=Paper=
{{Paper
|id=Vol-2728/master6
|storemode=property
|title=Ontological Model for Decision Support About Bariatric Surgery
|pdfUrl=https://ceur-ws.org/Vol-2728/master6.pdf
|volume=Vol-2728
|authors=Glauber Muzyka Oyarzabal Nunes,Rita Cristina Galarraga Berardi
|dblpUrl=https://dblp.org/rec/conf/ontobras/NunesB20
}}
==Ontological Model for Decision Support About Bariatric Surgery==
https://ceur-ws.org/Vol-2728/master6.pdf
Ontological Model for Decision Support about Bariatric
Surgery
Gláuber M. O. Nunes1, Rita C. G. Berardi1
Departamento Acadêmico de Informática
1
Universidade Tecnológica Federal do Paraná (UTFPR) – Curitiba – PR – Brazil
glauber@alunos.utfpr.edu.br, ritaberardi@utfpr.edu.br
Abstract. Bariatric surgery is a type of treatment for severe obesity but it can
bring several risks to the patient. In order to evaluate those risks correctly,
health professionals need to have access to the complete health history of the
patients. However, information systems are not always structured so that they
can extract this complete history, due to the existence of data silos and distinct
vocabularies. The use of linked data, ontologies and semantic rules
technologies is one of the possible solutions to this challenge. This research
proposal aims to test if an ontological model can map the complete health
history of patients who are candidate to bariatric surgeries, highlight risks
and support decisions. The research will use data from UNIMED Paraná.
1. Introduction
Obesity prevalence has been increasing in almost all the world in the last decades and it
is considered today a global epidemic. The World Health Organization [2018] estimated
that there were approximately 1.9 billion overweight adults and at least 650 million
obese adults in the world in 2016. In Brazil, a national research with self-reported
weight and height data showed that the obesity prevalence increased by 60% in ten
years: from 11.8% in 2006 to 18.9% in 2016 [BRASIL, 2017]. Obesity is a complex
disease, which requires preventive strategies in order to control and treat it.
Bariatric surgery is a type of treatment for cases of severe obesity with failure of
clinical treatment, reducing the patient mortality rate and attenuating comorbid
conditions, according to a thirty-year study [SJOSTROM, 2008]. However, this surgery
has multiple risks, being usually indicated only if the patient cannot lose weight or
sustain the weight loss even with appropriate medical care for over two years.
Bariatric surgery is offered in both the Sistema Único de Saúde (SUS) and the
supplementary health system. Supplementary health has a fundamental role in the health
area. In Brazil, over 47 million people have a health insurance plan [AGÊNCIA
NACIONAL DE SAÚDE, 2020]. In that system, UNIMED is one of the most
remembered health insurance providers and has a 37% share in the national market
[UNIMED, 2020]. This research will be done using data from UNIMED do Estado do
Paraná.
Doctors need to have access to the complete health history of the patients who
are candidate to bariatric surgery in order to properly care for them. Some of this data is
already digitalized in information systems. However, the data is not always organized so
that the information can be extracted because of the use of different information
Copyright © 2020 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
exam exam report
systems. Part of hospital
this data is in silos, with distinct data bases, increasing the challenge of
collecting the health history in a homogenous manner. Besides that, it becomes harder
bariatric surgery
to interpret the data and predict some type of risk to the patient after the surgery. It is
not possible to have a vision of the complete health examhistory
price of the patient to allow the
health professional to understand the diseaseexam better. exam report
In UNIMED do Estado do Paraná, data about the patients who are candidate to
exam price
bariatric surgery is split into two distinct data bases: one base for the administrative data
and one base for the clinical data, according to Figure 1. The bases use different
vocabularies
doctor without a total
patient standardization of the medical vocabulary.
exam-name exam
exam-price petitioner
exam-place laboratory
patient report
doctor patient
administrative data clinical data
Figure 1. Integration of administrative and clinical data
Data integration using traditional technologies like SQL is not viable due to the
complexity of the bases and the vocabulary involved. In a similar way, re-creating the
database structures would require huge investments of time and money. Therefore,
ontology technology could be an interesting alternative to unify the administrative and
clinical data, helping the clinical decision-making process. The use of ontology is
presented as an option because it is a data model with domain view, adding a domain
layer without necessarily changing the data in their original format. That allows for
flexibility in the integration format. Ontologies can be easily expanded or modified,
differently from the application databases.
2. Objectives
This research proposal aims to experiment if an ontological model can provide an
integrated view of the complete health history of patients who are candidate to bariatric
surgery, in order to help the medical decision-making, solving the challenges of distinct
databases and heterogeneous vocabularies.
Besides that, the specific objectives of the project are:
• Unify in a centralized way the administrative and clinical data of the patients;
• Compile knowledge that allows the use of the centralized data to evaluate the
risks associated to bariatric surgery and its comorbid conditions;
• Verify if the model answers the competency questions with real data of patients
who are candidate to bariatric surgery.
3. Justification
The obesity issue has been known since ancient times. Hippocrates said that the “sudden
death is more common in those who are naturally fat than in the lean” [CHADWICK,
1950]. Nowadays, obesity is defined as “eating more food than the organism can
metabolize and use, causing this excess to accumulate in the form of fat” [PORTER,
2011]. There are multiple types of obesity treatment, such as pharmacological treatment,
dietetical treatment, cognitive-behavioral therapy, heterodox therapies and nutritional
supplements. Bariatric surgery is usually indicated only after two years of regular
medical treatment [ABESO, 2016].
This being the case, there is interest in improving the quality of decisions
regarding bariatric surgery, in order to help health professionals to choose the best
treatment for obesity. This theme will be initially studied in the supplementary health
context because of the organized disposition of the data. The open data from SUS does
not allow for the individual identification of patients, rendering the trajectory creation
impossible. However, it is important to note that the research and model to be developed
could be used in the public health system if the open data allows in the future.
It is also important to highlight that decision support systems were introduced in
the 1960s and have been developed since that date [MCCALLIE, 2016]. When applied
in the health area, their objective is to help health professionals with the patient
treatment, being especially useful when it is needed to analyze big amounts of
information. Even though, the health professionals need to have the final decision in the
decision-making process and the system should never make the decisions alone
[CRESSWELL, 2013].
4. Systematic review
A systematic review was done in order to identify scientific papers that use ontology for
data integration related to obesity treatment. Besides that, it was aimed to know
different perspective regarding this treatment. This allowed for evaluating solutions in
this area.
The systematic review was done on February 2020. The utilized data bases were
Biblioteca IEEE Xplore, Biblioteca Digital ACM, Scopus, Springer Link, Periódicos
Capes and Google Scholar. The review was done using the methodology presented by
Kitchenham [2004]. The search was done using the words [ontology AND “linked data"
AND obesity] in any part of the text. Texts in English and Portuguese were considered.
The initial search found 586 articles. After applying the filtering criteria, we selected 24
articles for reading the full-texts.
We found out that there are several articles in literature regarding the use of
ontologies for data integration in the obesity treatment. However, there is no article
about this theme related to the bariatric surgery context. The article that was closest to
this area studied the use of ontologies for childhood obesity treatment [El-Hachem et al,
2012] and the article does not mention bariatric surgeries. An ontology similar to the
OMDP by Chen et al [2012] will be developed in this research but in the context of
bariatric surgery. Besides that, it was noticed that the most common challenge faced
during the data use for obesity treatment in literature is the data heterogeneity. The data
from multiple sources, such as administrative, utilization, social network and health data
are in distinct formats and usually use its own vocabularies.
5. Methodology
Figure 2 shows a flowchart with the development steps of the methodology. Each step
has a number that will be used to identify it later.
Figure 2. Methodology
Step 1 is the systematic review to identify relevant scientific papers about
ontologies applied to the obesity treatment. This step has already been done and the
results are a little bit detailed in section 5 of this text.
In order to define the terms and relations in the ontology, it is necessary to elicit
the knowledge in step 2. Sub-step 2.1 involves a bibliographic search about the medical
domain with focus on the complete health history of patients who are candidate to
bariatric surgery and their risk factors. It will be searched in the literature what data
should be part of the ontology about comorbid conditions. Sub-step 2.2 consists of
discussions with experts, including professional from UNIMED, to gather knowledge.
Sub-step 2.3 involves the documental analysis of the UNIMED database, in order to
know exactly what data is available in the databases and have more information to build
the ontology.
Steps 3-9 correspond to the ontology engineering and follow the ontology
development process known as Ontology 101, described by Noy et al [2001]. They
correspond to steps 2 to 7 of Ontology 101: Develop competency questions, consider
reusing existing ontologies (for example, the terminologies SNOMED CT, LOINC and
RxNorm have been developed significantly in the last decades [Bodenreider, 2018]),
enumerate important terms in the ontology, define the classes and the class hierarchy,
define the properties of the classes, define the restrictions of the properties and create
instances. The instance creation step involves sub-step 8.1, data cleaning. It is possible
that part of the data from UNIMED is unstructured or semi-structured. In that case, a
manual process of structuring and preparing the data will be needed.
Step 10 refers to the creation of semantic rules. This step will allow the creating
rules to map the risks of the bariatric surgery, in order to help the health professional to
make decisions. These semantic rules will be developed based on the bibliography
search of the domain (step 2.1) and discussions with experts (step 2.2).
Finally, the research ends at step 11, which aims to measure the performance of
the developed ontology. Two criteria will be used for this evaluation: data integration
and classification of the treatment steps and treatment risk. The first criterion will
measure the ontology performance in representing the health history of the patients in
an integrated way. If the ontological model can answer questions that required the data
that was previously split, the ontology successfully met the first evaluation criterion.
The second evaluation criterion measures the capacity of the ontological model to
classify the patient’s trajectory as “good” or “bad”, and the patient risk of complications
during the bariatric surgery. In order to do that, the data will be aleatorily split in two
parts, with 80% and 20% of the instances (this ratio can be changed depending on the
number of instances we will have). Using that second portion of the data, the health
history of the patients will be manually analyzed with help from health professionals,
classifying the patient’s trajectory as “good” or “bad” and classifying the risk of the
patient. This manual evaluation will be compared to the automatic evaluation done by
the ontology and the sensitivity and the specificity of the model will be calculated. The
closer the sensitivity and specificity are to 100%, the better is the result. It is important
to note that the ontology building steps are iterative, so there could be multiple
iterations of the steps above.
6. Expected results
It is expected that the developed ontology will allow for integrating the data from
patients who are candidate to bariatric surgery at UNIMED and highlight risks. The
ontology could be used to answer competency questions that the health professional
wants to answer about the data. The competency questions defined were defined by the
researcher based on the systematic review of the literature and they are listed below.
• What are the risk factors that patient X has to receive a bariatric surgery?
• What are the comorbid conditions of patient X?
However, these competency questions can be incremented or changed after
having access to the UNIMED data and starting developing the ontology, due to
iterative nature of the ontology development process. It is also expected that this
ontology could be easily expanded to include other domains in the medical field,
allowing for flexibility in the semantic representation of the data and avoiding changes
in the original structure of the data. All these advantages (and possible disadvantages)
will be known after the research execution.
7. Conclusion
This research proposes to investigate if this specific ontological model could solve data
integration challenges in the health domain and help health professionals in the decision
making. It is expected that the developed ontology, if successful, could be used as a
base for future application that can help the decision making in the health area and
improve the medical treatment of patients who are candidate to bariatric surgery.
References
Agência Nacional De Saúde (2020), “Dados Gerais”, https://www.ans.gov.br/perfil-do-
setor/dados-gerais.
Associação Brasileira para o Estudo da Obesidade e da Síndrome Metabólica (2016),
“Diretrizes Brasileiras De Obesidade”, AC Farmacêutica.
Bodenreider, O., Cornet, R. and Vreeman, D. J. (2018), “Recent developments in
clinical terminologies—SNOMED CT, LOINC, and RxNorm”, Yearbook of medical
informatics, v. 27, n. 01, p. 129-139.
Brasil (2017), “Vigitel Brasil 2016: Vigilância de Fatores de Risco e Proteção para
Doenças Crônicas por Inquérito Telefônico”, Ministério da Saúde, Secretaria de
Vigilância em Saúde, Brazil.
Chadwick, J. and Mann, W. N. (1950), “Medical Works of Hippocrates”, Blackwell
Scientific Publications, Boston, MA, P. 154.
Chen, L. et al (2019), “OMDP: An ontology-based model for diagnosis and treatment of
diabetes patients in remote healthcare systems”, International Journal of Distributed
Sensor Networks, V. 15, issue 5.
Cresswell, K. C. et al (2013), “Computerised Decision Support Systems for Healthcare
Professionals: An Interpretative Review”, Journal of Innovation in Health
Informatics 20, 2, 115–128.
El-Hachem, J. et al (2012), “An Owl 2-Based Knowledge Platform Combining The
Social and Semantic Webs For An Ambient Childhood Obesity Prevention System”,
Procedia Computer Science, [S.L.], V. 10, P. 110–119.
Kitchenham, B (2004), “Procedures for Performing Systematic Reviews”, Keele, UK,
Keele University, v. 33, n. 2004, p. 1–26.
Mccallie, D. P. (2016), “Clinical Decision Support: History and Basic Concepts”,
Healthcare Information Management Systems, Springer, 3–19.
Noy Nf, M. D. (2001), “Ontology Development 101: A Guide To Creating Your First
Ontology”. Stanford, CA, Stanford University.
Porter, R. S. and Kaplan, J. L. (2011). “The Merck manual of diagnosis and therapy”,
Section 1 - Nutritional Disorders, Chap. 5 - Obesity Topics, Merck Sharp & Dohme
Corp.
Sjostrom, L. (2008), “Bariatric Surgery and Reduction in Morbidity and Mortality:
Experiences from the SOS Study”, International Journal of Obesity 32, S93-S97.
UNIMED (2020), “UNIMED – Institucional”,
https://www.unimed.coop.br/home/sistema-unimed/a-unimed.
World Health Organization (2018), “Obesity and Overweight”.
https://www.who.int/news-room/fact-sheets/detail/obesity-and-overweight.