=Paper=
{{Paper
|id=Vol-3641/paper2
|storemode=property
|title=Estimation of Ukrainian Refugees Impact on COVID-19 Dynamics in Czech Republic using Prophet Model
|pdfUrl=https://ceur-ws.org/Vol-3641/paper2.pdf
|volume=Vol-3641
|authors=Dmytro Chumachenko,Vasyl Chomko,Ievgen Meniailov,Serhii Krivtsov,Halyna Padalko
|dblpUrl=https://dblp.org/rec/conf/profitai/ChumachenkoCMKP23
}}
==Estimation of Ukrainian Refugees Impact on COVID-19 Dynamics in Czech Republic using Prophet Model==
https://ceur-ws.org/Vol-3641/paper2.pdf
Estimation of Ukrainian Refugees Impact on COVID-19
Dynamics in Czech Republic using Prophet Model
Dmytro Chumachenko1,2, Vasyl Chomko3, Ievgen Meniailov4, Serhii Krivtsov1, and
Halyna Padalko1
1 National Aerospace University “Kharkiv Aviation Institute”, Chkalow str. 17, Kharkiv, 61070, Ukraine
2 Max Planck Institute for Demographic Research, Konrad-Zuse-Straße 1, Rostock, 18057, Germany
3 University of Waterloo, 200 University Ave W, Waterloo, N2L 3G1, Canada
4 V.N. Karazin Kharkiv National University, 4 Svobody Sq, Kharkiv, 61022, Ukraine
Abstract
This study aims to examine the impact of the Russian invasion of Ukraine on the epidemiological
trajectory of COVID-19, demonstrating the interplay between geopolitical instability and infectious
disease dynamics. We developed a machine learning model adapted to incorporate the onset of military
conflict into its predictive framework. The model's performance was evaluated against actual COVID-19
morbidity and mortality data before and after the invasion, considering a set of identified
epidemiological and conflict-related variables. The predictive accuracy of COVID-19 cases showed a
marked decline post-invasion, highlighting the perturbing effect of the conflict on disease spread
dynamics. Conversely, the mortality predictions exhibited increased accuracy, suggesting a differential
impact of the conflict on various aspects of the epidemic process. The study's findings underscore the
necessity of revising epidemiological models in conflict scenarios. It illustrates that incorporating
geopolitical events can significantly alter predictive outcomes and, therefore, should be considered in
public health planning and response strategies. The research calls for expanded models that can
dynamically integrate various global sociopolitical disturbances and for interdisciplinary efforts to
create robust forecasting tools capable of adapting to rapidly changing conditions.
Keywords
COVID-19, Prophet, machine learning, deep learning, war, epidemic model 1
1. Introduction
The COVID-19 pandemic has been a significant event worldwide, leading to widespread
challenges and major changes in how we handle public health, the economy, and social
interactions [1]. It has forced us to look closely at our healthcare systems and has pushed quick
advancements in technology and science. The pandemic has affected everyday life, international
travel, and the movement of goods worldwide, showing how closely connected we all are [2]. It
has also highlighted the need for countries to work together to handle such global crises. The
importance of the COVID-19 pandemic extends beyond its immediate impact; it has important
lessons for how we make policies, prepare for emergencies, and view health as an essential part
of our overall safety and well-being [3].
The COVID-19 pandemic's trajectory within the Czech Republic has been characterized by
several distinctive phases, mirroring global trends while exhibiting unique national peculiarities
[4]. Initially, the country was lauded for its prompt response to the initial outbreak, employing
stringent measures such as border closures, mandatory mask usage, and early lockdowns, which
effectively curtailed the spread of the virus [5]. However, the subsequent waves of infection
ProfIT AI 2023: 3rd International Workshop of IT-professionals on Artificial Intelligence (ProfIT AI 2023), November
20–22, 2023, Waterloo, Canada
dichumachenko@gmail.com (D. Chumachenko); vchomko@uwaterloo.ca (V. Chomko);
evgenii.menyailov@gmail.com (I. Meniailov); krivtsovpro@gmail.com (S. Krivtsov); galinapadalko95@gmail.com
(H. Padalko)
0000-0003-2623-3294 (D. Chumachenko); 0009-0009-4419-6651 (V. Chomko); 0000-0002-9440-8378
(I. Meniailov); 0000-0001- 5214-0927 (S. Krivtsov); 0000-0001-6014-1065 (H. Padalko)
© 2023 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org)
CEUR
ceur-ws.org
Workshop ISSN 1613-0073
Proceedings
witnessed a reversal of fortune as the nation grappled with rising cases, highlighting the strains
on its healthcare system [6]. The public health strategy oscillated between various levels of
restrictions, reflecting a balance between epidemiological caution and socio-economic pressures.
The pandemic further acted as a catalyst for the Czech Republic to enhance its healthcare
infrastructure, accelerate digital transformation in the provision of health services, and refocus
on the importance of epidemiological forecasting and preventive measures [7]. The national
response evolved, incorporating lessons from earlier phases to inform policy adaptations and
health directives aimed at mitigating the pandemic's impact and facilitating a recovery path
focused on resilience and health security [8].
The Russian invasion of Ukraine has severely exacerbated the COVID-19 crisis during a critical
phase of the Omicron variant's spread [9]. Predicted to face thousands of new cases daily,
Ukraine's healthcare system has been overwhelmed by the dual challenges of war and the
pandemic [10]. Medical facilities in conflict zones have been destroyed or rendered inoperative,
hindering the identification and diagnosis of COVID-19 cases, which has led to a substantial
underreporting of the actual infection rates [11]. Hospital capacities have been strained as beds
and resources allocated for COVID-19 patients are now redirected to treat those wounded in the
conflict, curtailing the treatment options for the virus and limiting access to essential medicines
and oxygen [12].
Moreover, standard public health measures such as social distancing and mask-wearing have
become nearly impossible to enforce, especially for individuals in bomb shelters or during
evacuations, which has likely intensified the virus's spread [13]. The invasion has also brought
Ukraine's vaccination campaign to a standstill, particularly in areas of active conflict, causing a
further decline in preventative measures against COVID-19.
The convergence of the war and the pandemic has shifted the focus of Ukraine's healthcare
system away from the virus, prioritizing immediate survival and care for injuries over pandemic
control [14]. This has inadvertently led to an increase in the virus's transmission, compounding
the country's humanitarian crisis.
Integrating information technologies and simulation models has been instrumental in
combating the pandemic, providing vital tools for forecasting, decision-making, and managing
healthcare resources [15]. Simulation models, leveraging vast datasets, have been pivotal in
predicting the spread of the virus, enabling governments and healthcare organizations to enact
preemptive measures and optimize response strategies [16]. Information technologies, such as
AI and machine learning, have facilitated the real-time analysis of infection rates, hospitalization,
and resource utilization, enhancing the ability to anticipate and respond to emerging hotspots
[17]. Furthermore, digital contact tracing apps have aided in mitigating transmission by
identifying and alerting individuals who may have been exposed to the virus [18]. These
technological approaches, when combined with robust data analytics, have not only contributed
to a more informed and agile response to the COVID-19 pandemic but have also laid the
groundwork for innovative solutions to future public health challenges [19].
The aim of the paper is to investigate the impact of the Russian full-scale invasion on the
COVID-19 dynamics in Czech Republic using Prophet model.
2. Materials and Methods
This research advances a methodological framework designed to evaluate the repercussions of
the Russian full-scale invasion of Ukraine on the epidemiological trends of infectious diseases.
Initially, the approach entails crafting a predictive machine learning algorithm tailored to map
the epidemic's trajectory within the chosen locale. Subsequently, the model's accuracy is
ascertained by applying it to pre-invasion health data, specifically morbidity and mortality
statistics associated with infectious diseases, collated from the 25th of January to the 23rd of
February 2022.
Upon validation, the model projects the expected disease incidence and related fatalities
within the initial thirty-day period following the onset of the Russian military invasion beginning
on the 24th of February 2022. A comparative analysis examines the divergence between the
model's projections and the health outcomes observed during the conflict.
Further, the study delves into dissecting the elements that impinge upon the progression of
infectious diseases, considering the peculiarities of both the pathogen in question and the
geographical setting. In the final phase of the methodology, a critical evaluation of the findings is
conducted to unearth potential risks to the public health infrastructure. This comprehensive
examination is pivotal for informing the strategic deployment of interventions to mitigate the
propagation of infectious diseases amidst wartime conditions.
The Prophet model, developed by Facebook's Core Data Science team, is a decomposable time-
series forecasting procedure for simulating COVID-19 dynamics, among other applications [21].
This model can be beneficial for public health officials to forecast the number of cases,
hospitalizations, and other critical variables over time.
The underlying principle of the Prophet model is to decompose the time series into three main
components: trend, seasonality, and holidays. It accommodates non-linear trends with daily,
weekly, and yearly seasonality, plus holiday effects. It does this by fitting the following equation:
y(t) = g(t) + s(t) + h(t) + 𝜖𝑡 , (1)
where y(t) is the predicted value; g(t) represents the trend function, which models non-periodic
changes; s(t) models periodic changes, such as seasonality; h(t) represents the effects of holidays
which occur on potentially irregular schedules over one or more days; 𝜖𝑡 represents the error
term which accounts for any unusual changes not accommodated by the model.
The trend component g(t) is typically modeled using a piecewise linear or logistic growth
curve to handle different types of growth patterns in the data:
For linear growth:
𝑔(t) = kt + m, (2)
For logistic growth:
𝐶 (3)
𝑔(t) = ,
1 + 𝑒 −𝑘(𝑡−𝑚)
where, k represents the growth rate, m dictates the offset parameter, and C is the carrying
capacity (the maximum achievable point for the metric being forecasted, which is particularly
relevant when forecasting something like an infection rate that has a natural cap).
Seasonality, s(t), is modeled using the Fourier series to provide a flexible model of periodic
effects. Prophet relies on a yearly seasonal component using a Fourier series with P yearly
seasonality, which is given by:
𝑁 2𝜋𝑛𝑡 2𝜋𝑛𝑡 (4)
𝑠(t) = ∑ (𝑎𝑛 cos ( ) + 𝑏𝑛 sin( )),
𝑛=1 𝑃 𝑃
where N is the number of Fourier terms used to model the seasonality, and a n, bn are coefficients
to be fitted.
Holiday effects, h(t), are included as an indicator function that equals 1 if the time t is a holiday
0 otherwise.
The error term ϵt generally assumes that these unpredictable effects are normally distributed.
The parameters for these models are fitted using historical data, and the model is then used to
forecast future trends. The Prophet model's strength lies in its robustness to missing data and
trend shifts, making it useful for scenarios such as COVID-19, where sudden changes in trend can
occur due to lockdowns or other interventions [21].
For COVID-19 dynamics, this model can incorporate effects such as lockdowns or vaccination
drives as holiday effects and accommodate sudden changes in the trend by adjusting the growth
rate k. The flexibility to incorporate these features makes the Prophet model a powerful tool for
public health forecasting and planning during the pandemic.
The Prophet model has been acknowledged for its intuitive parameters that are relatively
simple for analysts to adjust. It facilitates an ease of use not always present in more complex
forecasting algorithms. One of its significant advantages is the model's capacity to handle missing
data and outliers, as well as to incorporate potentially influential events like holidays, which can
be critical for predicting erratic behaviors in time-series data. Additionally, the model's
robustness against shifts in trends and its ability to capture a wide range of seasonalities – weekly,
monthly, or annually – render it particularly beneficial for long-term forecasting.
However, the Prophet model is not without its drawbacks. Its performance is heavily reliant
on the quality and granularity of the input data; insufficient or overly aggregated data can
substantially compromise its accuracy. Moreover, the model may fall short when dealing with
highly volatile data series or non-standard frequencies not accounted for in its core structure.
The simplistic assumption of independent holiday effects might also not hold in more complex
scenarios where holidays interact with underlying trends. Finally, while Prophet excels in its
simplicity and accessibility, it may not reach the sophistication required for capturing more
intricate relationships within the data that advanced statistical or machine learning models might
detect.
The model's hyperparameters, which are crucial for its predictive capacity, are outlined in
Table 1.
Table 1
Model’s hyperparameters
Hyperparameter Cumulative Cumulative New cases New deaths
cases deaths
changepoint_prior_scale 0.2445491199 0.49594308323 0.0162143017 0.01242620057
4868200 1842 991519 64601
changepoint_range 0.9044859898 0.80698749297 0.8736375440 0.88998889672
296930 9227 93367 8438
daily_seasonality FALSE TRUE TRUE FALSE
growth Linear Linear Linear Linear
seasonality_mode Multiplicative Additive Additive Additive
seasonality_prior_scale 0.0041598611 0.00579797001 0.1161495038 0.00304460686
9232672 191652 91177 463523
weekly_seasonality TRUE TRUE FALSE FALSE
yearly_seasonality FALSE FALSE FALSE TRUE
The table delineates the parameters for four different epidemiological measures: cumulative
cases, cumulative deaths, new cases, and new deaths. For example, the changepoint_prior_scale,
which governs the flexibility of the trend, is set to approximately 0.245 for cumulative cases and
is higher at nearly 0.496 for cumulative deaths, suggesting a more responsive adjustment to
changes for the latter. The changepoint_range, indicating the proportion of the history where
trend changes are sought, varies across the measures but remains above 0.8, providing a broad
window for detecting such changes.
Regarding seasonality, daily_seasonality is enabled only for cumulative deaths and new cases,
implying that daily fluctuations are significant for these measures. The growth parameter is
consistent 'Linear' across all measures, indicating a linear trend assumption in the model's
structure. The seasonality_mode differs; 'Multiplicative' is used for cumulative cases, whereas
'Additive' is preferred for the other measures, reflecting different approaches to how seasonality
impacts the trend.
Seasonality_prior_scale, which sets the importance of seasonal components, is notably low for
cumulative cases and new deaths, suggesting a conservative approach to fitting seasonal effects.
Weekly_seasonality is applied to cumulative cases and deaths, which could capture weekly
patterns like reporting fluctuations. Conversely, yearly_seasonality is considered only for new
deaths, potentially to account for long-term seasonal effects such as weather changes or
systematic variations in mortality reporting.
3. Results
Utilizing the Python programming environment, the Prophet model was deployed to evaluate
and refine its predictive performance regarding the COVID-19 pandemic's trend. The model's
forecasting accuracy was methodically examined across time horizons of three to thirty days. For
validation, this study utilized COVID-19 incidence and mortality statistics for the Czech Republic,
procured from the World Health Organization's COVID-19 Dashboard [22]. The data set for
assessment included daily reported cases and fatalities attributed to COVID-19 within the Czech
Republic, covering the period from the 25th of January to the 23rd of February 2022. This interval
reflects the initial 30 days after the beginning of the full-scale Russian invasion of Ukraine.
Figure 1 shows the results of forecasting of the COVID-19 cumulative cases in Czech Republic
from 25 January, 2022 to 23 February, 2022.
Figure 1: Forecast of the COVID-19 cumulative cases in Czech Republic (25.01.22 – 23.02.22)
Figure 2 shows the results of forecasting of the COVID-19 cumulative deaths in Czech Republic
from 25 January, 2022 to 23 February, 2022.
To evaluate the performance of the model we have used the Mean Absolute Percentage Error
(MAPE). MAPE is a statistical measure used to assess the accuracy of a forecasting model [23]. It
quantifies the average magnitude of errors as a percentage, providing a clear representation of
the forecast accuracy in relative terms. This is calculated by taking the mean of the absolute values
of the individual percentage errors between the forecasted and actual values. MAPE is
particularly useful as it offers a straightforward interpretability of the forecast precision, with
lower values indicating higher accuracy. However, its effectiveness may be limited when dealing
with data points close to zero, as the percentage error can become disproportionately high.
Figure 2: Forecast of the COVID-19 cumulative deaths in Czech Republic (25.01.22 – 23.02.22)
Table 2 shows MAPE for cumulative data forecast from 25 January, 2022 to 23 February, 2022.
Table 2
MAPE of the cumulative data forecast (25.01.22 – 23.02.22)
Period of the forecast (days) Infected Dead
3 days 3,24% 1,23%
7 days 5,58% 1,07%
14 days 9,08% 0,91%
21 days 11,63% 0,85%
30 days 13,64% 0,82%
The MAPE values for the infected category show an increasing trend as the forecast period
lengthens, starting from 3.24% for the 3-day forecast and rising to 13.64% for the 30-day
forecast. This increment suggests that the model's predictive accuracy decreases as the
forecasting horizon extends. In contrast, the MAPE for the dead category decreases slightly over
more extended forecast periods, starting from 1.23% at 3 days and reducing to 0.82% at 30 days,
indicating a relatively consistent and better forecasting performance for cumulative deaths
compared to infections over the given period.
Forecasts were calculated to examine the dynamics of the COVID-19 epidemic within the
Czech Republic from 24 February, 2022 through 25 March, 2022. This interval marks the first
month after the beginning of the full-scale Russian invasion of Ukraine.
Figure 3 shows the results of forecasting of the COVID-19 cumulative cases in Czech Republic
from 24 February, 2022 to 25 March, 2022.
Figure 4 shows the results of forecasting of the COVID-19 cumulative deaths in Czech Republic
from 24 February, 2022 to 25 March, 2022.
Table 3 shows MAPE for cumulative data forecast from 24 February, 2022 to 25 March, 2022.
Figure 3: Forecast of the COVID-19 cumulative cases in Czech Republic (24.02.22 – 25.03.22)
Figure 4: Forecast of the COVID-19 cumulative deaths in Czech Republic (24.02.22 – 25.03.22)
Table 3
MAPE of the cumulative data forecast (24.02.22 – 25.03.22)
Period of the forecast (days) Infected Dead
3 days 18,12% 0,78%
7 days 17,89% 0,76%
14 days 17,39% 0,67%
21 days 16,89% 0,52%
30 days 16,27% 0,45%
Throughout forecast periods extending from 3 to 30 days, the MAPE for the infected category
starts at 18.12% for the 3-day forecast and demonstrates a downward trend, decreasing to
16.27% for the 30-day forecast. This indicates a gradual improvement in the model's predictive
accuracy over time for forecasting infections. Similarly, the MAPE for the dead category shows a
consistent decrease from 0.78% at the 3-day forecast to 0.45% at the 30-day forecast, suggesting
a sustained enhancement in the precision of death forecasts as the forecast period lengthens.
4. Discussion
In the discussion of the interplay between the COVID-19 pandemic dynamics and significant
sociopolitical events, particularly the full-scale Russian invasion of Ukraine, several critical
insights emerge. This paper has engaged with the multifaceted repercussions of the conflict on
public health, primarily through the lens of infectious disease spread and management. The
empirical data gathered and forecasted via the Prophet model illustrate notable deviations in the
infection and mortality rates from the predicted trajectories post-invasion, underscoring the
profound impact of war on healthcare systems and disease control measures.
The observed decline in the accuracy of the cumulative case forecasts for the period following
the Russian full-scale invasion of Ukraine (24.02.22 - 25.03.22) suggests that the epidemiological
dynamics of COVID-19 in the Czech Republic were significantly altered during this time. This
decrement in predictive accuracy could be attributable to a myriad of factors induced by the
conflict, including but not limited to disruptions in routine healthcare practices, alterations in
population behavior due to the crisis, and the challenges in maintaining rigorous disease
surveillance and reporting amidst socio-political turmoil.
Conversely, the improvement in forecast accuracy regarding cumulative deaths within the
same timeframe offers a paradoxical narrative. It implies that, despite the upheaval caused by the
invasion and its impact on the health system's efficacy in tracking and managing COVID-19 cases,
the prediction of fatal outcomes became more precise. This increased accuracy could be
indicative of several scenarios. One possibility is that the data reporting for mortality could have
become more consistent or accurate despite the conflict. Another perspective might suggest that
while case numbers were subject to more significant fluctuation due to the reasons mentioned
above, the fatality rate may have followed a more predictable pattern, potentially due to
improved treatment protocols or a stable or decreasing virulence of the prevalent COVID-19
variants.
In the extended discussion of these outcomes, it is critical to integrate a nuanced
understanding of how the chaos of war can ripple through the public health infrastructure,
influencing not just the spread of infectious diseases but also the mechanisms of data collection
and analysis. The increase in predictive accuracy for cumulative deaths, despite the invasion, may
reflect a complex interplay of factors, including potentially more focused medical attention to
severe COVID-19 cases or an artifact of data prioritization that inherently occurs in crisis settings.
This dichotomy between case and death predictions highlights the complexity of
epidemiological modeling in the context of significant social disruptions. It further reinforces the
need for dynamic modeling approaches sensitive to rapid changes in public health landscapes.
The insights gained from analyzing these predictive discrepancies underscore the importance of
continuous model refinement and the incorporation of real-time data to inform public health
strategies, particularly in settings of socio-political instability.
The disruption of healthcare services due to the war cannot be overstated. It precipitated a
notable decline in routine medical care, vaccination efforts, and a reorientation of health
resources towards emergent war-related injuries and traumas. Consequently, the diversion of
medical resources has been paralleled by increased COVID-19 transmission risks, mainly due to
population displacements and the congregation of individuals in shelters devoid of appropriate
sanitary conditions.
Moreover, assessing machine learning models, like Prophet, in forecasting the pandemic's
trajectory has yielded insights into the necessity and utility of adaptable and robust predictive
tools in public health. Despite their utility, these models face limitations, especially when external
variables such as social upheaval are introduced. The forecast errors, represented by MAPE,
indicate variability in predictive accuracy, which tends to be more pronounced over extended
forecast periods. Nonetheless, it is imperative to acknowledge that forecasting models are
indispensable for early intervention strategies and resource allocation in public health crises.
Migration patterns in response to the invasion have also significantly affected the COVID-19
pandemic's landscape. The influx of refugees into neighboring countries like the Czech Republic
has introduced new challenges for public health surveillance and infrastructure. The movement
of large populations across borders during a pandemic can exacerbate the spread of the virus,
overwhelming local healthcare systems and complicating containment and mitigation efforts.
In reflecting upon these phenomena, it is evident that an integrative approach to public health
planning is essential - one that considers epidemiological trends in the context of geopolitical and
societal shifts. The pandemic's trajectory cannot be fully understood nor anticipated without
accounting for such extrinsic factors. Furthermore, the observed data underline the critical need
for international cooperation and support, significantly when national healthcare systems are
compromised due to factors like armed conflict.
Examining the cumulative data forecasts, particularly the MAPE across different timeframes,
further solidifies the argument that disease forecasting in societal upheaval requires models to
incorporate real-time changes in social behavior and healthcare access. While the model
displayed a decreasing trend in predictive error for cumulative deaths, indicating an
improvement over time, this could also reflect the adaptation of healthcare systems to crisis
conditions or changes in reporting accuracy under duress.
In summary, the interrelation between a global pandemic and an unanticipated military
conflict has revealed the complexities of managing public health in the contemporary world. The
experience underlines the importance of resilient health systems, the adaptability of predictive
models, and the need for comprehensive health policies to withstand the pressures of unforeseen
global events.
5. Conclusions
In conclusion, the study's findings delineate the multifaceted impact of the Russian invasion of
Ukraine on the epidemiological trajectory of COVID-19 within the Czech Republic. The decreased
precision in forecasting cumulative COVID-19 cases post-invasion underscores the profound
effect geopolitical instability can have on disease surveillance and prediction. It is indicative of
the myriad ways in which conflict can disrupt healthcare systems and the collation and reliability
of epidemiological data.
In contrast, the improved accuracy in predicting COVID-19-related mortality during the same
period presents an intriguing counterpoint, suggesting a differential effect of the conflict on
various aspects of the pandemic's progression. This improvement could reflect an unanticipated
standardization in mortality data reporting or a proper stabilization in mortality trends despite
the chaos of the surrounding circumstances.
These results illuminate the complexities of public health monitoring in times of crisis and the
crucial role that accurate, timely data plays in pandemic response efforts. They highlight the
resilience of mortality data amidst disruptions and underscore the potential need for augmented
predictive modeling techniques that can better account for the unexpected variability introduced
by such external factors.
Ultimately, this study emphasizes the need for robust, adaptable public health infrastructure
to maintain surveillance and care delivery during geopolitical crises. It also calls for international
collaboration in data sharing and model development to enhance preparedness and response
strategies in the face of concurrent public health and political emergencies. The profound
implications of these findings advocate for a reinforced global commitment to public health as an
integral component of national security and stability.
The scientific novelty of this research resides in its analytical approach, integrating
geopolitical events into epidemiological predictive models. By capturing the influence of the
Russian invasion of Ukraine, the study provides a novel empirical understanding of how such
crises can perturb disease transmission dynamics and forecasting. It extends current
epidemiological forecasting methodologies by introducing conflict parameters, thus offering a
more nuanced model sensitive to sudden sociopolitical shifts, bridging the gap between public
health research and political science.
On the practical front, the research introduces a new paradigm in public health surveillance
by accounting for the impact of military conflict on infectious disease progression. The enhanced
model is a tool for policymakers and healthcare providers, allowing for more informed decision-
making in crisis conditions. This approach enables a dynamic adjustment of health resources and
responses in real time, potentially mitigating the adverse effects of conflicts on the spread of
infectious diseases. The practicality of this model lies in its adaptability, providing a template for
how health systems can maintain operational integrity amidst the chaos of war or other
disruptions.
Future research should refine these models to include a broader spectrum of sociopolitical
disruptions, such as economic sanctions, natural disasters, or large-scale migration. Expanding
the data set to global incidences can enrich the model's predictive power and application across
different contexts. Further exploration into machine learning and artificial intelligence could
enhance the model's ability to learn from complex and non-linear data, providing more accurate
forecasts in the face of uncertainty. Moreover, interdisciplinary collaboration will be essential,
integrating insights from epidemiology, political science, and data science to bolster the model's
robustness and utility in anticipating public health needs during global crises.
Acknowledgements
Research was funded by a Max Planck Society Eirene Grant (awarded 2022) in the framework of
the research project “Data-driven investigation of the population dynamics caused by the war in
Ukraine”.
References
[1] A. Renieri, “COVID-19: a challenge and an opportunity,” European Journal of Human Genetics,
vol. 30, no. 8, Jul. 2022, doi: 10.1038/s41431-022-01142-6.
[2] H. Onyeaka, C. K. Anumudu, Z. T. Al-Sharify, E. Egele-Godswill, and P. Mbaegbu, “COVID-19
pandemic: A review of the global lockdown and its far-reaching effects,” Science Progress, vol.
104, no. 2, p. 003685042110198, Apr. 2021, doi: 10.1177/00368504211019854.
[3] D.-C. Thompson et al., “The Impact of COVID-19 Pandemic on Long-Term Care Facilities
Worldwide: An Overview on International Issues,” BioMed Research International, vol. 2020,
pp. 1–7, Nov. 2020, doi: 10.1155/2020/8870249.
[4] M. Tuček and V. Vaněček, “COVID-19 in the Czech Republic 2020 and 2021: comparative
analysis of probable work-related transmission of the coronavirus SARS-CoV-2,” Central
European Journal of Public Health, vol. 30, no. 3, pp. 201–204, Sep. 2022, doi:
10.21101/cejph.a7610.
[5] L. Berec et al., “Delays, Masks, the Elderly, and Schools: First Covid-19 Wave in the Czech
Republic,” Bulletin of Mathematical Biology, vol. 84, no. 8, p. 75, Jun. 2022, doi:
10.1007/s11538-022-01031-5.
[6] M. Fajfr et al., “Long-Term Antibody Response and Vaccination Efficacy in Patients with
COVID-19: A Single Center One-Year Prospective Study from the Czech Republic,” Viruses,
vol. 14, no. 3, p. 526, Mar. 2022, doi: 10.3390/v14030526.
[7] L. Přibylová, V. Eclerová, O. Májek, J. Jarkovský, T. Pavlík, and L. Dušek, “Using real-time
ascertainment rate estimate from infection and hospitalization dataset for modeling the
spread of infectious disease: COVID-19 case study in the Czech Republic,” PLOS ONE, vol. 18,
no. 7, pp. e0287959–e0287959, Jul. 2023, doi: 10.1371/journal.pone.0287959.
[8] M. Komenda et al., “Complex Reporting of the COVID-19 Epidemic in the Czech Republic: Use
of an Interactive Web-Based App in Practice,” Journal of Medical Internet Research, vol. 22,
no. 5, p. e19367, May 2020, doi: 10.2196/19367.
[9] D. Chumachenko and T. Chumachenko, “Impact of war on the dynamics of COVID-19 in
Ukraine,” BMJ Global Health, vol. 7, no. 4, p. e009173, Apr. 2022, doi: 10.1136/bmjgh-2022-
009173.
[10] I. Nesteruk, “Visible and Real Sizes of New COVID-19 Pandemic Waves in Ukraine,” Innovative
Biosystems and Bioengineering, vol. 5, no. 2, pp. 85–96, May 2021, doi:
10.20535/ibb.2021.5.2.230487.
[11] M. Dzhus and I. Golovach, “Impact of Ukrainian-Russian war on healthcare and humanitarian
crisis,” Disaster Medicine and Public Health Preparedness, vol. 17, pp. 1–9, Dec. 2022, doi:
10.1017/dmp.2022.265.
[12] B. S. Levy and J. Leaning, “Russia’s War in Ukraine — The Devastation of Health and Human
Rights,” New England Journal of Medicine, vol. 387, no. 2, Jun. 2022, doi:
10.1056/nejmp2207415.
[13] I. Pinchuk, R. Goto, Oleksiy Kolodezhny, Nataliia Pimenova, and N. Skokauskas, “Dynamics of
hospitalizations and staffing of Ukraine’s mental health services during the Russian
invasion,” International Journal of Mental Health Systems, vol. 17, no. 1, Jun. 2023, doi:
10.1186/s13033-023-00589-4.
[14] O. Uwishema et al., “Russia-Ukraine conflict and COVID-19: a double burden for Ukraine’s
healthcare system and a concern for global citizens,” Postgraduate Medical Journal, vol. 98, p.
postgradmedj-2022-141895, Jun. 2022, doi: 10.1136/postgradmedj-2022-141895.
[15] I. Izonin, R. Tkachenko, I. Dronyuk, P. Tkachenko, M. Gregus, and M. Rashkevych, “Predictive
modeling based on small data in clinical medicine: RBF-based additive input-doubling
method,” Mathematical Biosciences and Engineering, vol. 18, no. 3, pp. 2599–2613, 2021, doi:
10.3934/mbe.2021132.
[16] A. Nechyporenko et al., “Comparative Characteristics of the Anatomical Structures of the
Ostiomeatal Complex Obtained by 3D Modeling,” 2020 IEEE International Conference on
Problems of Infocommunications Science and Technology, PIC S and T 2020 - Proceedings, pp.
407–411, Oct. 2020, doi: 10.1109/picst51311.2020.9468111.
[17] K. Nixon et al., “Real-time COVID-19 forecasting: challenges and opportunities of model
performance and translation,” The Lancet Digital Health, vol. 4, no. 10, pp. e699–e701, Oct.
2022, doi: 10.1016/S2589-7500(22)00167-4.
[18] C.-E. Juneau, A.-S. Briand, P. Collazzo, U. Siebert, and T. Pueyo, “Effective contact tracing for
COVID-19: A systematic review,” Global Epidemiology, p. 100103, Mar. 2023, doi:
10.1016/j.gloepi.2023.100103.
[19] O. Kalinina, I. Skachkova, O. Skachkov, I. Kadykova, and I. Chumachenko, “Modeling the
Decision-Making Process in Project Management of Innovative Diffusion of Socio-Economic
Systems,” CEUR Workshop Proceedings, vol. 3453, pp. 24–35, 2023.
[20] S. J. Taylor and B. Letham, “Forecasting at Scale,” The American Statistician, vol. 72, no. 1, pp.
37–45, Sep. 2017, doi: 10.1080/00031305.2017.1380080.
[21] S. Sah, B. Surendiran, R. Dhanalakshmi, S. N. Mohanty, F. Alenezi, and K. Polat, “Forecasting
COVID-19 Pandemic Using Prophet, ARIMA, and Hybrid Stacked LSTM-GRU Models in India,”
Computational and Mathematical Methods in Medicine, vol. 2022, pp. 1–19, May 2022, doi:
10.1155/2022/1556025.
[22] World Health Organization, “WHO COVID-19 dashboard,” World Health Organization, 2023.
https://covid19.who.int/ (accessed Oct. 15, 2023).
[23] V. P. Mashtalir, V. V. Shlyakhov, and S. V. Yakovlev, “Group Structures on Quotient Sets in
Classification Problems,” Cybernetics and Systems Analysis, vol. 50, no. 4, pp. 507–518, Jul.
2014, doi: 10.1007/s10559-014-9639-z.