity

• Complex decisions for diagnosing and treatment

of patients: DMN offers a modular decision re-

quirements hierarchy which allows for the decom-

position of complex decision logic into smaller,

more manageable components

• In patient care processes, achieving evidence-

based decision making requires a high-quality

ﬂow of information : CMMN excels in its capac-

ity to model a wide range of data structures, mak-

ing it a versatile choice for representing the di-

verse and complex information inherent in patient

care processes

Nevertheless, it is imperative to acknowledge the

challenges associated with CMMN and DMN, includ-

ing complexity, limited tool support, and integration

issues. To ensure the successful implementation of

these modeling languages, stakeholders must invest

time and resources in learning and addressing these

challenges.

To enhance the applicability of our ﬁndings, future

research endeavors should encompass evaluations and

analyses in our work, with a dedicated focus on vali-

dating the conclusions drawn in this paper.

REFERENCES

Albahri, O. S., Albahri, A. S., Mohammed, K., Zaidan, A.,

Zaidan, B., Hashim, M., and Salman, O. H. (2018).

Systematic review of real-time remote health moni-

toring system in triage and priority-based sensor tech-

nology: Taxonomy, open challenges, motivation and

recommendations. Journal of medical systems, 42:1–

27.

Andree, K., Ihde, S., Weske, M., and Pufahl, L. (2022). An

exception handling framework for case management.

Software and Systems Modeling, 21(3):939–962.

Barbagallo, S., Corradi, L., De Ville de Goyet, J., Iannucci,

M., Porro, I., Rosso, N., Tanfani, E., and Testi, A.

(2015). Optimization and planning of operating the-

atre activities: an original deﬁnition of pathways and

process modeling. BMC medical informatics and de-

cision making, 15:1–16.

Combi, C., Oliboni, B., Zardini, A., and Zerbato, F. (2017).

A Methodological Framework for the Integrated De-

sign of Decision-Intensive Care Pathways—an Appli-

cation to the Management of COPD Patients. Journal

of Healthcare Informatics Research, 1(2):157–217.

Combi, C., Oliboni, B., Zardiniy, A., and Zerbato, F. (2016).

Seamless design of decision-intensive care pathways.

In 2016 IEEE International Conference on Healthcare

Informatics (ICHI), pages 35–45. IEEE.

Di Ciccio, C., Marrella, A., and Russo, A. (2015).

Knowledge-Intensive Processes: Characteristics, Re-

quirements and Analysis of Contemporary Ap-

proaches. Journal on Data Semantics, 4(1):29–57.

Dugar, S., Choudhary, C., and Duggal, A. (2020). Sepsis

and septic shock: Guideline-based management.

Gagne, D. (2016). Case Management Model and Notation

(CMMN): An Introduction. (December).

Herzberg, N., Kirchner, K., and Weske, M. (2015). Model-

ing and monitoring variability in hospital treatments:

a scenario using cmmn. In Business Process Man-

agement Workshops: BPM 2014 International Work-

Exploring the Power of Triple Crown Process Modeling in Healthcare: Sepsis Case

527

shops, Eindhoven, The Netherlands, September 7-8,

2014, Revised Papers 12, pages 3–15. Springer.

st, G., Heri

cko, M., and Polan

c, G. (2019). Theo-

retical foundations and implementation of business

process diagrams’ complexity management technique

based on highlights. Software & Systems Modeling,

18:1079–1095.

Kluza, K. and Nalepa, G. J. (2012). Proposal of square met-

rics for measuring business process model complexity.

In 2012 Federated Conference on Computer Science

and Information Systems (FedCSIS), pages 919–922.

IEEE.

Kurz, M., Schmidt, W., Fleischmann, A., and Lederer, M.

(2015). Leveraging cmmn for acm: examining the

applicability of a new omg standard for adaptive case

management. In Proceedings of the 7th international

conference on subject-oriented business process man-

agement, pages 1–9.

Lenz, R., Peleg, M., and Reichert, M. (2012). Healthcare

process support: achievements, challenges, current re-

search. International Journal of Knowledge-Based

Organizations (IJKBO), 2(4).

Lenz, R. and Reichert, M. (2007). It support for healthcare

processes–premises, challenges, perspectives. Data &

Knowledge Engineering, 61(1):39–58.

Mans, R. S., Van der Aalst, W. M., and Vanwersch, R. J.

(2015). Process mining in healthcare: evaluating and

exploiting operational healthcare processes. Springer.

Marin, M. A. Complexity metrics for the case management

modeling notation (cmmn).

Mei, J., Li, J., Yu, Y., Li, X., Liu, H., and Xie, G. (2014).

Embracing case management for computerization of

care pathways. In MIE, pages 3–7.

Mulyar, N., Pesic, M., Van Der Aalst, W. M., and Peleg, M.

(2008a). Declarative and procedural approaches for

modelling clinical guidelines: addressing ﬂexibility

issues. In Business Process Management Workshops:

BPM 2007 International Workshops, BPI, BPD, CBP,

ProHealth, RefMod, semantics4ws, Brisbane, Aus-

tralia, September 24, 2007, Revised Selected Papers

5, pages 335–346. Springer.

Mulyar, N., van der Aalst, W. M., and Russell, N. (2008b).

Process ﬂexibility patterns.

Neskovic, S. and Kirchner, K. (2016). Using context infor-

mation and cmmn to model knowledge-. pages 17–21.

OMG (2011). Business Process Model and Notation . (Jan-

uary):538.

OMG (2016). Case Management Model and Notation 1.1

Beta. (March):124.

OMG Healthcare Domain Taskforce (2020). The BPM+

Field Guide to Shareable Clinical Pathways. page 125.

Pufahl, L., Zerbato, F., Weber, B., and Weber, I. (2022).

Bpmn in healthcare: Challenges and best practices.

Information Systems, 107:102013.

Rojo, M. G., Rol

on, E., Calahorra, L., Garc

ıa, F.

O.,

anchez, R. P., Ruiz, F., Ballester, N., Armenteros,

M., Rodr

ıguez, T., and Espartero, R. M. (2008). Im-

plementation of the business process modelling nota-

tion (bpmn) in the modelling of anatomic pathology

processes. In Diagnostic pathology, volume 3, pages

1–4. BioMed Central.

Rol

on, E., Aguilar, E., Garc

ıa, F., Ruiz, F., Piattini, M.,

Calahorra, L., et al. (2008). Process modeling of the

health sector using bpmn: a case study. In Proceed-

ings of First International Conference on Health In-

formatics, HEALTHINF 2008, volume 2, pages 173–

Routis, I., Bardaki, C., Nikolaidou, M., Dede, G., and

Anagnostopoulos, D. (2023). Exploring cmmn ap-

plicability to knowledge-intensive process modeling:

An empirical evaluation by modelers. Knowledge and

Process Management, 30(1):33–54.

Routis, I., Nikolaidou, M., and Anagnostopoulos, D.

(2020). Empirical evaluation of CMMN models: a

collaborative process case study. Software and Sys-

tems Modeling, 19(6):1395–1413.

Rzepi

nski, T. M. (2007). The structure of diagnosis in

medicine: Introduction to interrogative characteris-

tics. Theoretical Medicine and Bioethics, 28:63–81.

anchez-Gonz

alez, L., Ruiz, F., Garc

ıa, F., and Cardoso, J.

(2011). Towards thresholds of control ﬂow complex-

ity measures for bpmn models. In Proceedings of the

2011 ACM symposium on Applied computing, pages

1445–1450.

Taj, M., Brenner, M., Sulaiman, Z., and Pandian, V. (2022).

Sepsis protocols to reduce mortality in resource-

restricted settings: A systematic review. Intensive and

Critical Care Nursing, page 103255.

Von Rosing, M., White, S. A., Cummins, F., and De Man,

H. (2014). Business process model and notation-

BPMN. The Complete Business Process Handbook:

Body of Knowledge from Process Modeling to BPM,

1(January):429–453.

Wiemuth, M., Junger, D., Leitritz, M., Neumann, J., Neu-

muth, T., and Burgert, O. (2017). Application ﬁelds

for the new object management group (omg) standards

case management model and notation (cmmn) and de-

cision management notation (dmn) in the periopera-

tive ﬁeld. International journal of computer assisted

radiology and surgery, 12:1439–1449.

Ye, J., Sun, S., Song, W., and Wen, L. (2008). Formal se-

mantics of bpmn process models using yawl. In 2008

Second International Symposium on Intelligent Infor-

mation Technology Application, volume 2, pages 70–

74. IEEE.

Zensen, A. and Kuster, J. (2018). A comparison of ﬂex-

ible BPMN and CMMN in practice: A case study

on component release processes. Proceedings - 2018

IEEE 22nd International Enterprise Distributed Ob-

ject Computing Conference, EDOC 2018, pages 105–

114.

HEALTHINF 2024 - 17th International Conference on Health Informatics

528