Formal Concept Analysis Applied to Characterize Longitudinal
Associations Between Depressive and Anxiety Disorders and
Somatization
Diogo Miranda
a
, Julio Neves
b
, Luiz Zarate
c
and Mark Song
d
Universidade Cat
´
olica de Minas Gerais, Computer Science Department, Minas Gerais, Brazil
Keywords:
Formal Concept Analysis, Triadic Concept Analysis, Bierdermann Conditional Attribute Association Rule
(BCAAR), Bierdermann Attributional Condition Association Rule (BACAR).
Abstract:
This study examines somatic syndromes as a significant public health challenge, highlighting the necessity
of longitudinal sampling to comprehend the evolution of physical symptoms over time. It investigates the
interplay between depressive and anxious symptoms and somatic symptoms related to disease. The research
characterizes these symptoms within a diverse population in Isfahan, Iran, over a three-year period, utilizing
Triadic Concept Analysis (TCA) as the primary analytical method to extract insights and establish correlations
across time. The findings emphasize the importance of longitudinal methodologies in exploring patterns and
rules associated with the symptoms under investigation. These insights enhance the understanding of the
relationship between mental and physical health, offering valuable insights for clinical decision-making and
treatment strategies.
1 INTRODUCTION
Somatic syndromes pose a significant public health
issue, affecting both individuals and healthcare sys-
tems. A somatic symptom is diagnosed when a person
significantly focuses on experiencing physical symp-
toms such as pain, weakness, shortness of breath, pal-
pitations, fatigue, and organ pain.
These abnormal sensations may be classified as
somatic syndromes when they cause substantial dis-
tress or functional impairment, without being linked
to a primary clinical condition.
Physical symptoms may or may not be associated
with a medical diagnosis, yet an individual experienc-
ing these symptoms believes they are unwell. Re-
search conducted in various countries over the past
few decades, primarily focusing on primary care set-
tings, indicates that somatization occurs in 16% to
50% of patient encounters(Schreiber et al., 2007)
(Spitzer et al., 2004) (WHO, 1997).
Identifying when a somatic symptom is directly
linked to a medical condition is a complex and nu-
a
https://orcid.org/0009-0009-4868-5700
b
https://orcid.org/0000-0002-0520-9976
c
https://orcid.org/0000-0002-3018-888X
d
https://orcid.org/0000-0001-5053-5490
anced task. Studies indicate that an individual ex-
periences an abnormal symptom approximately every
seven days (Birket-Smith and Mortensen, 2010), and
most often does not seek emergency care, as these
symptoms tend to resolve after a few days.
It is frequently necessary to observe patients over
several years to understand the abnormal symptoms
associated with physical conditions, highlighting the
importance of longitudinal sampling — often con-
ducted annually — within the context in which a pa-
tient is situated.
A well-known fact is that sadness and distress,
common human emotions, are always accompanied
by physical components. Similarly, persistent men-
tal disorders—such as depression and anxiety—are
associated with various physical symptoms, indicat-
ing that in 50% of cases, there is a direct relationship
between depression, anxiety, and somatization (L
¨
owe
et al., 2008).
Based on this principle, logistic regression analy-
ses were employed in (Bekhuis et al., 2015), revealing
results indicating that all groups of somatization were
more prevalent among patients with depressive and/or
anxiety disorders, as explained further in this article.
Thus, it becomes feasible to investigate and analyze
depressive and anxious symptoms primarily in rela-
390
Miranda, D., Neves, J., Zarate, L. and Song, M.
Formal Concept Analysis Applied to Characterize Longitudinal Associations Between Depressive and Anxiety Disorders and Somatization.
DOI: 10.5220/0013106500003911
Paper published under CC license (CC BY-NC-ND 4.0)
In Proceedings of the 18th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2025) - Volume 2: HEALTHINF, pages 390-397
ISBN: 978-989-758-731-3; ISSN: 2184-4305
Proceedings Copyright © 2025 by SCITEPRESS – Science and Technology Publications, Lda.

tion to somatization.
In this article, depressive symptoms, anxious
symptoms, and somatization were characterized us-
ing Triadic Concept Analysis (TCA) based on a mul-
tidimensional and longitudinal data repository (Adibi
et al., 2023) from the Isfahan region, Iran - Isfahan
is the third most populous city in the country, with
a population exceeding 1.8 million residents. This
setting was chosen primarily due to the diversity of
its population and the observable socio-cultural influ-
ences.
The data encompass various factors contributing
to somatization, involving 1,930 participants, primar-
ily adults aged between eighteen and sixty-five years,
who were randomly selected and monitored from
2017 to 2020. These factors include:
(a) functional symptoms across multiple body or-
gans,
(b) psychological assessments,
(c) lifestyle factors,
(d) demographic and socioeconomic aspects,
(e) laboratory studies,
(f) clinical analyses, and
(g) clinical history.
The aim is to primarily characterize psychological
phenomena that are independently related to somatic
symptoms through this longitudinal sampling.
Formal Concept Analysis (FCA) was introduced
in 1982 by Rudolf Wille (Dau and Wille, 2000) as a
derivation of concepts that possess hierarchies based
on a set of objects and properties. One of the primary
objectives of this method is to extract knowledge and
characterize a database, thereby enhancing the under-
standing of a context and facilitating better decision-
making in data analysis - the Triadic Concept Analy-
sis (TCA) is an extension of FCA, proposed by Wille
in 1995, which incorporates a third component in the
data analysis process.
The remainder of the paper is organized as fol-
lows: the background is outlined in Section 2. The
Related Work is described in Section 3. Section 4
presents the Methodology. Results are discussed in
Section 5. The conclusion and further research are in
Section 6.
2 BACKGROUND
Longitudinal study approaches aim to investigate a
sample of individuals with specific characteristics
over consecutive time periods, referred to as waves.
This methodology is widely utilized in the field of
health, as it provides flexibility in analyzing treatment
periods with medications and methodologies.
Longitudinal studies in the health field can en-
compass various characteristics and samples, referred
to as the dimensionality of the database. Databases
can have longitudinal characteristics, allowing for the
tracking and characterization of behaviors and pat-
terns of the objects and attributes involved in the
study. Therefore, a longitudinal study of such a
database can yield valuable insights to validate the
clinical procedures being adopted.
Studies may be conducted with thousands of sam-
ples and characteristics to assess, for example, how
a pre-existing symptom in a patient may affect an-
other clinical condition, representing a study based on
a high-dimensionality database. Conversely, studies
involving treatments that require specific characteris-
tics of a particular population may consist of only a
few hundred samples, indicating low or medium di-
mensionality.
Low dimensionality can lead to reduced learning
rates and knowledge extraction when utilizing artifi-
cial intelligence methods, such as machine learning.
Conversely, longitudinal studies are highly effec-
tive for detecting patterns and rules associated with
the study group. Overall, when longitudinal tech-
niques are applied to the analysis of a database, it be-
comes possible to extract important data, such as:
1. The evolution of clinical symptoms over years of
patient sampling;
2. The relationship between characteristics and
symptomatic variables;
3. The direct relationship between two clinically an-
alyzed symptoms considered in isolation.
All of this information can be of great value
for clinical analysis, providing support for decision-
making.
The database (Adibi et al., 2023) analyzed in this
study possesses longitudinal characteristics, as it in-
volves the observation of objects (patients) and at-
tributes (somatic symptoms) over the periods from
2017 to 2020. This work focused on the first three
years of patient sampling to maintain the highest pos-
sible dimensionality of the database, thereby avoid-
ing excessive sparse data—specifically, patients who
dropped out of the study from one year to the next.
Consequently, a total of 1,616 patients were analyzed
during the years 2017, 2018, and 2019, representing
approximately 83% of the original sample.
This database is well-suited for the application of
longitudinal studies, as it allows for the measurement
of the effects of symptoms related to depression and
anxiety on somatization across different periods, as
Formal Concept Analysis Applied to Characterize Longitudinal Associations Between Depressive and Anxiety Disorders and Somatization
391

well as the evolution and potential persistence of these
symptoms.
2.1 Formal Concept Analysis
Formal Concept Analysis (FCA) can be used to rec-
ognize patterns with the help of association rules and
their implications. FCA consists of a set of objects
forming a formal context, formal concepts, and rules.
A formal context can be represented as a triple
K = (G, M, I): G is a set of objects, M is a set of
attributes, and I ⊆ G× M an incidence relation where
(g, m) ∈ I indicating that the object g possesses the
attribute m.
The association rule A → B is valid only if for
every object that contains the attributes of B, it also
contains the attributes of A. Given a rule r, and the
parameters s and c, it can be denoted:
s = suppr (r) =
|
A
′
∩B
′
|
|
G
′
|
and
c = con f (r) =
|
A
′
∩B
′
|
|
A
′
|
S referred to as the support of the rule and C re-
ferred to as the confidence. When con f (r) = 100%,
the rule is referred to as an implication (Felde and
Stumme, 2023) .
2.2 Triadic Concept Analysis
The definitions of FCA extend to a third dimension,
resulting in a triadic context T, denoted by the quadru-
ple T = (G, M, B, Y ), where G, M, and B are, respec-
tively, the set of objects, attributes, and conditions, be-
longing to the ternary relation Y ⊆ (G × M × B). This
relation can be interpreted as: object g possesses at-
tribute m under condition b.
From the triadic context, it is possible to extract
triadic concept which is defined as (A, B, C). As in
dyadic concepts, A ⊆ K
1
(extent) and B ⊆ K
2
(Intent)
exist, however, there is also the addition of a third
component, C ⊆ K
3
(modus). From this concept, it is
possible to extract triadic rules, that are in the scope
of this paper.
The introduction of a third dimension allows for
a better characterization and representation of data.
Bi-dimensional data may receive the dimension time
to monitor the evolution of objects in relation to
attributes and discovery of hidden patterns in the
database, and this is exactly the shape of longitudinal
studies, such as the database used in this study.
Two types of triadic association rules can be de-
scribed for a context K := (G, M, B, Y ) (Biedermann,
1997a): the Bierdermann Conditional Attribute Asso-
ciation Rule (BCAAR) and the Bierdermann Attribu-
tional Condition Association Rule (BACAR).
The BCAAR rule is represented by:
(R → S)C (sup., con f .), where R, S ⊆ M and
C ⊆ B. This means that for every object possessing
all attributes of R, it also possesses all attributes
of S under condition C, with a support (sup) and a
confidence (conf).
The BACAR rule is represented by:
(P → Q) N (sup, con f ), where P, Q ⊆ B and N ⊆ M.
This means that for every object under the condition
in P, it will also be under the conditions in Q in
attribute N, with a support (sup) and a confidence
(conf).
The support corresponds to the proportion of ob-
jects in the subset g ⊆ G that satisfy the implication
P → Q, relative to the total number of objects |G| in
the formal context K:
Sup (P → Q) =
|(P ∪
{
Q
}
)|
|G|
Confidence corresponds to the ratio of objectsg ⊆
G that contain P and also contain Q, compared to the
total number of objects |G|:
Con f (P → Q) =
|P ∪
{
Q
}
|
|P
′
|
=
Sup (P → Q)
Sup (P)
2.3 Somatic Syndromes
The phenomenon known as ”somatization” can be
better defined by (Lipowski, 1988) as follows: ”Som-
atization is defined here as a tendency to experience
and communicate somatic discomfort and symptoms
that cannot be explained by pathological findings, at-
tributing them to physical illnesses and seeking med-
ical help for them.”
Analyzing somatic physical symptoms in a patient
is always a significant challenge for healthcare profes-
sionals, as many times a somatic symptom does not
directly reflect a clinical condition but rather an emo-
tional state of the patient.
One subfield of study on somatic symptoms is
psychosomatic disorders, where psychological dis-
tress can, in some way, cause or exacerbate a physical
symptom. This psychic suffering is often involuntary
and unconscious. On the other hand, (Bekhuis et al.,
2015) shows that somatization is greatly influenced
by other psychological clinical conditions, with sig-
nificant connections to depressive and anxious symp-
toms.
The characterization of these symptoms and clini-
cal conditions, combined with a professional analysis
of a patient’s history, can help identify a cause for a
HEALTHINF 2025 - 18th International Conference on Health Informatics
392

somatic condition. Thus, from a longitudinal sam-
ple of patients, it is possible to characterize, through
TCA, a database containing relevant information on
depressive symptoms, anxious symptoms, and soma-
tization.
This can be obtained from implication rules of the
form (P → Q) N (sup, con f ), meaning, for example,
that an anxious/depressive condition p implies a so-
matic condition q over a specific period n with a sup-
port (sup) and a confidence (conf), contributing to the
analysis of a clinical condition.
3 RELATED WORK
This work utilizes Formal Concept Analysis, with the
approach justified through a relationship between the-
oretical and practical knowledge of the subject. Re-
lated Work on this topic are presented below.
(Wei et al., 2018) analyzes the triadic approach of
formal concept analysis in four aspects: (i) the basic
approach of triadic concept analysis, (ii) triadic im-
plications and rules, (iii) the triadic factor of analysis,
and (iv) the analysis of fuzzy triadic concepts.
(Biedermann, 1997b) systematically Illustrates
the application of triadic formal concept analysis in
databases to represent complex concepts that are dif-
ficult to visualize. It also explains the generation
of rules and implications from an analyzed dataset.
In (Blevente Lorand Kis and Troanca, 2017), a tool
is presented that enables the visualization of these
concepts and rules, facilitating navigation and under-
standing of triadic structures .
The work presented in (Ganter and Obiedkov,
2004) shows several possible biases that can be gener-
ated from triadic formal concept analysis and various
implications in multiple scenarios.
Given the different interests that can be addressed
from a triadic context, the authors provide exten-
sive and compact descriptions through implication-
generating algorithms in the triadic context.
Examples of these interests, which can span dif-
ferent areas yet still be resolved in a triadic context,
are found in (Kent and Neuss, 1997), where the focus
is on hypertext analysis, and in (Carullo et al., 2015),
which presents an application of this method in online
recommendation systems.
(Hu et al., 2004) presents modeling techniques
based on a logical description language applied to
a cancer database. Results generated from the in-
tentions and extensions of entities present in these
databases are provided, obtained through formal con-
cept analysis.
In (Santos et al., 2022), an analysis of infant mor-
tality in two regions of Minas Gerais is conducted.
The process utilized the triadic formal concept analy-
sis approach to extract rules and implications from a
database. The study generated a series of rules with
certain hierarchies that characterized this database.
In (Ferreira et al., 2021), an application is made to
extract knowledge from a database generated by a sur-
vey conducted with women undergoing chemother-
apy for breast cancer. The application of Formal Con-
cept Analysis theory allowed for the extraction of a
set of hierarchically organized concepts, and from
these, rules that relate them were derived, thus de-
scribing the results of the antiemetic treatments in this
database.
(Lana et al., 2022) conducts a longitudinal anal-
ysis of a database on COVID-19 using the processes
of triadic formal concept analysis. The results of this
work present implication rules that longitudinally de-
scribe the evolution of the COVID-19 pandemic at
different points in time. The literature shows that
FCA can be applied in many different contexts ((Ana-
nias et al., 2021), (Alves et al., 2023)).
4 METHODOLOGY
The database discussed in this article (Adibi et al.,
2023) consists of a multidisciplinary longitudinal
sampling of somatic symptoms, primarily represented
by adults aged eighteen to sixty-five from Isfahan,
Iran, who were randomly selected and monitored
over four consecutive years. In this study, seven
databases were collected:
(a) assessment of functional symptoms in various
organs,
(b) psychological assessment,
(c) lifestyle variables,
(d) demographic and socioeconomic variables,
(e) laboratory measurements,
(f) clinical examination, and
(g) historical information.
Psychological details, including depression, anxi-
ety, post-traumatic stress, and other attributes related
to psychosomatic symptoms, were also included in
this database. Additionally, this study addressed the
issue of post-traumatic stress focused on COVID-19
through questionnaires. The questionnaires offered a
choice between 1 and 4, where 1 indicated the pres-
ence of a symptom and 4 indicated its absence.
The study was initially conducted with 1,943 pa-
tients in 2017, and by 2020, it was completed with
1,176 patients who remained. This particular article
Formal Concept Analysis Applied to Characterize Longitudinal Associations Between Depressive and Anxiety Disorders and Somatization
393

analyzes the periods from 2017 to 2019, which had
the highest patient participation rate, with 1,697 par-
ticipants and a response rate of 88%, comprising an
average age of 40.03 years, with 756 being male.
Figure 1: Methodology.
There are many articles that address formal con-
cept analysis applied to the field of health and human
behavior. This particular article seeks to characterize
depressive symptoms, anxious symptoms, and soma-
tization based on a multidimensional and longitudinal
data repository using: data collection, exploration, at-
tribute selection and transformation, and the extrac-
tion of contexts and rules (Figure 1).
In the first stage, it was necessary to collect the re-
quired data and its description. For this, access to the
objects and attributes to be worked on was requested.
Due to legal restrictions, the data cannot be processed
outside the servers of the Isfahan Cardiovascular Re-
search Center. Therefore, as shown in Figure 1, it
became necessary to install systems and analyze the
data through a remote connection to the servers.
After the installation of the necessary systems,
data preprocessing is required in the second stage. For
this, the selection of attributes to be the focus of this
article was made. The choice was primarily based on
the aim of conducting a direct analysis of the effects
of anxiety and depression on somatization.
The attributes related to depression, anxiety, and
somatic symptoms — described in Table 1 — were
selected according to their relevance to the study. In
Table 1, it is possible to observe the division into
groups that separates depressive and anxious symp-
toms (H) from somatic symptoms (S).
Additionally, the data was cleaned to ensure the
maximum number of participating patients. Due
to limitations in the computational resources of the
server and the computational cost of the algorithms
used in TCA, it became necessary to divide the origi-
nal database in the third stage into several subsets.
This division was based on the product of an at-
tribute related to depression and anxiety with all the
attributes characterizing a somatic symptom. This ap-
proach would allow for a longitudinal characteriza-
tion of the relationship between depression and anxi-
ety attributes and those of somatization.
Table 1: Symptoms of Depression and Anxiet (H), Somatic
(S).
Group Sub Group Description
H A Sleep problems
H B Panic symptoms
H C Sadness symptoms
H D Anxiety symptoms
H E Sudden palpitations
H F Existential problems
H G Sudden restlessness
H H Irritation
H I Constant worry
Group Sub Group Description
S A Headaches
S B Stomach aches
S C Back pain
S D Joint pain
S E Chest pain
S F Swelling
S G Restlessness
S H Heart palpitations
S I Pressure in the heart
In the fourth stage, a transformation of the data
was performed using discretization techniques. This
discretization considered the range of the questions,
initially defining [1..2] → no and [3..4] → yes. This
discretization was necessary for the algorithm to work
correctly.
In the fifth stage, the creation of the triadic formal
context was carried out, as shown in Tables 2 and 3,
along with the input data for the Lattice Miner soft-
ware mentioned throughout the work.
For the creation of the triadic context, each ob-
ject is described by three waves: 2017 (w1), 2018
(w2) and 2019 (w3). The example of the formal
context also illustrates the division of the database
according to the mathematical relationship a × S =
{
(a, b)|a ∈ H, b ∈ S
}
.
The tool used for the FCA algorithms is Lattice
Miner 2.0. This is a data mining prototype developed
under the supervision of Professor Rokia Missaoui at
the University of Quebec.
It is a public access Java platform whose main
functions include all low-level operations that allow
for the manipulation of input data, structures, and rule
associations.
HEALTHINF 2025 - 18th International Conference on Health Informatics
394

Table 2: Triadic Context of the Base Sleep Problems (HA).
w1 w2 w3
Obj HA SA SB ... SI HA SA SB ... SI HA SA SB ... SI
1 x x x x x x x
2 x x x x x x x x x
3 x x x x
Table 3: Triadic Context of the Base Panic Symptoms (HB).
w1 w2 w3
Obj HB SA SB ... SI HB SA SB ... SI HB SA SB ... SI
1 x x x x x x x x x
2 x x x x x x x x
3 x x x x x x x
The platform enables the generation of groups,
called formal concepts, which include logical impli-
cations, thus showing binary relationships between
collections of objects and their sets of attributes or
properties.
5 RESULTS
Based on a triadic analysis conducted on the database
in question, it was possible to relate how aspects
of depression and anxiety interfere with somatic
symptoms in the population of Isfahan, Iran.
Considering:
• w1: Data from the first wave (1) conducted in
2017;
• w2: Data from the second wave (2) conducted in
2018;
• w3: Data from the third wave (3) conducted in
2019.
The following rules were obtained:
BACAR Implications.
1. (HA → SG) w1 [sup = 62, 7%, conf = 93, 3%]
2. (HA → SG) w2 [sup = 64, 3%, conf = 92, 3%]
3. (HA → SG) w3 [sup = 62, 7%, conf = 93, 5%]
4. (HB → SE)w1[sup = 70, 8%, conf = 90, 4%]
5. (HB → SG) w3 [sup = 68, 2%, conf = 92, 9%]
6. (HC → SG) w3 [sup = 70, 5%, conf = 93, 7%]
7. (HC → SG) w2 [sup = 67, 9%, conf = 91, 3%]
8. (HD → SG) w2 [sup = 81, 4%, conf = 91, 1%]
9. (HD → SG) w3 [sup = 73, 5%, conf = 91, 4%]
10. (HE → SF)w2[sup = 70, 5%, conf = 90, 3%]
11. (HF → SG)w2 [sup = 71, 0%, conf = 91, 0%]
12. (HF → SF)w2[sup = 71, 7%, conf = 91, 8%]
The attributes in the presented rules follow the
definitions from Table 1, as well as the indicated
waves. A support threshold of 60% was used for gen-
erating the rules. Several results were generated; for
the interpretation of the rules in this article, only the
main rules with a confidence level above 90% were
considered.
For rules 1, 2, and 3, an interesting relationship
can be found between the symptom of sleep problems
(HA) and the somatic symptom of restlessness (SG).
In the three indicated waves for the periods of
2017, 2018, and 2019 (w1, w2, and w3), this relation-
ship remained consistent with a support close to 63%
and a confidence of 93%. This means that a portion
of the database characterized by symptoms of sleep
problems also exhibits symptoms of headaches.
On the other hand, rules 6 and 7 illustrate a clinical
condition that emerged during the periods of 2018 and
2019 (w2 and w3).
These rules indicate that symptoms related to sad-
ness and loss of motivation (HC) may be associ-
ated with somatic symptoms of restlessness (SG) and
sudden loss of attention. This characteristic of the
database also supports the context in which patients
may have found themselves, such as during a global
crisis or a pandemic.
It can also be inferred that the characteristics of
the database lead us to understand that the somatic
symptoms most directly related to depressive and
anxious symptoms are the clinical conditions of rest-
lessness (SG), sudden sensations of swelling (SF),
and chest pain (SE), particularly during the periods
of 2018 (w2) and 2019 (w3).
Formal Concept Analysis Applied to Characterize Longitudinal Associations Between Depressive and Anxiety Disorders and Somatization
395

Conversely, other depressive and anxious symp-
toms did not appear in the analysis when considering
the established support and confidence thresholds;
for example, regarding the symptom of constant and
sudden worry (HI), no rules were found that charac-
terize the database according to the set parameters.
BCAAR Implications
1. (w2 → w3)SE[sup = 72, 6%, conf = 92, 5%]
2. (w3 → w2)SG[sup = 75%, conf = 92, 8%]
3. (w1 → w3)HB[sup = 62, 9%, conf = 90, 3%]
4. (w1 → w3)SE[sup = 72, 9%, conf = 91, 8%]
5. (w1 → w3)SI[sup = 70, 6%, conf = 94, 2%]
6. (w2w1 → w1)SG[sup = 71, 6%, conf = 95, 7%]
7. (w1w2 → w3)SI[sup = 68, 6%, conf = 94, 0%]
8. (w1w2 → w3)HC[sup = 63, 0%, conf = 90, 9%]
9. (w1w3 → w2)HC[sup = 73, 5%, conf = 93, 6%]
10. (w2 → w1)HD[sup = 81, 4%, conf = 91, 1%]
11. (w2 → w1)SG[sup = 83, 5%, conf = 93, 6%]
12. (w2w3 → w1)SH[sup = 60, 5%, conf = 90, 1%]
Applying the BACAR rules similarly, no rules
were generated with support lower than 60%. Like-
wise, the interpreted rules must have support greater
than 50% and confidence of 90%.
Rule 1 shows that 72.6% of the patients who par-
ticipated in the study had chest pain (SE) in 2018
(w2), and 92.5% continued to have chest pain in 2019
(w3).
It can also be observed for Rule 3 that 62.9% of
the patients who had panic symptoms in 2017 (w1)
continued to have this symptom in 2019 (w3), main-
taining the same behavior for the somatic symptoms
in Rules 4 and 5, with support around 70% and confi-
dence around 92%.
Rule 7 shows that for every patient who partici-
pated in the study during the observation period be-
tween 2017 and 2018 (w1, w2), 94% continued to ex-
hibit a somatic symptom of heart pressure (SI) in the
following year (w3).
Similarly, Rule 8 shows the same behavior for a
depressive and anxious symptom. Between 2017 and
2018 (w1, w2), 73.5% of the patients had the symp-
tom of sudden sadness, and in 2019 (w3), 93.6% of
the individuals continued to exhibit the same symp-
tom.
Another interpretation we can derive is that
BACAR rules 6, 7, 8, 9, and 10 allow us to identify
a certain prevalence of both depressive and anxious
symptoms as well as somatic symptoms during the
three waves (w1, w2, and w3). This may indicate cer-
tain characteristics that remain persistent over a long
observation period, potentially supporting clinical de-
cisions during the analysis of these patients.
Thus, by analyzing these implications, it is pos-
sible to generate practical results, especially as assis-
tance in decision-making during the evaluation of a
medical condition.
6 CONCLUSIONS
The objective of this article, following the analysis
and selection of results, is to understand and primar-
ily characterize the relationship between somatic syn-
dromes and symptoms of depression and anxiety.
Additionally, the research seeks to highlight the
contrasts in the results obtained within the context of
the population of Isfahan, Iran, using a database main-
tained and provided by the Isfahan Cardiovascular
Research Institute (ICRI).
This database provided information on a study
conducted with patients who experienced clinical
conditions between 2017 and 2020, also considering
the developments of the COVID-19 pandemic during
this period.
The information was collected and validated by
professionals through questionnaires and made avail-
able in the Data in Brief repository. Furthermore, it
is important to emphasize that the database contains
information on various variables that may influence
a somatic symptom, making it a non-linear analysis
rather than a simple one.
Conversely, the objective of this work was to en-
hance the applicability of Triadic Concept Analysis.
This analysis has proven to be an effective approach
for identifying aspects in certain variables that are not
easily discernible in a primary analysis, related to var-
ious contexts across fields of knowledge.
With this result, it is possible to support a profes-
sional’s analysis and be useful in the foundation and
justification of decision-making. However, the study
has the limitation of not considering all aspects and
contexts that may correspond and relate to a somatic
condition.
Therefore, by conducting a complex analysis of
indirectly related data, the study reveals a correspon-
dence and a relationship between these elements, pro-
viding an alternative perspective on the database
Thus, it is expected that various fields of study and
scenarios will benefit from the methodology used in
this work during the investigation and analysis stages,
utilizing Triadic Concept Analysis (TCA).
HEALTHINF 2025 - 18th International Conference on Health Informatics
396

ACKNOWLEDGEMENTS
The authors thank the Pontif
´
ıcia Universidade
Cat
´
olica de Minas Gerais – PUC-Minas and
Coordenac¸
˜
ao de Aperfeic¸oamento de Pessoal de
N
´
ıvel Superior — CAPES (CAPES – Grant PROAP
88887.842889/2023-00 – PUC/MG, Grant PDPG
88887.708960/2022-00 – PUC/MG - Inform
´
atica,
and Finance Code 001). The present work was also
carried out with the support of Fundac¸
˜
ao de Amparo
`
a Pesquisa do Estado de Minas Gerais (FAPEMIG)
under grant number APQ-01929-22.
REFERENCES
Adibi, P., Kalani, S., Ani, A., Shahoon, H., Feizi, A., and
Roohafza, H. (2023). A multidimensional longitudi-
nal dataset on functional somatic syndromes. Data in
Brief, 48:109267.
Alves, A., Zarate, L., Freitas, H., and Song, M. (2023). Par-
allelism in the generation of concepts through the for-
mal context object partitioning using the in-close 4 al-
gorithm. pages 195–202.
Ananias, K. H., Missaoui, R., Ruas, P. H., Zarate, L. E., and
Song, M. A. (2021). Triadic concept approximation.
Information Sciences, 572:126–146.
Bekhuis, E., Boschloo, L., Rosmalen, J. G., and Schoev-
ers, R. A. (2015). Differential associations of specific
depressive and anxiety disorders with somatic symp-
toms. Journal of Psychosomatic Research, 78(2):116–
122.
Biedermann, K. (1997a). How triadic diagrams represent
conceptual structures. In Lukose, D., Delugach, H.,
Keeler, M., Searle, L., and Sowa, J., editors, Con-
ceptual Structures: Fulfilling Peirce’s Dream. ICCS
1997, volume 1257 of Lecture Notes in Computer Sci-
ence, Berlin, Heidelberg. Springer.
Biedermann, K. (1997b). How triadic diagrams represent
conceptual structures. In Lukose, D., Delugach, H.,
Keeler, M., Searle, L., and Sowa, J., editors, Concep-
tual Structures: Fulfilling Peirce’s Dream, pages 304–
317, Berlin, Heidelberg. Springer Berlin Heidelberg.
Birket-Smith, M. and Mortensen, E. L. (2010). Pain in
somatoform disorders: Is somatoform pain disorder
a valid diagnosis? Acta Psychiatrica Scandinavica,
106(2):103–108.
Blevente Lorand Kis, C. S. and Troanca, D. (2017). Fca
tools bundle – a tool that enables dyadic and tri-
adic conceptual navigation. In Proceedings of the In-
ternational Conference on Formal Concept Analysis,
ICFCA ’17, pages 214–219.
Carullo, G., Castiglione, A., De Santis, A., and et al. (2015).
A triadic closure and homophily-based recommenda-
tion system for online social networks. World Wide
Web, 18(6):1579–1601.
Dau, F. and Wille, R. (2000). On the modal understanding
of triadic contexts. In Decker, R. and Gaul, W., edi-
tors, Classification and Information Processing at the
Turn of the Millennium, pages 83–94, Berlin, Heidel-
berg. Springer Berlin Heidelberg.
Felde, M. and Stumme, G. (2023). Triadic exploration and
exploration with multiple experts. Knowledge & Data
Engineering Group, University of Kassel, Germany.
Ferreira, L., Nobre, C., Z
´
arate, L., and Song, M. (2021).
Study of the evolution of antiemetic treatment through
the application of triadic formal concept analysis. KD-
MILE.
Ganter, B. and Obiedkov, S. (2004). Implications in tri-
adic formal contexts. In Wolff, K. E., Pfeiffer, H. D.,
and Delugach, H. S., editors, Conceptual Structures
at Work, pages 186–195, Berlin, Heidelberg. Springer
Berlin Heidelberg.
Hu, B., Dasmahapatra, S., Dupplaw, D., Lewis, P., and
Shadbolt, N. (2004). Managing patient record in-
stances using dl-enabled formal concept analysis. In
Motta, E., Shadbolt, N. R., Stutt, A., and Gibbins,
N., editors, Engineering Knowledge in the Age of the
Semantic Web, pages 172–186, Berlin, Heidelberg.
Springer Berlin Heidelberg.
Kent, R. E. and Neuss, C. (1997). Conceptual analysis of
hypertext, pages 70–89. Springer Berlin Heidelberg,
Berlin, Heidelberg.
Lana, P., Nobre, C., Zarate, L., and Song, M. (2022). For-
mal concept analysis applied to a longitudinal study
of covid-19. Scitepress.
Lipowski, Z. J. (1988). Somatization: the concept and its
clinical application. American Journal of Psychiatry,
145(11):1358–1368.
L
¨
owe, B., Spitzer, R. L., Williams, J. B., Mussell, M.,
Schellberg, D., and Kroenke, K. (2008). Depression,
anxiety and somatization in primary care: Syndrome
overlap and functional impairment. General Hospital
Psychiatry, 30(3):191–199.
Santos, D., Nobre, C., Zarate, L., and Song, M. (2022). Ap-
plication of formal concept analysis and data mining
to characterize infant mortality in two regions of the
state of minas gerais. ICEIS 2022 - 24th International
Conference on Enterprise Information Systems.
Schreiber, D., Kolb, N. R., and Tabas, G. (2007). Soma-
tizing patients: part i. practical diagnosis. American
Family Physician, 61:1073–8.
Spitzer, R. L., Williams, J. B., Kroenke, K., Linzer, M.,
Hahn, S. R., and Brody, D. (2004). Utility of a new
procedure for diagnosing mental disorders in primary
care. Journal of the American Medical Association,
272:1749–1762.
Wei, L., Qian, T., Wan, Q., and et al. (2018). A research
summary about triadic concept analysis. Interna-
tional Journal of Machine Learning and Cybernetics,
9(4):699–712.
WHO, W. H. O. (1997). Somatization in cross-cultural per-
spective: A world health organization study in pri-
mary care. American Journal of Psychiatry, 154:989–
995.
Formal Concept Analysis Applied to Characterize Longitudinal Associations Between Depressive and Anxiety Disorders and Somatization
397