Mathematical Modeling of Infectious Diseases: Frameworks, Parameter Estimation, and Applications in Veterinary Epidemiology

Introduction

Mathematical modeling of infectious diseases provides a quantitative framework for understanding pathogen transmission dynamics, evaluating intervention strategies, and forecasting outbreak trajectories in animal populations. These models translate biological mechanisms of host-pathogen interactions into mathematical equations that can be analyzed analytically or simulated computationally. In veterinary medicine, modeling approaches inform decisions regarding vaccination campaigns, culling protocols, quarantine measures, and surveillance system design across livestock, companion animal, and wildlife populations.

The fundamental objective of infectious disease modeling is to capture the essential features of transmission processes while abstracting away unnecessary biological detail. This abstraction allows researchers to identify critical parameters governing outbreak dynamics, estimate the basic reproduction number (R0), and compare the relative effectiveness of different control measures. The field has evolved substantially from simple deterministic compartmental models to complex stochastic frameworks incorporating spatial heterogeneity, host behavior, and pathogen evolution.

Core Modeling Frameworks

Compartmental Models

The most widely used class of models in infectious disease epidemiology is the compartmental framework, in which individuals are classified according to their infection status. The classic susceptible-infected-recovered (SIR) model partitions a host population into three compartments: susceptible individuals (S) who can acquire infection, infected individuals (I) who are capable of transmitting the pathogen, and recovered individuals (R) who have gained immunity following infection [1]. The dynamics are governed by a system of ordinary differential equations:

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

where beta represents the transmission rate and gamma represents the recovery rate. The basic reproduction number R0 = beta / gamma defines the average number of secondary infections generated by a single infected individual in a fully susceptible population. When R0 exceeds 1, an epidemic can occur; when R0 is less than 1, the pathogen cannot sustain transmission.

Extensions to the basic SIR framework include the SEIR model, which incorporates an exposed (E) compartment representing individuals who are infected but not yet infectious. This latency period is critical for pathogens such as canine distemper virus and feline coronavirus, where viral shedding begins after a measurable incubation period. The SEIR model introduces an additional parameter sigma representing the rate at which exposed individuals become infectious.

Deterministic versus Stochastic Formulations

Deterministic compartmental models assume that population sizes are large enough that random fluctuations are negligible. However, in veterinary contexts involving small herds, endangered wildlife populations, or early outbreak phases, stochastic effects can dominate dynamics [2]. Stochastic models treat transitions between compartments as probabilistic events, typically using continuous-time Markov chains or discrete-time branching processes.

Djuikem and Arino [2] demonstrated that stress-induced changes in host susceptibility can fundamentally alter infection dynamics in both deterministic and stochastic frameworks. Their analysis showed that stochastic models capture extinction probabilities and fade-out dynamics that deterministic models cannot represent, particularly when the effective population size is small. This distinction is critical for modeling pathogens in isolated livestock operations or fragmented wildlife habitats.

Fractional Order Models

Recent advances have introduced fractional calculus into epidemiological modeling. Chauhan et al. [3] developed a fractional SIR model for childhood disease transmission with vaccination, demonstrating that fractional order derivatives capture memory effects and non-local dynamics that integer-order models cannot represent. The fractional parameter introduces a degree of freedom that improves model fit to empirical incidence data, particularly for diseases with long-term immunity waning or heterogeneous exposure patterns.

Parameter Estimation and Model Identifiability

Structural Identifiability

Before a model can be calibrated to data, it must be determined whether the model parameters can be uniquely estimated from the available observations. Structural identifiability analysis examines whether different parameter sets can produce identical model outputs. Liyanage et al. [4] provided a comprehensive tutorial on structural identifiability of epidemic models using the StructuralIdentifiability.jl software package. Their work demonstrated that many commonly used epidemiological models have parameters that are not structurally identifiable, meaning that multiple parameter combinations yield identical model predictions. This finding has profound implications for interpreting model outputs and designing data collection protocols.

Bayesian Inference and Uncertainty Quantification

Bayesian methods provide a principled framework for estimating model parameters from data while quantifying uncertainty. Warne et al. [5] developed a Bayesian uncertainty quantification approach to identify population-level vaccine hesitancy behaviors. Their methodology incorporated prior knowledge about vaccine uptake patterns and used Markov chain Monte Carlo sampling to generate posterior distributions for behavioral parameters. This approach is directly applicable to veterinary vaccination campaigns where owner compliance and herd-level uptake rates are uncertain.

Bayesian functional hierarchical models offer additional flexibility for spatially correlated disease outcomes. Ma et al. [6] introduced a spatially correlated analysis framework using Bayesian functional hierarchical models that can accommodate complex spatial dependencies in disease incidence data. This approach is particularly valuable for analyzing surveillance data from multiple geographic regions, such as county-level reporting of bovine tuberculosis or state-level monitoring of highly pathogenic avian influenza.

Network and Spatial Models

Contact Network Epidemiology

Compartmental models assume homogeneous mixing, meaning that every individual has equal probability of contacting every other individual. This assumption is violated in most real populations, where contact patterns are structured by social relationships, spatial proximity, and management practices. Network models represent individuals as nodes and contacts as edges, with infection spreading along these edges according to transmission probabilities.

Li et al. [7] analyzed the time evolution of infection state correlation of nodes in an SIR model on a random contact network. Their work revealed that the correlation structure of infection states evolves dynamically during an outbreak, with implications for targeted surveillance and intervention strategies. In veterinary contexts, network models have been applied to livestock movement networks, where animal transport between farms creates pathways for pathogen spread.

Shedding-Weighted Network Approaches

Barroso et al. [8] developed shedding-weighted network approaches for understanding tuberculosis maintenance in multihost systems using camera trap data. Their methodology assigned weights to network edges based on pathogen shedding rates from different host species, allowing the identification of key reservoir hosts and transmission bottlenecks. This approach is directly relevant to wildlife-livestock interfaces where multiple species contribute to pathogen maintenance, such as bovine tuberculosis transmission between badgers and cattle or African swine fever transmission between wild boar and domestic pigs.

Spatial and Environmental Drivers

Environmental factors including temperature, humidity, and precipitation profoundly influence vector-borne disease transmission. Huxley et al. [9] demonstrated that relative humidity systematically shifts juvenile thermal performance and projected population growth in a malaria vector, challenging the common practice of using temperature alone to predict vector population dynamics. Their findings have direct implications for modeling Culicoides-borne bluetongue virus in livestock and Aedes-borne arboviruses in poultry operations.

Schwaiger et al. [10] developed an agent-based model for temperature and precipitation-dependent population dynamics of Aedes vexans, a mosquito vector for several arboviruses affecting livestock and wildlife. Their model incorporated detailed life stage-specific responses to environmental conditions, allowing prediction of vector abundance peaks and associated transmission risk windows.

Advanced Modeling Applications

Multihost and Zoonotic Systems

Many veterinary pathogens circulate among multiple host species, complicating transmission dynamics and control efforts. Fraser et al. [11] developed a mathematical modeling study of yellow fever outbreak potential in Djibouti, Somalia, and Yemen, demonstrating how modeling can assess invasion risk in naive populations. Their framework incorporated vector population dynamics, human movement patterns, and livestock density to estimate R0 and outbreak probability.

Huang et al. [12] applied transmission and cost-effectiveness modeling to estimate progress towards elimination and future strategy optimization for gambiense human African trypanosomiasis in Uganda. Their model incorporated vector control, active surveillance, and treatment components, demonstrating how modeling can guide resource allocation for neglected zoonotic diseases.

Vaccine Impact Modeling

Mathematical models are essential tools for evaluating vaccine effectiveness and optimizing immunization strategies. Gazeau et al. [13] used mathematical modeling to highlight the crucial role of early childhood immunization in preventing congenital cytomegalovirus in countries with high seroprevalence. While this study focused on human populations, the methodological framework is directly transferable to veterinary vaccination programs, including feline leukemia virus vaccination in multi-cat households and bovine respiratory syncytial virus vaccination in feedlot cattle.

Naffeti et al. [14] projected the public health benefits of a hypothetical HSV-1 vaccine using mathematical modeling analysis. Their approach demonstrated how modeling can inform vaccine development priorities by quantifying potential population-level impacts before clinical trial completion. Similar frameworks can guide investment decisions for veterinary vaccine development, particularly for pathogens with significant economic impact.

Cost-Effectiveness and Resource Allocation

Economic modeling integrated with transmission dynamics provides decision support for resource allocation. Liu et al. [15] conducted a mathematical modeling study comparing the impacts and cost-effectiveness of tuberculosis systematic screening strategies in prisons in Brazil, Colombia, and Peru. Their framework combined transmission modeling with health economic evaluation to identify optimal screening frequencies and diagnostic algorithms. This approach is directly applicable to veterinary settings such as bovine tuberculosis surveillance in dairy herds or porcine reproductive and respiratory syndrome monitoring in swine operations.

Ward et al. [16] examined whether real-world scaling up of hepatitis C treatment in people who inject drugs could be cost-saving using an economic modeling study. Their analysis demonstrated that treatment as prevention can be cost-effective when transmission dynamics are considered, a principle that applies to veterinary pathogens such as feline immunodeficiency virus in cat colonies or equine influenza in training facilities.

Behavioral and Feedback Models

Behavioral Feedback and Resource Accessibility

Host behavior influences disease transmission, and disease prevalence in turn influences host behavior. Zhang et al. [17] developed a behavioral feedback model driven by resource accessibility under static-dynamic optimal control. Their framework captured the bidirectional relationship between infection risk and behavioral responses, including vaccination uptake, social distancing, and movement restrictions. In veterinary contexts, behavioral feedback models can represent farmer responses to disease outbreaks, including changes in biosecurity practices, testing frequency, and culling decisions.

Memory Effects and Global Dynamics

Wu et al. [1] investigated how memory effects govern global dynamics thresholds in the SIR model. Their analysis demonstrated that incorporating memory of past infection states into transmission and recovery rates fundamentally alters the conditions for epidemic emergence and persistence. Memory effects are particularly relevant for pathogens with long-term immunity or chronic carrier states, such as feline leukemia virus or bovine leukemia virus.

Model Validation and Comparison

Scenario Analysis and Outbreak Simulation

Scenario analysis provides a structured approach to exploring possible outbreak trajectories under different assumptions about transmission parameters, intervention effectiveness, and host population dynamics. Marziano et al. [18] conducted scenario analysis for potential community spread of Andes virus, a rodent-borne hantavirus. Their approach systematically varied key parameters including transmission rate, incubation period, and case fatality ratio to generate a range of possible outcomes. This methodology is directly applicable to emerging veterinary pathogens where parameter uncertainty is high.

Comparative Modeling Studies

Comparative modeling studies evaluate multiple modeling approaches applied to the same epidemiological question. Li et al. [19] conducted a review of epidemiological modeling studies on monkeypox, comparing compartmental, network, and agent-based approaches. Their analysis identified strengths and limitations of each framework for different research questions, providing guidance for model selection in future studies.

Computational Implementation

Software and Algorithmic Considerations

Model implementation requires careful consideration of numerical methods, computational efficiency, and reproducibility. Liyanage et al. [4] demonstrated the use of StructuralIdentifiability.jl for identifiability analysis, while other researchers have employed R, Python, and MATLAB for simulation and inference. Agent-based models such as those developed by Schwaiger et al. [10] require substantial computational resources for large populations but offer flexibility in representing individual-level heterogeneity.

Model Calibration to Empirical Data

Model calibration involves estimating parameters from observed data, typically using maximum likelihood or Bayesian methods. Feng et al. [20] modeled the risk of West Nile virus infection in seven European countries from published serological and case notification data. Their approach demonstrated how to combine multiple data sources, including seroprevalence surveys, case reports, and vector surveillance, to estimate infection risk and identify high-risk areas.

Limitations and Future Directions

Model Assumptions and Their Consequences

All models make simplifying assumptions that limit their applicability. Common assumptions in veterinary disease models include homogeneous mixing, constant population size, and fixed transmission rates. Violations of these assumptions can lead to biased parameter estimates and inaccurate predictions. Sensitivity analysis and model validation against independent data are essential for assessing model robustness.

Integration with Molecular Epidemiology

The integration of mathematical modeling with molecular epidemiology, including pathogen genomic data, represents a promising frontier. Phylodynamic approaches combine phylogenetic inference with transmission modeling to estimate key epidemiological parameters from genetic sequence data. These methods are particularly valuable for pathogens with limited surveillance data or where contact tracing is impractical.

One Health Modeling Frameworks

One Health approaches that integrate human, animal, and environmental health data require modeling frameworks capable of representing cross-species transmission and shared risk factors. The livestock zoonoses interface, including pathogens such as highly pathogenic avian influenza H5N1 and Nipah virus, requires models that capture transmission dynamics across domestic animal, wildlife, and human populations.

Conclusion

Mathematical modeling of infectious diseases provides essential tools for understanding transmission dynamics, evaluating intervention strategies, and guiding surveillance efforts in veterinary populations. From simple compartmental models to complex agent-based frameworks, these approaches translate biological mechanisms into quantitative predictions that inform decision-making. Continued methodological advances in parameter estimation, spatial modeling, and behavioral feedback will further enhance the utility of these tools for veterinary epidemiology and disease control.

References

[1] Wu Z, Cai Y, Wang Z, et al. Reshaping transmission and recovery: How memory effects govern global dynamics thresholds in the SIR model. Math Biosci. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42203106/

[2] Djuikem C, Arino J. Deterministic and Stochastic Infection Dynamics in a Population Subject to Stress. Bull Math Biol. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42249193/

[3] Chauhan JP, Jebran S, Khirsariya SR. Stability analysis and numerical investigation of fractional SIR model for childhood disease transmission with vaccination. Sci Rep. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42177245/

[4] Liyanage YR, Saucedo O, Tuncer N, et al. A Tutorial on Structural Identifiability of Epidemic Models Using StructuralIdentifiability.jl. Bull Math Biol. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42234202/

[5] Warne DJ, Varghese A, Brewster A, et al. Bayesian uncertainty quantification to identify population level vaccine hesitancy behaviours. PLoS One. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42189916/

[6] Ma S, Yu K, Pan J, et al. Spatially Correlated Analysis of Infectious Disease Outcomes Based on Bayesian Functional Hierarchical Models. Stat Med. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42271131/

[7] Li M, Shi H, Ma J, et al. Time evolution of infection state correlation of nodes in an SIR model on a random contact network. Math Biosci. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42173237/

[8] Barroso P, Silk MJ, Perelló A, et al. Shedding-weighted network approaches for understanding tuberculosis maintenance in multihost systems using camera traps. Ecol Appl. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42244255/

[9] Huxley PJ, Brown JJ, St Laurent B, et al. Beyond Temperature: Relative Humidity Systematically Shifts Juvenile Thermal Performance and Projected Population Growth in a Malaria Vector. Ecol Lett. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42281274/

[10] Schwaiger A, Schweitzer M, Hackl WO, et al. Agent-Based Model for Temperature and Precipitation-Dependent Population Dynamics of Aedes vexans. Stud Health Technol Inform. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42175388/

[11] Fraser K, Cibrelus L, Horton J, et al. Yellow fever outbreak potential in Djibouti, Somalia and Yemen: a mathematical modelling study. BMC Glob Public Health. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42260694/

[12] Huang CI, Antillon M, Crump RE, et al. Transmission and cost-effectiveness modelling to estimate the progress towards elimination and future strategy optimisation for gambiense human African trypanosomiasis in Uganda. BMC Infect Dis. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42251250/

[13] Gazeau S, Byrne C, Coombs D, et al. Mathematical modelling highlights the crucial role of early childhood immunization in preventing congenital CMV in countries with high CMV seroprevalence. Vaccine. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42269282/

[14] Naffeti BS, Ayoub HH, Johnston C, et al. Projected public health benefits of a hypothetical HSV-1 vaccine in the United States: A mathematical modeling analysis. Sci Adv. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42234759/

[15] Liu YE, Bampi JVB, Arthur RF, et al. Comparative impacts and cost-effectiveness of tuberculosis systematic screening strategies in prisons in Brazil, Colombia, and Peru: A mathematical modeling study. PLoS Med. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42241460/

[16] Ward Z, Fraser H, Pherson SM, et al. Can real-world scaling up HCV treatment in people who inject drugs be cost-saving: an economic modelling study in the UK. Int J Drug Policy. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42217525/

[17] Zhang Y, Liu H, Tang S, et al. Sustainable Coexistence with Infectious Diseases: A Behavioral Feedback Model Driven by Resource Accessibility under Static-Dynamic Optimal Control. Bull Math Biol. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42234344/

[18] Marziano V, De Bellis A, Del Manso M, et al. Scenario analysis for potential community spread of Andes virus (ANDV). Euro Surveill. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42246066/

[19] Li J, Luo Q, Beltran E, et al. A Review of Epidemiological Modeling Studies on Monkeypox. China CDC Wkly. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42180661/

[20] Feng H, Marini G, Barabás É, et al. Modelling the risk of West Nile virus infection in seven European countries from published serological and case notification data, 2008 to 2022. Euro Surveill. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42246068/

[21] Park SJ, Jeong SG, Choi MY. Infection dynamics for fluctuating infection or removal rates regarding the number of infected and susceptible individuals. Infect Dis Model. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42222650/

[22] Ramadan S, Salman S, El-Sayed A. A modified HIV model with Beddington-DeAngelis incidence and cure rate. Sci Rep. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42191796/

[23] Lisowski B, V Rezelj V, Vignuzzi M, et al. In Vitro Analysis and Dynamic Modeling of SARS-CoV-2 Infection Inhibition by Sigma-1 Receptor Antagonist PB28. Bull Math Biol. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42176142/