Spatial-Temporal Visualization Tool for Hospital Support for Infection

Spread and Outbreaks

Denisse Kim

1 a

, Manuel Campos

1,2 b

, Bernardo Canovas-Segura

1 c

and Jose M. Juarez

1 d

Med AI Lab, University of Murcia, Campus Espinardo, Murcia, 30100, Spain

Murcian Bio-Health Institute (IMIB-Arrixaca), El Palmar, Murcia, 30120, Spain

Keywords:

Visualization Application, Infection Control, Outbreak, Health, Hospital-Acquired Infection.

Abstract:

Hospital-acquired infections (HAIs) are a major concern today, especially when related to multidrug-resistant

bacteria, as they are associated with increases in healthcare costs, prolonged length of stay, and attributable

mortality. Tracking the presence of these infections requires interweaving spatial-temporal information from

patients and microbiological laboratory results. However, this is normally a manual process and the big

amounts of daily clinical data makes it error-prone and time-consuming. In these processes, the temporal

dimension is usually taken into account, but not the topology and spatial distribution of patients within a hos-

pital building. Interactive Information Visualization can be used to bring together information from various

data sources and to make these spatial-temporal relationships understandable to the human eye. We propose a

new interactive visual tool for the exploration of infection spreads within hospitals. The tool presents several

connected views to help analyze the epidemic situation of a hospital over time and understand the information

contained in the epidemiological indicators.

1 INTRODUCTION

Multidrug-resistant microorganisms are a growing

challenge for public health since their treatment is

a complex process. These types of microorganisms

have also a high impact on hospital-acquired infec-

tions (HAIs) (Cassini et al., 2019), as infections

caused by these pathogens are associated with in-

creases in healthcare costs, prolonged length of stay

and attributable mortality (Serra-Burriel et al., 2020).

In a healthcare setting, pathogens can be transmit-

ted from contact with an infected patient, a health-

care worker, or a contaminated environment (Mone-

gro et al., 2023). Tracking the presence of these in-

fections requires interweaving spatial-temporal infor-

mation about the patients and microbiological labora-

tory results. However, the large amounts of clinical

data that are captured daily pose some challenges and

make the process error-prone and time-consuming to

clinicians in the task of combining this information

to make any inference, and to hospital administrators

https://orcid.org/0000-0002-1790-372X

https://orcid.org/0000-0002-5233-3769

https://orcid.org/0000-0002-0777-0441

https://orcid.org/0000-0003-1776-1992

in their decision-making processes (Caban and Gotz,

2015).

The development of interactive Information Visu-

alization can help overcome this information overload

(Rind et al., 2010) and discover new knowledge, such

as patterns in the clustering of the pathogen and the

transmission among patients. However, current ap-

proaches tend to study diseases at the population level

in geographic areas, rather than local spatial-temporal

studies at the building level.

We introduce a new interactive visual tool for the

exploration of infection spreads inside hospitals. The

tool presents several connected views to help analyze

the epidemic situation of a hospital through time and

understand the information contained in epidemiolog-

ical indicators. In this paper, we describe the tasks

and requirements deﬁned with experts in epidemiol-

ogy and hospital management, and we present the

views developed to perform these tasks as well as the

interactions between them and with the user.

2 RELATED WORK

The development of visualization tools for epidemi-

ologists and policy-makers has focused mainly on

308

Kim, D., Campos, M., Canovas-Segura, B. and Juarez, J.

Spatial-Temporal Visualization Tool for Hospital Support for Infection Spread and Outbreaks.

DOI: 10.5220/0012312600003657

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 308-314

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

the task of disease surveillance over a population.

In these developments, the spatial-temporal informa-

tion of a disease normally describes the evolution and

spread at the population level for a post-analysis of

what happened usually in a geographic area (Chori-

anopoulos and Talvis, 2016; Sakai et al., 2004; Su

et al., 2021). However, despite the increasing ac-

knowledgment of the importance of tracing people

for the study of the spread of infections in recent

years, very little has been done on the analysis of con-

tacts and spread of diseases at the individual level and

taking into account the spatial characteristics of the

building they are in over time (Oppermann and Mun-

zner, 2020), especially in hospital environments.

Regarding the representation of the spatial topol-

ogy and data in buildings, Opperman and Munzner

(Oppermann and Munzner, 2020) presented a set of

visual decision-support tools centered around occu-

pancy data for management and planning in buildings.

However, their goal is to visualize non-trajectory

spatial-temporal data relating to large-scale indoor en-

vironments. Another example is the one presented by

onig et al. (K

onig et al., 2021), which is an interac-

tive visualization system that uses a 3D representation

of building interiors for a uniﬁed sensor data display

(e.g. temperature or humidity).

Regarding the development of visual analytics for

contact tracing and epidemiological studies, Baum-

gartl et al. (Baumgartl et al., 2021) presented a visual

analytics approach to support the analysis and recon-

struction of transmission pathways, patient contacts,

and the progression of an outbreak at the patient level.

They designed multiple specialized views and high-

lighted their adaptation of the storyline visualization

for contact tracing. Their views include an epidemic

curve, a contact network, a timeline, and a storyline-

like view. M

uller et al. (M

uller et al., 2020) applied

an RNN model for the detection of potential infec-

tions, transmissions, and infection factors, and they

proposed a visual interface to explore the model re-

sults. Sondag et al. (Sondag et al., 2022) presented

a visual analytics approach for the inspection of in-

fection maps in an evolving emergency response sit-

uation. They introduce the concept of representative

trees to visualize a time-varying infection map of dis-

ease spread, and interactive visualization techniques

for the assessment of different control policies.

3 METHODS

3.1 Data

To carry out the study of the spread of a disease and

the presence of infectious outbreaks in hospitalized

patients, it is necessary to intertwine microbiologi-

cal reports with the spatial and temporal localization

of patients. This consists of a complex and time-

consuming process that is not usually found in real

open datasets, both due to the challenge of recording

it and the privacy problems it can involve.

To assess the tool, we study the spread of the

Clostridioides Difﬁcile (CD) pathogen in a hospital

setting. The CD is the main cause of infectious diar-

rhea in hospitalized patients, with increasing rates of

mortality, incidence, and hospitalizations (Hota et al.,

2012). The choice of the best method for its treatment

continues to be a topic of debate today, while its inci-

dence increases, reaching values of 92 per 100,000 in-

habitants in North America and Europe (Lital Meyer

et al., 2014).

To represent the transmission of this pathogen at

a high level of detail, we use a realistic generated

dataset of patients with their information (i.e. demo-

graphic data, length of stay, admission day, and in-

formation about their treatment, among others). In

each moment, these patients are in a bed of a ser-

vice (this service can be a Ward, the ICU, the ER,

a Radiology Room or an Operating room) and have

a state of health. This state takes its values from the

SEIRD (Susceptible, Exposed, Infected, Recovered,

Dead) epidemiological model (Brauer, 2008), which

we adapted to the infection we are dealing with. In

this case, we included the NS (non-susceptible) state

for those that have immunity, and the possibility of

being already colonized in newly admitted patients.

We generated this dataset with a simulation model,

that combines an agent-based model with an adapta-

tion of the SEIRD epidemiological model(Kim et al.,

2023).

In our simulation, time is discrete, divided into

steps or periods. During each step, patients can move

from one room to another, their states of health can

change, and places can be infected or decontaminated.

Steps are tuned to 8 hours, representing usual work

shifts. For each step, information about the patients

and places is saved: whether there are newly admitted

or discharged patients, what movements took place,

which places were contaminated or cleaned, and in

what state of health each patient is. For further de-

tails, we refer the reader to (Kim et al., 2023).

Spatial-Temporal Visualization Tool for Hospital Support for Infection Spread and Outbreaks

309

3.2 User Tasks

With the help of epidemiologists, we have deﬁned a

series of tasks that they need to perform in order to de-

tect when and where the infection transmissions take

place and how infectious outbreaks originate. To be

able to understand the ﬁnal action that experts must

carry out to complete each task, we applied the frame-

work deﬁned by Munzner (Munzner, 2009), which

is useful for describing the reason why each visual-

ization is necessary, thus it helps differ between the

many goals the tool will have. With this framework,

we transformed the description of each task from a

domain-speciﬁc language to a more abstract form,

thus we could see similarities and differences between

them. In this way, we avoid redundancies and over-

lapped tasks. The resulting tasks are presented below,

an example of this process is presented in T1:

T1. Analysis of the situation of the infection in

diverse places in the hospital over time. To identify

and compare the epidemic state of the hospital

between different places and over time using several

epidemiological indicators, which were discussed

and chosen with the experts. These are mortality

rate, incidence, incidence density, period prevalence,

and point prevalence (e.g. when are there more

deceased? Which service presents more incidence?).

Application of Munzner’s framework:

• High level of abstraction: to discover new infor-

mation and generate or verify a hypothesis.

• Mid level of abstraction: to locate cases in the

hospital, the targets are known, but not the loca-

tions.

• Low level of abstraction: to summarize cases and

show an overview of all possible targets.

T2. Detection of sources of infection in a period of

time. To discover when and where the ﬁrst concen-

trations of cases occurred, this is, to locate the focal

point of the infection.

T3. Detection of sources of infection in a point of

time. To trace and identify when and where the ﬁrst

case of infection occurred and how the spread took

place from there by locating the ﬁrst patient that

appears infected.

T4. Study of potential outbreaks. To discover whether

or not an outbreak occurred in those concentrations

of cases. The deﬁnition of infectious outbreak de-

pends on the pathogen that is being studied: in the

case of the CD, an outbreak is deﬁned as three or

more epidemiologically linked cases within a period

of seven days or fewer (West Virginia Bureau for Pub-

lic Health, 2013).

3.3 Tools

We developed the visual tool in the game engine

Unity 3D with C#. This allowed us to create a 3D

representation of a hospital and to show from dif-

ferent perspectives the topology and the interactions

between patients at a low level of abstraction, which

would ease the tracing of the disease. We modeled the

different parts of the hospital and beds with Blender,

and we used PostgreSQL for data persistence. This

way, we could study diverse scenarios and make sev-

eral tests by running the simulator and saving the

newly generated datasets in a database automatically.

4 PROPOSED DESIGN

Based on the deﬁned tasks, we present an interac-

tive visual tool that has three main views (Figure 1):

the Hospital view, which shows information about

the spread of the disease in a spatial and temporal

plane; the Epi view, which shows temporal informa-

tion regarding calculated epidemiological indicators;

and the Tabular view, which shows the values of said

epidemiological indicators of the hospital in a tabular

format sorted by location and time. Next, we are go-

ing to analyze each one in detail and, afterwards, to

list the interactions that the user can perform with the

visual tool.

4.1 Hospital View

The Hospital view shows the patients’ movements at a

low level of abstraction, as well as the evolution of the

endemic situation of the places on a 3D representation

of the hospital (Tasks 1-3). This view is composed of

the visualization of the hospital and a toolbar, which

allows the user to perform certain actions (Figure 1a).

The 3D hospital allows to analyze the spatial distri-

bution of patients at each moment, study where in-

fected patients are located at an individual level, what

contacts occur (i.e. when do they share a room), and

where a greater number of patients are concentrated

in different ranges of time.

Patients can be in one state of health at a time,

which is color-coded: green represents susceptible

patients, yellow exposed patients, red infected pa-

tients, purple recovered patients, black deceased pa-

tients, and blue non-susceptible patients. Through

user-controlled animations, we can identify which

HEALTHINF 2024 - 17th International Conference on Health Informatics

310

Figure 1: The interactive visual interface. (a) The Hospital view shows the spread of the disease in a 3D spatial-temporal

representation. (b) The Tabular view shows the information of the other views in a tabular format. (c) The Epi view shows the

epidemiological indicator selected by the user.

rooms and services the patients went to and how the

infection evolved.

4.2 Epi View

The Epi view helps to study the evolution of the dis-

ease (Tasks 1, 4). In this view, the user can analyze

several epidemiological indicators using line charts

(Figure 1c): on the one hand, cases and point preva-

lence, that are displayed by day; on the other hand,

incidence, period prevalence, mortality rate and in-

cidence density, which are calculated by week. We

present the cases in the SEIRD-NS format, which

means, showing the number of patients in each state

of health by day. Thus, it is possible to have a general

view of the situation throughout the entire simulation,

and then focus on a period of time by means of se-

mantic zooming (Figure 2). Besides this, by hovering

at each point we are able to see detailed information

at each moment of time.

4.3 Tabular View

The Tabular view allows an analysis of the disease

present in the hospital with daily aggregated informa-

tion in space (Tasks 1, 2). In this view (Figure 3),

the user can choose to see an epidemiological indi-

cator (i.e. cases, mortality rate, incidence, incidence

density, period prevalence, and point prevalence) cal-

culated over the entire hospital or grouped by ﬂoor or

service (Figure 3). In the case of the latter, the differ-

ent services are color-coded in coordination with the

Hospital view, thus making it easier to locate them on

the map.

4.4 Interactions

In this section, we describe the interactions between

the user and the tool, as well as between the visual

components. The tasks that epidemiologists and hos-

pital administrators have to carry out are complex in

themselves, and searching for a case of infection in

a hospital with 200 or more beds can be a long and

consuming process. To assist in this work, we have

applied a segmented color scale to easily identify the

different states of health of each patient as indicated

by Aigner et al. (Aigner et al., 2011). In this way,

it is possible to see when transmissions happen – i.e.

when a patient is red (infected) and a roommate turns

yellow (exposed or incubating) –, and which patients

might be future cases (those in yellow) that could sup-

port the spread. This also connects with the Epi view,

in which we use the SEIRD-NS model obtained from

the input data to represent the evolution of cases. In

this model, we encode each state of health with the

same color as in the Hospital view.

The Hospital view gives the user an overview of

the space and allows them to interact with the cam-

era by movements to go to one area or another, zoom

in and out, or rotate to see from another angle. It

is well known that the use of 3D representations can

lead to problems of occlusion, perspective distortions,

and shadows, among others (Munzner, 2015). In or-

der to avoid these problems, we have implemented an

orthographic camera, thus objects are rendered uni-

formly, without a sense of perspective.

In this view, the user can also perform other inter-

actions, such as changing between the hospital ﬂoors

to see one or another; advance, go back, pause, or re-

sume the animation of the simulation; and advance

or go back step by step in the simulation without the

need for animation.

The user can also ﬁlter out the information that

does not interest them at that moment (Shneiderman,

1996). This is possible by means of a toolbar, in

which they can ﬁlter the patients by a range of age,

gender, and state of health. The ﬁltering is applied

to the 3 views. Besides this, users can also perform

zoom actions as deﬁned by Shneiderman (Shneider-

man, 1996) to focus on subsets of patients, by being

able to distinguish patients’ transitions between dif-

Spatial-Temporal Visualization Tool for Hospital Support for Infection Spread and Outbreaks

311

Figure 2: Semantic zooming process. At ﬁrst, the user will see the epidemiological indicator for the whole period of time.

Through a timeline, they can choose the range of time they want to study, and this will be applied to the Epi view.

Figure 3: Tabular view shown with the Hospital view with several infected patients zooming by service. (a) First ﬂoor

containing the remaining services (ER, Radiology and Surgery). (b) Second ﬂoor containing the Wards and the ICU.

ferent categories. This is done by highlighting infor-

mation by ﬂoor or service to see changes in speciﬁc

different areas (Figure 3).

5 DISCUSSION

We have developed an interactive visual tool for the

analysis of the spread of a disease by multidrug-

resistant bacteria, and the investigation of infectious

outbreaks inside hospitals. This tool focuses on the

spatial and temporal dimensions of the movements of

hospitalized patients, as well as on the information

provided by calculated epidemiological indicators.

When developing a medical visual tool, one of the

main risks is the treatment of real clinical data. This

can lead to problems regarding data quality, bias, or

patient privacy, among others (WHO, 2023). By us-

ing realistic data generated with a simulation model

(Kim et al., 2023), we have avoided these problems

associated with handling sensitive and sometimes in-

complete data. Through the use of synthetic data, we

can know exactly where the patients were when their

states of health changed. Regarding applying the tool

to real data, there would be no differences between

a simulated and a real environment in which the data

was obtained with the same precision. Although we

built the tool with simulated data, the current imple-

mentation provides the integration of data from log

ﬁles and relational databases, allowing the integration

from both simulated data and real health information

systems.

The use of Unity for the development involved

both advantages and limitations. Unity is a platform

prepared for developing a 3D tool, so it offers a wide

variety of libraries and great efﬁciency in its compi-

lation time. However, it is not intended for data visu-

alization, and so the asset store does not count with a

great range of chart packages.

Furthermore, the scalability in the hospital mod-

eling entails a challenge and a possible limitation of

this tool. Nevertheless, we are working on an autom-

atization of the creation of hospital ﬂoors with their

rooms with beds in the necessary arrangements (i.e.

orientation and position). The hospital presented in

Figure 3 was created in this way.

This visual tool can have applications in health

and in education: it can be used by epidemiolo-

HEALTHINF 2024 - 17th International Conference on Health Informatics

312

gists and administrators in hospitals in their decision-

making process (e.g. better admission control in the

ER, more isolation of infected patients, or greater care

in the movement of at-risk patients). It can also be ap-

plied, for example, in the Objective Structured Clin-

ical Examination (OSCE), where the skills acquired

by students upon completing their medical degree are

evaluated. In these tests, they create simulated sce-

narios and students must make a series of decisions

as doctors.

Our following step is to validate the visual tool

by means of qualitative evaluations with specialized

users to ensure its usability and usefulness.

6 CONCLUSIONS

In this paper, we propose a decision-support visual

tool to help intertwine the spatial-temporal locations

of hospitalized patients with information about their

health conditions during the spread of an infection

by multidrug-resistant bacteria. To do this, the tool

combines several views which include a 3D visualiza-

tion of a hospital, a 2D visualization of the temporal

progression of the disease, and a tabular visualization

to detail the information depicted in the other views.

The interactivity of the created 3D model allows for

a faithful representation of the movements and the in-

fection processes, while the use of widespread charts

can help understand the temporal progress through

epidemiological indicators. The tasks were deﬁned

together with epidemiologists from hospitals from

Murcia, and we consider that this tool is going to be of

utility in different areas (infection control, decisions

in a hospital, teaching). Our next step is to validate

the usefulness and usability of the tool with said users.

ACKNOWLEDGEMENTS

This work was partially funded by the CON-

FAINCE project (Ref: PID2021-122194OB-I00) by

MCIN/AEI/10.13039/501100011033 and by ”ERDF

A way of making Europe”, by the ”European

Union” or by the ”European Union NextGenera-

tionEU/PRTR”. This research is also partially funded

by the FPI program grant (Ref: PRE2019-089806).

REFERENCES

Aigner, W., Miksch, S., Schumann, H., and Tominski, C.

(2011). Visualization of Time-Oriented Data. Human-

Computer Interaction Series. Springer, London.

Baumgartl, T., Petzold, M., Wunderlich, M., H

ohn, M.,

Archambault, D., Lieser, M., Dalpke, A., Schei-

thauer, S., Marschollek, M., Eichel, V. M., Mut-

ters, N. T., Consortium, H., and von Landesberger, T.

(2021). In Search of Patient Zero: Visual Analytics of

Pathogen Transmission Pathways in Hospitals. IEEE

Transactions on Visualization and Computer Graph-

ics, 27(2):711–721.

Brauer, F. (2008). Compartmental Models in Epidemiology.

Mathematical Epidemiology, 1945:19–79.

Caban, J. J. and Gotz, D. (2015). Visual analytics in health-

care – opportunities and research challenges. Jour-

nal of the American Medical Informatics Association,

22(2):260–262.

Cassini, A., H

ogberg, L. D., Plachouras, D., Quattrocchi,

A., Hoxha, A., Simonsen, G. S., Colomb-Cotinat,

M., Kretzschmar, M. E., Devleesschauwer, B., Cec-

chini, M., Ouakrim, D. A., Oliveira, T. C., Strue-

lens, M. J., Suetens, C., and Monnet, D. L. (2019).

Attributable deaths and disability-adjusted life-years

caused by infections with antibiotic-resistant bacteria

in the EU and the European Economic Area in 2015:

a population-level modelling analysis. The Lancet. In-

fectious Diseases, 19(1):56–66.

Chorianopoulos, K. and Talvis, K. (2016). Flutrack.org:

Open-source and linked data for epidemiology. Health

informatics journal, 22(4):962–974.

Hota, S. S., Achonu, C., Crowcroft, N. S., Harvey, B. J.,

Lauwers, A., and Gardam, M. A. (2012). Determining

Mortality Rates Attributable to Clostridium difﬁcile

Infection. Emerging Infectious Diseases, 18(2):305–

307.

Kim, D., Canovas-Segura, B., Jimeno-Almaz

an, A., Cam-

pos, M., and Juarez, J. M. (2023). Spatial-temporal

simulation for hospital infection spread and outbreaks

of Clostridioides difﬁcile. Scientiﬁc Reports, 13(1).

Number: 1 Publisher: Nature Publishing Group.

onig, B., Limberger, D., Klimke, J., Hagedorn, B., and

ollner, J. (2021). RoomCanvas: A Visualization Sys-

tem for Spatiotemporal Temperature Data in Smart

Homes. The Eurographics Association.

Lital Meyer, S., Ricardo Espinoza, A., and Rodrigo Quera,

P. (2014). Infecci

on por clostridium difﬁcile: epidemi-

olog

ıa, diagn

ostico y estrategias terap

euticas. Revista

edica Cl

ınica Las Condes, 25(3):473–484. Pub-

lisher: Elsevier.

Monegro, A. F., Muppidi, V., and Regunath, H. (2023).

Hospital-Acquired Infections. In StatPearls. Stat-

Pearls Publishing, Treasure Island (FL).

Munzner, T. (2009). A Nested Model for Visualization De-

sign and Validation. IEEE Transactions on Visualiza-

tion and Computer Graphics, 15(6):921–928.

Munzner, T. (2015). Visualization Analysis and Design. A

K Peters/CRC Press.

uller, M., Petzold, M., Wunderlich, M., Baumgartl, T.,

ohn, M., Eichel, V., Mutters, N. T., Scheithauer, S.,

Marschollek, M., and Landesberger, T. v. (2020). Vi-

sual Analysis for Hospital Infection Control using a

RNN Model. The Eurographics Association.

Spatial-Temporal Visualization Tool for Hospital Support for Infection Spread and Outbreaks

313

Oppermann, M. and Munzner, T. (2020). Ocupado: Vi-

sualizing Location-Based Counts Over Time Across

Buildings. Computer Graphics Forum, 39(3):127–

138.

Rind, A., Miksch, S., Aigner, W., Turic, T., and Pohl, M.

(2010). Visuexplore: Gaining new medical insights

from visual exploration. pages 149–152.

Sakai, T., Suzuki, H., Sasaki, A., Saito, R., Tanabe, N., and

Taniguchi, K. (2004). Geographic and temporal trends

in inﬂuenzalike illness, Japan, 1992-1999. Emerging

infectious diseases, 10(10):1822–1826.

Serra-Burriel, M., Keys, M., Campillo-Artero, C., Agodi,

A., Barchitta, M., Gikas, A., Palos, C., and L

opez-

Casasnovas, G. (2020). Impact of multi-drug re-

sistant bacteria on economic and clinical outcomes

of healthcare-associated infections in adults: Sys-

tematic review and meta-analysis. PLoS ONE,

15(1):e0227139.

Shneiderman, B. (1996). The eyes have it: A task by data

type taxonomy for information visualizations. In Pro-

ceedings 1996 IEEE symposium on visual languages,

pages 336–343. IEEE.

Sondag, M., Turkay, C., Xu, K., Matthews, L., Mohr, S.,

and Archambault, D. (2022). Visual Analytics of Con-

tact Tracing Policy Simulations During an Emergency

Response. Computer Graphics Forum, 41(3):29–41.

Su, W., Fu, W., Kato, K., and Wong, Z. S.-Y. (2021). Japan

LIVE Dashboard for COVID-19: A Scalable Solution

to Monitor Real-Time and Regional-Level Epidemic

Case Data. Studies in health technology and infor-

matics, 286:21–25. Place: Netherlands.

West Virginia Bureau for Public Health (2013). Guidelines

for Clostridium difﬁcile (C. diff) Outbreaks in Long-

Term Care Facilities (LTCFs). (Last access Sep 26,

2023).

WHO (2023). Ethics and governance of artiﬁcial intelli-

gence for health.

HEALTHINF 2024 - 17th International Conference on Health Informatics

314