Epidemiological Modeling: Lessons from the Spanish Flu to COVID-19

Introduction

Epidemiological modeling provides a quantitative framework for understanding pathogen transmission dynamics within host populations. While much of the public discourse surrounding pandemic modeling has focused on human pathogens, the underlying mathematical and computational principles are equally applicable to veterinary medicine. The Spanish influenza pandemic of 1918 and the COVID-19 pandemic offer distinct case studies from which veterinary epidemiologists can extract valuable lessons for modeling emerging and re-emerging animal diseases. This review examines the evolution of epidemiological modeling approaches from the era of the Spanish flu to the COVID-19 pandemic, with a focus on applications relevant to veterinary virology, diagnostics, and population-level disease control.

Historical Context: The Spanish Flu and Early Modeling

The Spanish influenza pandemic, caused by an H1N1 influenza A virus, remains one of the most devastating infectious disease events in recorded history. Brüssow [1] provided a comprehensive analysis of the beginning and ending of this respiratory viral pandemic, highlighting the epidemiological patterns that characterized its three distinct waves. The first wave in the spring of 1918 exhibited relatively low mortality, followed by a highly lethal second wave in the autumn of 1918, and a third wave in early 1919. This temporal structure, with waves of varying severity, is a hallmark of many respiratory pathogens and has direct parallels in veterinary diseases such as highly pathogenic avian influenza (HPAI) H5N1 in poultry and wild birds.

Early epidemiological models for the Spanish flu were necessarily simple, relying on basic case counts, mortality data, and crude estimates of transmission rates. The basic reproduction number (R0) for the 1918 pandemic has been retrospectively estimated using compartmental models, with values typically ranging from 1.4 to 2.8 depending on the population and modeling assumptions. These retrospective analyses, while limited by the quality of historical data, established foundational principles for understanding how respiratory viruses spread through susceptible populations.

Compartmental Models in Veterinary Epidemiology

The most widely used class of epidemiological models in veterinary medicine is the compartmental model, which partitions the host population into discrete states based on infection status. The classic susceptible-infected-recovered (SIR) model and its extensions (SEIR, SEIRS, SIS) form the backbone of most transmission dynamic analyses.

The SIR Framework

The SIR model divides the population into three compartments:

• Susceptible (S): individuals capable of acquiring infection
• Infected (I): individuals currently infectious
• Recovered (R): individuals who have cleared infection and are immune

The system is governed by ordinary differential equations:

dS/dt = -beta * S * I / N dI/dt = beta * S * I / N - gamma * I dR/dt = gamma * I

where beta is the transmission rate, gamma is the recovery rate, and N is the total population size. The basic reproduction number R0 = beta / gamma.

For veterinary applications, these models must be adapted to account for species-specific factors such as population density, management practices, and diagnostic test characteristics. For example, modeling the spread of Mycoplasma bovis in feedlot cattle requires incorporating the effects of commingling, stress-induced immunosuppression, and the chronic carrier state that characterizes this pathogen.

Extensions for Veterinary Pathogens

Veterinary epidemiological models often require additional compartments to capture relevant biological complexity. The SEIR model adds an exposed (E) compartment for individuals that are infected but not yet infectious, which is critical for pathogens with prolonged incubation periods such as Feline Coronavirus and FIP. The SEIRS model allows for waning immunity and reversion to susceptibility, which is relevant for pathogens like Canine Parvovirus variants where maternal antibody interference complicates vaccination strategies.

Table 1 summarizes common compartmental model structures and their veterinary applications.

Lessons from COVID-19 Modeling

The COVID-19 pandemic accelerated the development and application of epidemiological modeling tools, many of which have direct relevance to veterinary medicine. Key lessons include the importance of incorporating spatial heterogeneity, age-structured contact patterns, and diagnostic test performance into model frameworks.

Spatial and Metapopulation Models

COVID-19 modeling demonstrated that spatially explicit models, which account for movement between subpopulations, provide more accurate predictions than homogeneous mixing models. This principle is directly applicable to veterinary diseases such as African swine fever, where computational models for early detection and spread prediction in wild boar populations have become essential tools. Metapopulation models treat each farm or geographic region as a subpopulation connected by animal movement, wildlife corridors, or fomite transmission.

Diagnostic Test Characteristics in Models

A critical lesson from COVID-19 was the need to incorporate diagnostic test sensitivity and specificity into epidemiological models. In veterinary medicine, this is particularly important for pathogens where molecular and serological tests have different performance characteristics at different stages of infection. For example, modeling the spread of Feline Leukemia Virus progressive infection requires accounting for the window period between infection and detectable antigenemia, as well as the possibility of regressive infection where proviral DNA is present but p27 antigen is not detectable.

The integration of diagnostic data into models can be represented as a decision tree for surveillance design.

flowchart TD
    A[Population at Risk], > B{Surveillance Objective}
    B, >|Early Detection| C[Select Diagnostic Test]
    B, >|Prevalence Estimation| D[Select Sampling Strategy]
    C, > E{Test Characteristics}
    E, >|High Sensitivity| F[PCR-based assay]
    E, >|High Specificity| G[Serological confirmatory test]
    D, > H{Sample Type}
    H, >|Individual| I[Random sampling within herds]
    H, >|Pooled| J[Pooled PCR for cost efficiency]
    F, > K[Model Transmission Dynamics]
    G, > K
    I, > K
    J, > K
    K, > L[Estimate R0 and Intervention Impact]
    L, > M[Implement Control Measures]
    M, > N[Monitor with Repeated Testing]
    N, > O{Outbreak Controlled?}
    O, >|No| C
    O, >|Yes| P[Maintain Surveillance]

Parameter Estimation and Uncertainty

All epidemiological models require parameter estimates, which are derived from experimental studies, field observations, or literature review. Key parameters include the transmission rate (beta), recovery rate (gamma), incubation period, and case fatality rate. For veterinary pathogens, these parameters often vary by host species, age class, and management system.

Bayesian Approaches

Bayesian statistical methods have become increasingly important for parameter estimation in veterinary epidemiological models. These approaches allow for the incorporation of prior knowledge (e.g., from experimental challenge studies) and the quantification of uncertainty in model predictions. Bayesian networks, as probabilistic graph models, have been applied to veterinary biological inference and disease risk assessment.

Sensitivity Analysis

Robust epidemiological models require thorough sensitivity analysis to identify which parameters most strongly influence model outputs. For veterinary models, parameters related to diagnostic test performance, animal movement rates, and environmental persistence of pathogens often dominate uncertainty. For example, models of Escherichia coli transmission in poultry operations are highly sensitive to assumptions about environmental contamination and the efficacy of biosecurity protocols.

Modeling Specific Veterinary Pathogens

Highly Pathogenic Avian Influenza

The modeling of HPAI H5N1 in poultry and wild birds has benefited directly from lessons learned during both the Spanish flu and COVID-19 pandemics. Key modeling considerations include:

• The role of wild waterfowl as asymptomatic reservoirs
• The spatial structure of poultry production systems
• The effectiveness of stamping-out policies versus vaccination
• The impact of diagnostic test turnaround time on outbreak detection

Models for HPAI typically use stochastic compartmental frameworks with explicit spatial structure, incorporating data from surveillance maps and molecular epidemiology.

African Swine Fever

African swine fever modeling has advanced considerably, with computational models now used for early detection and spread prediction in wild boar populations. These models incorporate:

• Wild boar population density and home range size
• Carcass persistence and environmental transmission
• Human-mediated spread through fomites and contaminated feed
• Diagnostic test performance for both acute and chronic infections

Bovine Respiratory Disease Complex

Modeling bovine respiratory disease complex, which involves multiple viral and bacterial pathogens, requires multi-pathogen frameworks. Pathogens such as Bovine Coronavirus respiratory disease, Mycoplasma bovis, and Mannheimia haemolytica interact synergistically, and models must account for these interactions to predict disease dynamics accurately.

The Role of Molecular Diagnostics in Modeling

Modern epidemiological models increasingly incorporate molecular diagnostic data, including pathogen genome sequences, to infer transmission chains and estimate epidemiological parameters. This approach, known as phylodynamics, combines phylogenetic analysis with epidemiological modeling.

Genomic Epidemiology

Whole-genome sequencing of pathogens allows for the reconstruction of transmission networks with greater resolution than traditional epidemiological data alone. For veterinary applications, genomic epidemiology has been applied to:

• Tracking the introduction and spread of Lumpy Skin Disease Virus epidemiology
• Identifying sources of Salmonella contamination in poultry products
• Monitoring antimicrobial resistance gene dissemination in livestock-associated Staphylococcus aureus

Quantitative PCR and Viral Load Dynamics

Models that incorporate quantitative viral load data from PCR assays can provide more accurate estimates of transmission potential. For respiratory viruses, viral load kinetics influence the duration and magnitude of infectiousness. In veterinary settings, this approach has been used to model the transmission of West Nile Virus in horses and Canine Coronavirus variants.

Limitations and Challenges

Despite their utility, epidemiological models have important limitations that must be acknowledged. All models are simplifications of reality, and their predictions are only as reliable as the data and assumptions on which they are based. Key challenges in veterinary epidemiological modeling include:

• Incomplete or biased surveillance data
• Uncertainty about pathogen biology (e.g., duration of immunity, role of environmental transmission)
• Heterogeneity in host populations and management practices
• Difficulty in validating model predictions against real-world outcomes

The Spanish flu pandemic, as analyzed by Brüssow [1], illustrates the importance of understanding how pandemics begin and end. The factors that contributed to the eventual decline of the 1918 pandemic (herd immunity, viral evolution, public health interventions) are the same factors that must be incorporated into modern models for veterinary pathogens.

Future Directions

The integration of machine learning and artificial intelligence with traditional compartmental models represents a promising frontier for veterinary epidemiology. These hybrid approaches can identify complex patterns in large datasets (e.g., from automated diagnostic platforms, animal movement records, and environmental monitoring) and improve model predictions.

Additionally, the development of real-time modeling platforms that can ingest streaming diagnostic data and update predictions dynamically will enhance the utility of models for outbreak response. Point-of-care molecular diagnostics, such as those used for feline upper respiratory pathogens, can provide the rapid data inputs needed for such systems.

Conclusion

Epidemiological modeling has evolved substantially from the simple retrospective analyses applied to the Spanish flu to the sophisticated, multi-scale models used during the COVID-19 pandemic. Veterinary medicine has both contributed to and benefited from these advances. The lessons learned from these two pandemics underscore the importance of robust data collection, transparent model assumptions, and the integration of diagnostic test characteristics into modeling frameworks. As emerging infectious diseases continue to threaten animal populations and the human-animal interface, the continued refinement of epidemiological modeling approaches will remain essential for effective disease surveillance and control.

References

[1] Brüssow H. The beginning and ending of a respiratory viral pandemic-lessons from the Spanish flu. Microb Biotechnol. 2022. URL: https://pubmed.ncbi.nlm.nih.gov/35316560/