=Paper=
{{Paper
|id=Vol-1920/BPM_2017_paper_155
|storemode=property
|title=PALIA-ER: Bringing Question-Driven Process Mining Closer to the Emergency Room
|pdfUrl=https://ceur-ws.org/Vol-1920/BPM_2017_paper_155.pdf
|volume=Vol-1920
|authors=Eric Rojas,Carlos Fernández-Llatas,Vicente Traver,Jorge Munoz-Gama,Marcos Sepúlveda,Valeria Herskovic,Daniel Capurro
|dblpUrl=https://dblp.org/rec/conf/bpm/RojasFTMSHC17
}}
==PALIA-ER: Bringing Question-Driven Process Mining Closer to the Emergency Room==
https://ceur-ws.org/Vol-1920/BPM_2017_paper_155.pdf
PALIA-ER: Bringing Question-Driven Process
Mining Closer to the Emergency Room
Eric Rojas1 , Carlos Fernández-Llatas2 , Vicente Traver2 , Jorge Munoz-Gama1 ,
Marcos Sepúlveda1 , Valeria Herskovic1 , and Daniel Capurro3
1
Computer Science Department, School of Engineering
Pontificia Universidad Católica de Chile, Santiago, Chile
{eric.rojas}@uc.cl, {jmun,marcos,vherskov}@ing.puc.cl
2
ITACA Institute. Universitat Politècnica de València, España
{cfllatas,vtraver}@itaca.upv.es
3
Internal Medicine Department, School of Medicine
Pontificia Universidad Católica de Chile, Santiago, Chile
{dcapurro}@med.puc.cl
Abstract. This paper presents PALIA-ER, a web-based tool for question-
driven process mining in Emergency Room. PALIA-ER uses Palia dis-
covery algorithm and includes model simplification and filtering features
specially domain-specific for ER. Most PALIA-ER functionalities can be
easily applied to other interdisciplinary contexts such as other healthcare
units, education, or logistics.
Keywords: process mining, emergency room, healthcare, BPM
1 Challenges of Question-Driven Process Mining in ER
In recent years, Process Mining has emerged as a powerful and popular tool
for the analysis of processes in a wide range of disciplines, such as Education or
Healthcare [2]. Healthcare is a good example of this interdisciplinary research: in
[3] the authors identify and characterize 74 case studies of process mining in the
medical domain. The event data recorded by the Hospital Information Systems
(HIS) is used to obtain detailed models and knowledge about the performed
processes. Within healthcare, Emergency Room (ER) processes have particu-
lar characteristics that make them di↵erent: they have short duration (e.g., the
episodes last hours instead of days), a well-defined end-point (e.g., an episode
typically ends with the patient being hospitalized or ambulatory discharged),
and particular phases (e.g., the triage, where the condition of the patient is ini-
tially evaluated, and a color-coded priority is assigned). Given the success of
Process Mining in healthcare, in [4] the authors proposed a specific question-
driven methodology to apply Process Mining in healthcare. This methodology
is specially indicated for interdisciplinary research, since it is not technology
driven, but instead the questions formulated by the domain-expert are consid-
ered the first-class citizens of the analysis. This interdisciplinary methodology
presents a series of challenges: 1) process models should be simple enough to be
2 Rojas, Fernández-Llatas, Traver, Munoz-Gama, Sepúlveda, et al.
Fig. 1. PALIA-ER Process Mining Tool
understood by researchers not familiar with the BPM discipline (e.g., health-
care professionals), 2) process models should be able to represent in an intuitive
way complex behavior typical of such interdisciplinary contexts, 3) event data
and process models should be able to be filtered and aggregated to represent
the process at the desired level of detail, 4) case statistics should be available
in order to lead the process analysis interactively, and finally, 5) it should in-
clude domain-specific filter possibilities (e.g., in the ER case, triage, discharge
destination).
This paper presents PALIA-ER, a tool for Question-driven Process Mining
analysis in ER. PALIA-ER addresses all the aforementioned challenges, and
provides the healthcare expert a tool with a complete set of features to analyze
their own ER processes. Notice that, although PALIA-ER is ER-specific, most
of the concepts and features presented in this work are easily adaptable to other
interdisciplinary domains.
This article is organized as follows: Section 2 presents PALIA-ER and its
main features. Two illustrative case examples using the tool are illustrated in
Section 3. Finally, Section 4 concludes the article.
2 PALIA-ER Process Mining Tool
In this article we present PALIA-ER, a web-based process mining tool designed
for question-driven process mining analysis of the ER domain. Figure 1 shows
an overview of the tool main elements: the main center panel of the tool is
PALIA-ER 3
b. Activity Naming
a. ER filters
c. Statistics
Fig. 2. PALIA-ER Features
used to display the discovered process models and the process statistics; the side
bar on the left contains the menus to set the filters, model simplifications, and
discovery parameters; finally the panel at the bottom displays information about
the analyzed process. PALIA-ER has several features specially designed for ER.
In particular, the main ones are:
B Interdisciplinary Process Models and Discovery: Process Mining tools for in-
terdisciplinary research require process models to be easy to interpret by
actors of all disciplines. PALIA-ER uses the Palia algorithm [1] to automat-
ically discover graph-based process models from event data. Those models
were first used for indoor location systems data analysis, where complex in-
formation needed to be presented in a comprehensive way [1], making them
suitable for other complex domains such as healthcare and ER. For the sake
of space, we refer the reader to [1] for more details on the models or the
discovery algorithm.
B Model Simplification: PALIA-ER includes features to easily merge or remove
activities in order to provide high or low level overviews of the same process,
depending on the question being answered. Figure 2b partially illustrates
this feature, where providing an existing alias to an activity or unchecking
it collapses or hides such activity.
B ER Domain Specific Filters: PALIA-ER includes case filters based on transver-
sal case properties, common in other process mining tools. For example, filter
by date or filter by duration, among other examples. However, the tool also
provides easy access to ER domain specific filters. For example, filter by
4 Rojas, Fernández-Llatas, Traver, Munoz-Gama, Sepúlveda, et al.
triage color (i.e., the color-based priority of the ER episode assigned dur-
ing the triage phase), or filter by last discharge (i.e., the type of discharge
of the last visit, including ambulatory discharge or hospitalize discharge for
example). Other examples include filters by the general characteristics of the
patient, such as age or gender. For the sake of space, we refer the reader to
[4] for more details on the ER domain specific reference model. Figure 2a
partially illustrates this feature.
B Statistics: PALIA-ER provides detailed and aggregated statistics about the
cases, allowing the final user to go back and forward on an interactive analysis
of the process, filtering such cases not relevant for answering the current
question. Figure 2c partially illustrates this feature.
To demonstrate the use of PALIA ER, a screencast with a walk-through of
the tool can be found at http://pmuc.ing.puc.cl/tools/paliaer/.
3 Case Examples
In this section we illustrate the applicability and potential impact of the tool,
by applying it to answer two frequently asked questions by experts in the ER,
following our methodology [4]. The scope of the analysis was provided by the
needs of the ER experts, who also validated the results. Given the space con-
straints, the cases are merely teased, but similar, more extensive analysis can be
found in [4].
Q1: What is the followed process by all the ER episodes? : The first
question relates to the general view of the ER unit. In order to analyze the
high-level view, the grouping feature shown in Figure 2b is used to group all the
activities in 5 main ER stages (Triage and Diagnostic, Examination, Referral,
Treatment and Discharge). Then, the relations between the ER stages and the
start and end points of the process are analyzed. The resulting process model
can be seen in Figure 3. A more detailed analysis concludes that significant
interactions are been held between the Triage and Diagnostics, Examinations
and Treatment stages, followed by the Discharge stage and finally, with less
interaction, the Referral stage.
Q2: What is the followed process for patients over 70 years? : Re-
garding the analysis necessary to answer Q2: First, the data was filtered accord-
ing to the age of the patient, selecting only patients with 70 years or more. The
discovered process model can be seen in Figure 4. When we analyze the statis-
tics of these episodes (cf. Figure 4), we discover one particular characteristic:
hospitalization is the highest destination after the patient has been treated in
the ER. A more detailed analysis concludes that in these cases, hospitalization
is higher than the typical episodes studied in Q1.
4 Conclusions
This paper presented PALIA-ER, a tool for a question-driven process mining
for the ER domain. Two conclusions arise: First, the tool can help reduce the
PALIA-ER 5
Fig. 3. Q1 Process Model
Fig. 4. Q2 Process Model and Statistics
dependency on the process mining expert from the clinicians looking for solutions
to analyze their processes. And second, even though the PALIA-ER solution is
ER-specific, it could potentially be applied to other interdisciplinary domains.
References
1. Fernandez-Llatas, C., Lizondo, A., Monton, E., Benedi, J.M., Traver, V.: Process
mining methodology for health process tracking using real-time indoor location sys-
tems. Sensors 15(12), 29821–29840 (2015)
2. Mans, R., van der Aalst, W.M.P., Vanwersch, R.J.B.: Process Mining in Healthcare
- Evaluating and Exploiting Operational Healthcare Processes. Springer Briefs in
Business Process Management, Springer (2015)
3. Rojas, E., Munoz-Gama, J., Sepúlveda, M., Capurro, D.: Process mining in health-
care: A literature review. Journal of Biomedical Informatics 61, 224 – 236 (2016)
4. Rojas, E., Sepúlveda, M., Munoz-Gama, J., Capurro, D., Traver, V., Fernandez-
Llatas, C.: Question-driven methodology for analyzing emergency room processes
using process mining. Applied Sciences 7(3), 302 (2017)