Review: Theoretical Epidemiology Needs Urgent Attention in China

Based on our comprehensive review of theoretical epidemiology literature, several seminal contributions have shaped our understanding of disease dynamics. R. A. Ross, in 1902, employed mathematical models to evaluate malaria interventions (7), while W. H. Hamer in 1906 highlighted the importance of reducing susceptible populations to control epidemics (8–10). In 1928, Reed-Frost elucidated infection spread across generations (11). McKendrick and Kermack between 1927 and 1935 introduced differential equations to depict disease transmission and the transmission threshold concept (12–14). George MacDonald further explored transmission dynamics by posing critical questions regarding infected individuals entering a population: What would occur if an infected individual entered a population? How many new infections would result (15)? The Threshold Theory in 1970 suggested that 80% vaccine coverage could eradicate smallpox. Mathematical modeling gained prominence following the 2003 SARS outbreak, aiding in transmissibility calculations and evaluating prevention measures (16–20). The 2009 Influenza A (H1N1) pandemic witnessed extensive modeling to assess vaccination effectiveness and establish influenza surveillance systems (21–24). The COVID-19 pandemic prompted accelerated development in theoretical epidemiology, with various models predicting trends, simulating transmission, and evaluating control measures to inform government decision-making (3–4,25–30). Subsequently, theoretical epidemiology has focused on refining models and addressing previous limitations, such as real-time database optimization and dynamic modeling to monitor disease progress (31–32).

Upon reviewing the literature, we identified two main types of mathematical models in theoretical epidemiology: “data-driven” models and “mechanism-driven” models.

“Data-driven” models encompass various approaches to investigate the connection between disease occurrence and time. These methods involve temporal regression models, control charts, time series models like autoregressive integrated moving average model (ARIMA), Monte Carlo algorithmic models, gray theory models, neural network models, and related derivatives (33).

This category of methods focuses on analyzing large datasets of case information to comprehend the trends of diseases or health conditions over time. It involves examining the correlation between disease occurrences and time or other variables, developing models, and forecasting disease prevalence trends.

This model type offers the advantage of rapid and simple forecasting based on historical data. However, its primary drawbacks include using time as an independent variable and attributing disease development solely to temporal changes. These models often rely heavily on data, overlooking the infectious and epidemiologic aspects of disease (34). Typically, such models focus solely on prediction, utilizing historical data to simulate scenarios for forecasting, and rarely assess the effectiveness of infectious disease prevention and control strategies (35–36).

“Mechanism-driven” models are grounded in a comprehensive understanding of the disease process. These models simulate population dynamics, progressing from susceptibility to incubation, symptomatic or asymptomatic infection, and eventual recovery or mortality. They incorporate key epidemiological factors such as infectiousness, pathogenicity, and virulence (1,6,37–39). Additionally, they integrate practical prevention and control measures, including pharmacological interventions (e.g., antiviral drugs, antibiotics, vaccines) and non-pharmacological strategies (such as contact tracing, testing, school closures, hand hygiene, social distancing, mask-wearing), environmental disinfection, and vector control. These models are parameterized to reflect variations across different populations and adjust for interrelated parameters, ensuring alignment with real-world scenarios of disease transmission and control.

The models can be classified based on the object of study and parameter properties into two categories: 1) models with a group object and deterministic parameters, exemplified by the transmission dynamics model (1,4,6,40–45); 2) models with an individual object and stochastic parameters, illustrated by agent-based models (46–47), multiagent system (MAS) (48), cellular automata (CA) (49–50), and others.

Mechanism-driven models offer several key advantages: 1) They excel in uncovering the fundamental mechanisms of diseases and providing detailed explanations of epidemiological occurrences; 2) These models offer high flexibility in modeling. By employing a dynamics model that considers the natural history of the disease, clinical features, and epidemiological characteristics like the “three links” and “two factors,” supplemented by biological evidence, and utilizing mathematical tools such as differential equations and computer technology, these models can effectively simulate disease transmission processes. This leads to enhanced predictive reliability, aiding in public health decision-making and the formulation of effective intervention strategies; 3) Mechanism-driven modeling necessitates a multidisciplinary approach, integrating knowledge and techniques from epidemiology, biology, statistics, and computer science. This interdisciplinary collaboration results in a more holistic comprehension of disease complexity, facilitating the proposition of comprehensive solutions.

The model exhibits significant limitations: 1) It is highly sensitive to the initial values and relies heavily on subjective and empirical factors for parameter settings, with unclear significance of individual parameters in practical usage; 2) Its rigorous mathematical and theoretical derivation demands a deep understanding of mathematics and computer science, limiting its universal applicability in frontline systems.

When developing mechanism-driven models, researchers must consider key factors such as the research’s objective and the compatibility of the data with the selected model. Subsequently, selecting a suitable model to evaluate transmissibility, morbidity, and other anticipated outcome measures is vital. The precise construction and selection of mechanism-driven models are essential to accurately replicate real-world scenarios and achieve desired outcomes. Hence, the process of model building and assessment calculations should adhere to a standardized and reproducible methodology. As such, our research team has introduced the “MODELS” modeling framework, comprising six primary steps and 19 sub-steps in the modeling process (51).

The current mathematical model has limitations in practical application. Therefore, it is essential to validate its accuracy using real data and enhance it through adjustments. The model serves as a theoretical foundation and enhances data support for disease control. Here, personal opinions on its future development are discussed.

The “connotation” of theoretical epidemiology is expected to be further enhanced: The complexity and professional limitations of mathematical modeling have hindered extensive research and application of infectious disease mathematical modeling in Chinese public health. This is attributed to several factors: 1) Studies primarily focus on methodologies rather than practical implementation in public health; 2) Some research lacks clear parameter interpretation and comprehensive system dynamics; 3) Inconsistencies exist among experts regarding the natural history and transmission mechanisms of infectious diseases, for example, the percentage and transmissibility of asymptomatic infections are often overlooked.

The “outreach” of theoretical epidemiology is expected to be further expanded: The emergence of new infectious disease outbreaks present opportunities for the advancement of theoretical epidemiology, encompassing both theory and practical application. The theoretical aspect involves interdisciplinary collaboration, crucial for addressing complex public health issues by integrating various fields like infectious disease epidemiology, ecology, pathogen biology, genetics, applied mathematics, and computer science in the age of real-world and big data (52). The practical application aims to extend the reach of theoretical epidemiology from infectious diseases to areas such as chronic non-communicable diseases. This expansion is poised to stimulate the progress of public health strategies and aid in addressing global health challenges.

Theoretical epidemiology will play an important role in the training and pooling of human resources: Our research team surveyed the status of theoretical epidemiology instruction in academic and Center for Disease Control and Prevention (CDC) settings throughout China. Findings reveal a significant demand for theoretical epidemiology research and application within CDCs nationwide. However, there is a notable absence of theoretical epidemiology education in undergraduate programs across Chinese universities, hindering its practical application in epidemiology. Theoretical epidemiology is critical to understanding epidemic trends and transmission patterns of disease, and we therefore call for its inclusion in basic public health and preventive medicine curricula. This integration not only enhances the preparedness of future public health professionals but also reinforces the knowledge base within the discipline. By fostering theoretical epidemiology education, we aim to fortify the public health system to effectively combat infectious disease outbreaks and health emergencies.

No conflicts of interest.