Korea University Sejong Campus Professor Choi Boseung Proposes New Method to Overcome the Limitations of Existing Infectious Disease Spread Models

 

Estimating the Reproduction Number (R) More Accurately with Less Information… Published in Nature Communications

 

Professor Choi Boseung from the Division of Big Data Science at Korea University Sejong Campus, in collaboration with Professor Lee Hyojeong’s team (Kyungpook National University’s Department of Statistics), Chief Investigator Kim Jaekyung (KAIST’s Dempartment of Mathematical Sciences, Institute for Basic Science(IBS) and Center for Mathematical and Computational Science Biomedical Mathematics Group), and Senior Researcher Choi Sunhwa (National Institute for Mathematical Sciences), has proposed a new predictive model for infectious disease spread. This model dramatically improves accuracy, allowing the reproduction number (R) to be estimated without additional epidemiological data, supporting public health experts in developing effective containment strategies.

 

When a novel virus emerges, scientists analyze its structure and behavior and pharmaceutical companies develop vaccines and treatments, while, public health authorities act as a defensive barrier to protect citizens and minimize damage. Mathematics play a key role in accurately predicting impacts, allocating medical resources, and securing hospital beds.

 

The COVID-19 pandemic highlighted the importance of mathematical models in understanding the spread of infectious diseases. Metrics such as the R, incubation period, and infectious period were crucial in analyzing the spread of the virus and designing containment policies.

 

However, traditional models have limitations. Most assume that every contact has the same probability of transmitting the virus, regardless of when the contact occurs. These models, based on Markovian systems that consider only the present state, fail to account for past influences on future conditions.

 

In reality, both past and present states affect future outcomes. The non-Markovian system reflects this by incorporating the incubation period between contact and the onset of infection, making earlier contacts more likely to lead to transmission.

 

Professor Choi stated, “The complexity of mathematical estimation and modeling required by the non-Markovian system, which considers both past and present states, has historically led infectious disease models to rely on Markovian assumptions, which do not fully capture real-world dynamics.”

 

To address this, the research team developed a new infectious disease spread model that considers both past and present states. By introducing delay differential equations that account for past states alongside present ones, the team overcame the limitations of traditional models based on ordinary differential equations.

 

From January 20 to November 25, 2020, using data on COVID-19 cases in Seoul, the team evaluated the accuracy of their new model. During the early rapid spread phase (January 20 to March 3, 2020), the traditional model estimated the reproduction number at 4.9, while the new model estimated it at 2.7. The actual value, determined through contact tracing, was 2.7, demonstrating that traditional models can overestimate the reproduction number by nearly double, leading to exaggerated perceptions of infectiousness.

 

Senior Researcher Choi stated, “Traditional models required additional epidemiological information, such as the infectious period, to adjust for overestimations. The new model’s advantage is its ability to accurately estimate the reproduction number without requiring additional data.”

 

CI Kim stated, “Based on this new model, our team developed a program called ‘IONISE (Inference Of Non-markovIan SEir model)’ and made it freely available for researchers in the field. We hope this will help public health experts better understand the dynamics of infectious disease spread and design more effective containment strategies.”

The research was published in the October 9^th issue of Nature Communications(Impact Factor 14.7).

[Figure 1] A New Method for Estimating Epidemiological Metrics in Infectious Diseases Based on Realistic Assumptions

A collaborative team from IBS, KAIST, Korea University, and NIMS developed a novel method to overcome the fundamental limitations of traditional infectious disease estimation models. Unlike traditional methods that assume future states depend solely on the present based on ordinary differential equations, this new method uses delay differential equations to incorporate past states into future predictions.

[Figure 2] Comparison of Traditional and New Methods for Estimating Epidemiological Metrics

Comparison of results obtained by applying existing and new methods to actual COVID-19 case data.
(a) Both methods accurately approximated the number of confirmed cases.
(b) However, in estimating the effective R, the new method closely approximated the values calculated based on actual contact and infection data (indicated by the dotted line), whereas the existing method overestimated the value by nearly twice as much.

[Figure 3] Analysis of Cumulative COVID-19 Case Data Used for Model Comparison

 

From January 20 to November 25, 2020, the analysis focused on three periods of rapid case growth in Seoul. The new model consistently produced accurate reproduction number estimates, while the traditional model overestimated reproduction numbers by nearly twice during the initial spread phase and showed inconsistent results in subsequent waves, heavily influenced by the model's input parameters. In the third wave phase, the new model and the existing model produced similar results.

[Figure 4] The Research Team

 

△ From left: CI Kim Jaekyung (IBS/KAIST, Co-Corresponding Author), Dr. Hong Hyukpyo (IBS/KAIST, currently at University of Wisconsin, Co-First Author), Senior Researcher Choi Sunhwa (NIMS, Co-Corresponding Author), Dr. Um Eunjin (Korea University, currently at the Korea Disease Control and Prevention Agency, Co-First Author), and Professor Choi Boseung (IBS/Korea University, Co-Corresponding Author).

 

[Figure 5] The Research Team

 

△ From left: Kim Jaekyung (IBS/KAIST, Co-Corresponding Author), Professor Choi Boseung (IBS/Korea University, Co-Corresponding Author), Dr. Hong Hyukpyo (IBS/KAIST, currently at University of Wisconsin, Co-First Author), Professor Lee Hyojeong (Kyungpook National University), Senior Researcher Choi Sunhwa (NIMS, Co-Corresponding Author), and Dr. Um Eunjin (Korea University, currently at Korea Disease Control and Prevention Agency, Co-First Author).

 

KU Sejong Student PR Team, KUS-ON

Translator: Seo Yujeong