Clinical Processes and Its Data, What Can We Do with Them?

Eric Rojas, Michael Arias and Marcos Sepúlveda

Computer Science Department, School of Engineering, Pontificia Universidad Católica de Chile, Macul, Santiago, Chile

Keywords: Process Mining, Healthcare Processes, Process Discovery, Healthcare Data, Process Mining Methodology.

Abstract: Global healthcare services have evolved over time, and nowadays they are expected to follow high-quality

optimized standards. Analyzing healthcare processes has become a relevant field of study, and different

techniques and tools have been developed to promote improvements in the efficiency and effectiveness of

these processes. There is a research field called process mining that can be used to extract knowledge from

the event data stored in the hospital information systems. With the help of this, it is possible to discover the

real executed process, examine its performance and analyze the resource interaction during its execution.

The goal of this article is to provide a bibliographic survey about the use of process mining algorithms,

techniques, and tools in the analysis of healthcare processes, providing a general overview about the main

approaches previously used and the information required to apply them in the medical field. We provide

important insights about data, algorithms, techniques and methodologies that are required to help answer

medical expert questions about their processes, motivating and inspiring a broader usage. So, if we have the

information and it is possible to analyze and understand the healthcare processes, why are we not doing it?

1 INTRODUCTION

Medical centers around the world perform a large

variety of processes that are relevant because,

depending on their proper execution, they have a

direct impact on people’s health. Therefore, there is

a clear need to ensure that the implementation of

these medical processes are carried out in the most

effective and efficient manner possible. The

information currently stored in healthcare

information systems might provide valuable insights

about how these processes are being performed

helping to propose potential improvement that could

enhance their performance.

Process mining has emerged as a research discipline

that can discover, monitor and improve real

processes by extracting knowledge from event logs

readily available in today's information systems

(Van der Aalst et al., 2012). This discipline, has a

great potential for application in the healthcare

domain because medical processes are considered

complex, large and with a lot of variability in time

(Homayounfar, 2012).

The objective of this work is to present a systematic

overview of the different approaches that have been

used to analyze healthcare processes using process

mining, including a compilation of types of data

needed (Kaymak & Mans, 2012; RS Mans & Aalst,

2013), questions that can be answered (RS Mans &

Aalst, 2013), methodologies that have been applied

(Ferreira, 2012; Van der Aalst et al., 2012), and

techniques/algorithms available to perform the

analysis (e.g. (Günther & van der Aalst, 2007;

Weijters & van der Aalst, 2003)). It also covers a

detailed analysis of the geographical location of the

main case studies that have been done in this domain

and the leading medical fields where it has been

applied. Gathering all this information should help

professionals, in both process mining and healthcare

domains, to identify new research opportunities to

improve healthcare services using process mining

techniques.

The outline of the paper is as follows. In section 2,

we introduce the basics of process mining in

healthcare. In section 3, we give a list of the main

approaches applied. Main challenges are explained

in Section 4 and the conclusions in Section 5.

2 RELATED WORK

2.1 Process Mining

Process mining is a relatively young research

discipline. It focuses on extracting knowledge from

data generated and stored in databases of (corporate)

642

Rojas E., Arias M. and Sepúlveda M..

Clinical Processes and Its Data, What Can We Do with Them?.

DOI: 10.5220/0005287206420647

In Proceedings of the International Conference on Health Informatics (HEALTHINF-2015), pages 642-647

ISBN: 978-989-758-068-0

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

information systems. Process-Aware Information

Systems (PAISs) (Dumas, Van der Aalst, & Ter

Hofstede, 2005) are systems that are readily able to

produce event logs. Specific examples of such

applications are ERP systems (e.g. SAP) and

Customer Relationship Management systems

(CRM). Event log data is not limited only to the data

from these applications, because many other systems

also provide useful data. Moreover, the data about a

complex process might not come from a single

source of information.

According to (Van der Aalst et al., 2007), it is

possible to store event information where: (i) each

event refers to an activity, (ii) each event refers to a

case, (iii) each event can have a performer, also

referred to as originator (the person executing or

initiating the activity), and (iv) each events have a

timestamp and are totally ordered.

There are three main types of process mining:

process discovery, conformance checking and

enhancement. In (Van der Aalst et al., 2012), it is

described that automatic process discovery allows

extracting process models from an event log;

conformance checking to monitor deviations by

comparing a given model and the event log, and

enhancement allows extending or improving an

existing process model using information about the

actual process recorded in the event log. It is

possible to extend the analysis through

organizational mining, the automatic construction of

simulation models, extension models, predicting

cases and other approaches.

2.2 Process Mining in the Health

Domain

Healthcare processes are seen as a domain that has

complex models and vary a lot in time

(Homayounfar, 2012). Being able to use techniques

to discover workflow models and analyze the

performance of them, are great opportunities to

examine the information stored in the events logs of

hospital information systems.

Using process mining techniques in healthcare

processes, allows not only to understand what is

really happening with them, but also, can generate

benefits associated with process efficiency,

improving the quality of services provided, as well

as a positive impact on the management of medical

centers. For this, it is relevant to take advantage of

the capabilities offered by processes mining to

collaborate in the construction of event logs and

analyze healthcare processes. Also, incorporated

medical knowledge can generate results that can

provide data to improve medical practices in

hospitals. The process mining research area has been

used in the healthcare processes field to discover

process models from event logs (Bose & Aalst,

2011; Mans, Reijers, van Genuchten, & Wismeijer,

2012; RS Mans & Schonenberg, 2009), do

conformance checking (Kirchner, Herzberg, Rogge-

solti, & Weske, 2012; Zhou, 2009) and evaluate

social networks analysis (Bose & Aalst, 2011; Lang,

Bürkle, Laumann, & Prokosch, 2008; Mans et al.,

2012).

3 MAIN APPROACHES APPLIED

In this section, the type of processes, the type of

data, the frequently posed questions, the

methods/algorithms, and the used methodologies,

are described. Besides, the case studies are

characterized according to their medical field and

their geographical location.

3.1 Types of Processes

It is critical to understand the different types of

processes that are executed in the hospital

environments, so the algorithms/techniques and

process mining tools are correctly applied.

According to (Dumas et al., 2005; Ferreira, 2012)

and (Kaymak & Mans, 2012), there are two types of

processes in the healthcare domain: the medical

treatment processes and the organizational

processes.

Medical Treatment Processes: These are the clinical

processes of managing a patient. These processes

include from the diagnostic actions based on

symptoms, to the execution of a series of actions, to

relief the patient.

Organizational Processes: These are the ones

focused on the organizational knowledge of the

process, capturing the collaborative information of

the healthcare professionals and their organizational

units.

3.2 Types of Data

Based on the type of process and the type of analysis

that is needed, it is necessary certain data to do it. In

this section, 2 classifications are described. The first

classification is the one included in (Kaymak &

Mans, 2012), based on the data used on their process

mining case study. The second one is presented on

ClinicalProcessesandItsData,WhatCanWeDowithThem?

643

(RS Mans & Aalst, 2013), and it is based on the data

source and its level of abstraction, accuracy,

granularity, directness, and correctness. To allow a

correct execution of the process mining techniques,

and give the expert the knowledge of its processes it

is necessary to have clarity of the type of data that it

is available and how it should be managed.

3.3 Frequently Asked Questions

The process mining techniques gives the medical

specialists the ability to answer some of the

questions related to their processes, acknowledging

them the improvement opportunities. According to

the questions that are raised by the specialists, data

and information must be gathered from different

sources. Four typical questions that want to be

answered by medical process specialists (RS Mans

& Aalst, 2013), are:

- What are the most followed paths and what

exceptional paths are followed?

- Are there any differences in care path followed

by different patient groups?

- Do we comply with internal and external

guidelines?

- Where are the bottlenecks in the process?

Besides these interrogations, we propose a fifth

question to include the organizational collaboration

process between the specialists:

- What are the roles and social relationships

between the medical staff?

These questions guide the use and application of the

process mining techniques, to provide correct and

accurate answers to the real needs of the specialists.

3.4 Used Methods and Algorithms

Through the tools available for process mining there

are several techniques implemented to execute the

correct and desired analysis. Some of the main

techniques that have been applied are: Trace

Clustering (Caron et al., 2014; RS Mans &

Schonenberg, 2009), Performance Sequence

Analyzer (Butler-Henderson, 2012a; Caron et al.,

2014), Fuzzy Miner (Kim, Kim, Song, Kim, & Yoo,

2013; RS Mans & Schonenberg, 2009), Alpha Miner

(Lang et al., 2008), Genetic Miner (Fei & Meskens,

2010; Lang et al., 2008), Heuristic Miner (Caron,

Vanthienen, & Baesens, 2013; Kaymak & Mans,

2012; Kim et al., 2013) and Conformance Checker

(Dewandono, Fauzan, Sarno, & Sidiq, 2013; Zhou,

2009).

3.5 Used Methodologies

Two main methodologies have been used: a

methodology including clustering techniques (Caron

et al., 2013; Doremalen, 2012), and a methodology

following the L* life-cycle model (Binder et al.,

2012; Van der Aalst et al., 2012).

3.6 Implementation Strategies

A classification of strategies used to implement

process mining is described. The first one is the

basic direct strategy, which involves the direct

application of the process mining tools to a set of

data extracted directly from a data source and used

to build an event log (Caron et al., 2014; Mans et al.,

2012; RS Mans & Schonenberg, 2009; Rebuge,

Lapao, Freitas, & Cruz-Correia, 2013). This strategy

has two basic challenges: data extraction and event

log construction. The second strategy is the semi-

automated, where the data and event log

construction is done through a specific solution.

These solutions connect to one or several data

sources and extract the correct data to build the

event log, but still need the knowledge of the

process mining tools to execute it (Helmering,

Harrison, Iyer, Kabra, & Van Slette, 2008). This

strategy has the disadvantage that is a local solution.

The third strategy to implement process mining is to

do it in a specific suite, where you can connect to the

data sources, extract the data, build an event log and

execute the process mining techniques. This has the

advantage that the person that will use the suite does

not need to know in detail how to connect to the data

sources, and how to use the process mining tools.

The disadvantage is that the suites are developed for

a specific environment and its data sources.

Examples of this implementation include the

Medtrix Process Mining Studio (Ferreira, 2012) and

the Emotiva Tool (Fernández-Llatas et al, 2013).

3.7 Types of Case Studies

There are several ways to build or execute a case

study using process mining techniques/algorithms.

They all include the use of data and the creation of

an event log, but can vary according to the process

mining tool they use. The tool they use determines

the type of case study, it can be the typical tools, a

new development made for the specific case study or

the use of the existing techniques with additional

techniques from outside the process mining field.

Following, is a proposed classification. The first

type defined is the basic case study, which takes the

HEALTHINF2015-InternationalConferenceonHealthInformatics

644

data from one or several data sources and builds the

event log and executes the process mining

techniques using the tools available. There is no new

implementation done in this type of case study and

the main objective is to give knowledge of the

healthcare process. Several case studies have been

done of this type (Mans, R., & Schonenberg, H.,

2008; RS Mans & Aalst, 2013; RS Mans &

Schonenberg, 2009). The second one is the case

study where a new technique or algorithm has been

developed to complement the actual available tools.

Some of these case studies are (Bose & Aalst, 2011;

Butler-Henderson, 2012b; Gupta, 2007). The third

type is the one that uses the process mining tools and

techniques available, but also incorporates

techniques from other fields like statistical analysis

(Butler-Henderson, 2012b), extended data analysis

(Bose & Aalst, 2011), data mining and CRISP-DM

(Zhou, 2009) and DECLARE (Grando, Aalst, &

Mans, 2011; Grando, Schonenberg, & Aalst, 2011).

These are important due to the mixture of

techniques.

3.8 Process Mining Tools

A group of software applications is available to do

analysis through process mining. The most used tool

is ProM, a tool that has a large number of

algorithms/techniques implemented and it is

considered as the “pluggable” environment for

process mining (van Dongen, et al, 2005). In

healthcare domain, ProM has been used in several

case studies, for instance: (Bose & Aalst, 2011;

Lang et al., 2008; RS Mans & Schonenberg, 2009;

Zhou, 2009). DECLARE is a flexible constraint-

based workflow management suite, used to model

medical guidelines in (Grando, Aalst, et al., 2011;

Grando, Schonenberg, et al., 2011). It is interesting

that the popular process mining toolkit DISCO is not

mentioned as much as expected in these case studies.

Only in (Perimal-lewis, Vries, & Thompson, 2014)

is reported as a main tool. An emerging process

mining tool called PmLab, has not been used so far.

3.9 Geographical Analysis and Medical

Fields

The use of process mining has been growing in the

last couple of years, becoming an important tool to

analyze the medical processes and generate

improvements opportunities. A quantitative count

and geographical classification was performed based

on the case studies available. The highest

concentration of case studies is in Europe, existing

only a few in North America, Asia and Australia. No

case studies have been done in Africa or South

America. In more detail, Netherlands is the country

that has published most studies, followed by

Belgium and Germany.

Besides the geographical analysis, the case studies

have been divided and classified into medical

domains. Some of these test cases include

Cardiology data, Caregiving Processes data,

Dentistry data, Diabetes data, Intensive Care Unit

data, Medication data, Oncology data, Radiotherapy

data and surgical data. Oncology and surgery (nine

and five case studies) are the medical fields where

most case studies have been done. In average two or

fewer case studies have been done per field.

4 CHALLENGES

The main challenges and limitations that experts

have found are the following: (1) Satisfying medical

protocols and guidelines, (2) Including medical

knowledge, (3) Including the physical information

and conditions of the patients, (4) Identifying and

accessing data sources, (5) Data integration from

different sources, (6) Quality of data (incorrect and

incomplete data), (7) Granularity and preprocessing

of the data, (8) Using real event logs and data (and

its complexity of cases), and not only synthetic data,

and (9) Building the correct and complete event log.

These limitations relate to data sources, event log

construction and how to include semantic

knowledge from the medical experts.

5 CONCLUSIONS

In process mining is possible to discover a process

model, perform conformance analysis and propose

opportunities to enhance any process. Healthcare

experts can benefit from these tasks, allowing them

to find process improvement opportunities. We

provide a useful bibliographic survey, which gives

an overview of main approaches needed to apply

process mining in the healthcare domain. Some

challenges and limitations were identified, related

with data sources, event log construction, and adding

semantic knowledge from medical specialists.

Overall, this article aims to serve as a motivational

guide that collects some case studies previously

conducted in the healthcare field, and provides an

outline about what must be taken into consideration

to carry out a process mining project. These outlines

ClinicalProcessesandItsData,WhatCanWeDowithThem?

645

involve the approaches explained and the different

kinds of algorithms, techniques and process mining

tools. We are able to gain significant insight as a

result of performing process mining in healthcare,

taking advantage of the data stored in the medical

information systems is what makes this possible.

REFERENCES

Binder, M., Dorda, W., Duftschmid, G., Dunkl, R., Fr, K.

A., Hronsky, M., & Rinderle-ma, S. (2012). On

Analyzing Process Compliance in Skin Cancer

Treatment : An Experience Report from the Evidence-

Based Medical Compliance Cluster ( EBMC2 ). In

CAiSE 2012 (pp. 398–413).

Bose, R., & Aalst, W. van der. (2011). Analysis of Patient

Treatment Procedures. Business Process Management.

Butler-Henderson. (2012a). Health Informatics and

Knowledge Management 2012 (Vol. 129).

Butler-Henderson. (2012b). Health Informatics and

Knowledge Management 2012 (Vol. 129).

Caron, F., Vanthienen, J., & Baesens, B. (2013).

Healthcare Analytics: Examining the Diagnosis–

treatment Cycle. Procedia Technology, 9, 996–1004.

Caron, F., Vanthienen, J., Vanhaecht, K., Limbergen, E.

Van, De Weerdt, J., & Baesens, B. (2014). Monitoring

care processes in the gynecologic oncology

department. Computers in Biology and Medicine, 44,

88–96.

Dewandono, R. D., Fauzan, R., Sarno, R., & Sidiq, M.

(2013). Ontology and Process Mining For Diabetic

Medical Treatment Sequencing. In 7th International

Conference on Information & Communication

Technology and Systems (ICTS) (pp. 171–178).

Doremalen, B. Van. (2012). Process Mining in Healthcare

Systems : An Evaluation and Refinement of a

Methodology.

Dumas, M., Van der Aalst, W. M. P., & Ter Hofstede, A.

H. M. (2005). Process-aware information systems:

bridging people and software through process

technology. John Wiley & Sons.

Fei, H., & Meskens, N. (2010). Discovering Patient Care

Process Models From Event Logs. 8th International

Conference of Modeling.

Fernández-Llatas, C., Benedi, J.-M., García-Gómez, J. M.,

& Traver, V. (2013). Process mining for

individualized behavior modeling using wireless

tracking in nursing homes. Sensors (Basel,

Switzerland), 13(11), 15434–51.

Ferreira, D. R. (2012). Business process analysis in

healthcare environments : A methodology based on

process mining, 37, 99–116.

Grando, M. A., Aalst, W. M. P. Van Der, & Mans, R. S.

(2011). Reusing a Declarative Specification to Check

the Conformance of Different CIGs. In BPM 2011

Workshops (pp. 188–199).

Grando, M. A., Schonenberg, M. H., & Aalst, W. Van

Der. (2011). Semantic-Based Conformance Checking

of Computer Interpretable Medical Guidelines. In

BIOSTEC 2011 (pp. 285–300).

Günther, C. W., & van der Aalst, W. M. P. (2007). Fuzzy

Mining – Adaptive Process Simplification Based on

Multi-Perspective Metrics. In Business Process

Management, 328–343.

Gupta, S. (2007). Workflow and process mining in

healthcare. Master’s Thesis, Technische Universiteit

Eindhoven, (May).

Helmering, P., Harrison, P., Iyer, V., Kabra, A., & Van

Slette, J. (2008). Process Mining of Clinical

Workflows for Quality and Process Improvement.

Mercy Health System (pp. 1–7).

Homayounfar, P. (2012). Process mining challenges in

hospital information systems. In Computer Science

and Information Systems (FedCSIS), 2012 Federated

Conference on, 1135–1140.

Kaymak, U., & Mans, R. (2012). On process mining in

health care. … Conference on, 1859–1864.

Kim, E., Kim, S., Song, M., Kim, S., & Yoo, D. (2013).

Discovery of Outpatient Care Process of a Tertiary

University Hospital Using Process Mining, 19(1), 42–

49.

Kirchner, K., Herzberg, N., Rogge-solti, A., & Weske, M.

(2012). Embedding Conformance Checking in a

Process Intelligence System in Hospital Environments.

In 2012 BPM workshops (pp. 126–139).

Lang, M., Bürkle, T., Laumann, S., & Prokosch, H.

(2008). Process Mining for Clinical Workflows :

Challenges and Current Limitations. eHealth Beyond

the Horizon – Get IT There S.K. Andersen et Al. (Eds.)

Mans, R., & Aalst, W. van der. (2013). Process mining in

healthcare: Data challenges when answering

frequently posed questions. … in Health Care, 107–

120.

Mans, R., Reijers, H., van Genuchten, M., & Wismeijer,

D. (2012). Mining processes in dentistry. Proceedings

of the 2nd ACM SIGHIT Symposium on International

Health Informatics - IHI ’12, 379.

Mans, R., & Schonenberg, H. (2008). Process mining

techniques: an application to stroke care. Studies in

Health …, 136, 573–578.

Mans, R., & Schonenberg, M. (2009). Application of

process mining in healthcare–a case study in a dutch

hospital. Biomedical Engineering, 425–438.

Perimal-lewis, L. U. A., Vries, D. D. E., & Thompson, C.

H. (2014). Health intelligence : Discovering the

process model using process mining by constructing

Start-to-End patient journeys, (Hikm), 59–67.

Rebuge, A., Lapao, L. V., Freitas, A., & Cruz-Correia, R.

(2013). A process mining analysis on a Virtual

Electronic Patient Record system. Proceedings of the

26th IEEE International Symposium on Computer-

Based Medical Systems, 554–555.

Van der Aalst, W. M. P., Adriansyah, A., Karla, A.,

Medeiros, A. De, Arcieri, F., Baier, T., … Song, M.

(2012). Process Mining Manifesto. In Business

Process Management Workshops, 169–194.

HEALTHINF2015-InternationalConferenceonHealthInformatics

646

Van der Aalst, W. M. P., Reijers, H. a., Weijters, a. J. M.

M., van Dongen, B. F., Alves de Medeiros, a. K.,

Song, M., & Verbeek, H. M. W. (2007). Business

process mining: An industrial application. Information

Systems, 32(5), 713–732.

Van Dongen, B. F., Medeiros, A. K. A. De, Verbeek, H.

M. W., & Weijters, A. J. M. M. (2005). The ProM

Framework: A New Era in Process Mining Tool

Support. In Applications and Theory of Petri Nets

2005, 444–454.

Weijters, A. J., & van der Aalst, W. M. P. (2003).

Rediscovering Workflow Models from Event-Based

Data using Little Thumb. Integrated Computer-Aided

Engineering, 10, 151–162.

Zhou, W. (2009). Process mining: acquiring objective

process information for healthcare process

management with the CRISP-DM framework.

ClinicalProcessesandItsData,WhatCanWeDowithThem?

647