Spatial Epidemiology in Veterinary Science: Mapping Disease Clusters

By Dr. Zubair Khalid, DVM, MS, PhD · 2026-08-11

Key Takeaways

• Spatial epidemiology in veterinary science quantifies disease distribution across geographic space to identify areas of concentrated occurrence (clusters) and investigate underlying environmental, demographic, or management drivers.
• Data structures are critical, with point data (individual animal locations) offering high resolution but potential confidentiality issues, and aggregate data (counts per administrative unit) protecting privacy but introducing the modifiable areal unit problem and scale dependence.
• Disease mapping employs smoothing methods, such as empirical Bayes, to stabilize rate estimates in sparsely populated areas and generate risk maps that guide targeted surveillance and intervention strategies, for example, identifying rainfall as a key predictor for liver fluke in dairy herds.
• Cluster detection methods, including global tests like Moran's I and local methods like the spatial scan statistic, statistically evaluate whether observed disease patterns exceed random expectation, with the spatial scan statistic being widely used for its ability to account for multiple testing and identify specific high-risk locations.
• Spatial regression models are essential for identifying risk factors by accounting for spatial autocorrelation in residuals, which is common due to shared environmental exposures or transmission pathways, and can reveal spatially varying covariate effects, such as deforestation and livestock density influencing zoonotic disease clusters.
• Reproducibility and defensibility require meticulous documentation of data sources, cleaning steps, spatial units, coordinate reference systems, and software versions, aligning with WOAH terrestrial animal health standards for surveillance and reporting.

Spatial epidemiology examines how disease occurrence distributes across geographic space and what environmental, demographic, or management factors explain that distribution. For veterinary researchers, the discipline provides a formal framework for asking where disease concentrates, why it concentrates there, and whether observed patterns exceed what chance would produce. This article introduces the conceptual foundations and core methods of spatial epidemiology applied to animal health data, with emphasis on disease mapping and cluster detection. It serves researchers designing observational studies, analyzing surveillance data, or evaluating control programs that operate across defined geographic areas. The methods described here apply across species, from companion animal populations to livestock production systems to wildlife reservoirs.

The relevance of spatial thinking in veterinary medicine has grown with the recognition that many pathogens move through landscapes in predictable ways. Directly and indirectly transmitted infectious diseases are spatial-related, because transmission depends on distances between susceptible hosts and the environments they share, including contaminated materials and infectious animal species. Emerging viruses, most of which originate from animal hosts, frequently appear at the interface between wildlife, domestic animals, and human populations, and their emergence is difficult to predict without spatial context. Improvements in computing performance and statistical approaches have expanded the options for visualizing and analyzing disease spatial data, making these methods accessible to researchers beyond specialist geographers.

This article covers the data structures used in veterinary spatial analysis, the principles of disease mapping, and the main approaches to cluster detection. It distinguishes point data, representing individual animal locations, from aggregate data, representing counts or rates summarized over administrative or ecological units. The article does not address spatio-temporal modeling, which extends these methods to incorporate time explicitly. Later sections in this series address Bayesian hierarchical models and specific cluster detection algorithms in greater depth.

At a Glance

Spatial Data Structures in Veterinary Epidemiology

Spatial data in veterinary research take two fundamental forms. Point data record the geographic coordinates of individual events, such as the location of a farm with a positive test result or the GPS position of a wildlife carcass. Aggregate data summarize events within predefined spatial units, such as postal areas, counties, or veterinary districts. The choice between these structures shapes every subsequent analytical decision.

Point data preserve the full spatial resolution of the original observations but raise privacy and confidentiality concerns, particularly for commercial livestock operations. Aggregate data protect confidentiality and align naturally with administrative reporting systems, but they impose the boundaries of the reporting unit on the analysis. A disease process that crosses administrative borders becomes fragmented, and associations detected at one scale may not hold at another. This scale dependence, known as the modifiable areal unit problem, requires researchers to test whether conclusions remain stable across alternative boundary definitions.

Disease Mapping Principles

Disease mapping produces a visual representation of disease frequency across space, typically as a map of rates or risks. The purpose is descriptive: to show where disease burden concentrates and to generate hypotheses about underlying drivers. Maps of crude rates, calculated as cases divided by population at risk within each unit, are straightforward but misleading when population sizes vary widely across units. Small populations produce unstable rate estimates, and extreme values often appear in sparsely populated areas.

Smoothing methods address this instability by borrowing information from neighboring units. Empirical Bayes approaches shrink extreme estimates toward the overall mean, with the degree of shrinkage inversely proportional to the population at risk. Fully model-based approaches, described in the companion article on Bayesian hierarchical models, treat the true underlying risk as a random process and estimate it with associated uncertainty. The output is a map of smoothed risk with credible intervals, allowing readers to distinguish areas of genuinely elevated risk from areas where estimates are simply imprecise.

Risk maps have direct application in veterinary control programs. Spatial models developed with geographic information systems can generate risk maps for production diseases, supporting the formulation of control programs targeted to high-risk areas. For example, linear regression models using climatic, environmental, soil, livestock, and pasture variables have explained over 70% of the variation in exposure to liver fluke in dairy herds, with rainfall as a consistent predictor across model specifications. Such maps identify where surveillance should be intensified and where preventive interventions will yield the greatest return.

Cluster Detection Methods

Cluster detection asks whether cases concentrate in space more than expected under a null hypothesis of random distribution. Methods divide into two broad families: global clustering tests and local cluster detection.

Global clustering tests assess whether clustering exists anywhere in the study region, producing a single summary statistic. They do not identify where clusters are located. Examples include Moran's I, which measures spatial autocorrelation of case counts or rates, and the Tango and Besag-Newell tests, which evaluate the overall tendency toward clustering. These tests are useful for screening, but their lack of localization limits their practical value for control programs.

Local cluster detection methods identify specific locations where risk is elevated. The spatial scan statistic, implemented in SaTScan and related software, imposes circular or elliptic windows of varying size across the study region and compares the observed case count inside each window with the expected count under the null hypothesis. Windows with significantly elevated risk are reported as clusters, with likelihood ratio statistics and Monte Carlo p-values. Kulldorff's spatial scan statistic remains the most widely used approach in veterinary applications because it accounts for the multiple testing inherent in scanning many potential windows.

Applications in veterinary research illustrate the value of these methods. A spatial clustering analysis of more than 65,000 cases of cutaneous leishmaniasis in Colombia identified seven statistically significant clusters distributed across all major biomes. Deforestation and livestock production were significantly associated with cluster membership, along with urbanization, travel time to major cities, and proximity to water bodies. The analysis linked human environmental disturbance to the spatial structure of a zoonotic disease epidemic, demonstrating how cluster detection can identify modifiable drivers of transmission.

Spatial Regression and Risk Factor Identification

Once clusters are identified, the next question is what explains them. Spatial regression methods extend conventional regression to account for spatial dependence among observations. Standard regression assumes independence of residuals, but disease outcomes in nearby locations are often correlated because they share environmental exposures, host populations, or transmission pathways. Ignoring this autocorrelation produces biased standard errors and inflated significance tests.

Spatial regression approaches include simultaneous autoregressive models, conditional autoregressive models, and geographically weighted regression. The choice depends on whether the research question concerns global relationships, where one set of coefficients applies across the entire study region, or local relationships, where the effect of a covariate varies across space. Geographically weighted regression fits local models at each location, producing maps of covariate effects that reveal spatial heterogeneity in disease drivers.

The integration of spatial epidemiology with other analytical traditions strengthens inference. Combining phylogeographic reconstruction with spatial regression has identified predictors of viral diffusion for highly pathogenic avian influenza, including host densities, elevation, and meteorological variables. Such integrated approaches recognize that the spatial pattern of disease reflects both ecological drivers and the evolutionary history of the pathogen, and that neither perspective alone fully explains observed distributions.

Data Quality and Reporting Considerations

Spatial analysis is only as reliable as the data feeding it. Surveillance systems vary in completeness, diagnostic sensitivity, and reporting timeliness across regions and species. The World Organization for Animal Health maintains international standards for animal health surveillance and notification, and researchers should align their data collection and reporting with these frameworks. Where surveillance is passive, underreporting may correlate with access to veterinary services, creating spatial bias that mimics true disease clustering.

Ethical considerations also apply. Point data on individual farms or herds can identify premises with disease, with potential economic consequences for owners. Researchers must balance the scientific value of precise locations against confidentiality obligations, and should consider aggregation or spatial perturbation where re-identification risk is high. The same principles that govern human health data, including the need for ethics review and data protection, apply to veterinary spatial data, particularly when zoonotic pathogens are involved.

Applied Spatial Analysis Workflow

A structured workflow reduces the risk of method misapplication and improves reproducibility. The sequence below assumes the analyst has already defined the study question and assembled the case and population data. Each stage produces an output that feeds the next, and each has distinct failure modes.

Step 1: Define the Spatial Unit and Reference Population

Choose the spatial unit before any analysis. Point data, such as individual animal locations or farm coordinates, preserve maximum information but raise confidentiality concerns and often require address geocoding. Aggregate data, such as counts per postcode, census tract, or veterinary practice catchment, are more readily available but introduce the modifiable areal unit problem, where results change with the choice of boundary. The reference population must match the spatial unit. For livestock, this means the number of animals at risk per unit, not the human population or the number of holdings, unless the research question concerns holding-level incidence.

Species and production system change this decision. For companion animals, the reference population is often the number of registered patients or estimated pet ownership per area, both of which are frequently unavailable at fine spatial resolution. For wildlife, the reference population may be estimated from habitat suitability models or census counts, and the uncertainty in these estimates must be carried into the analysis. For production livestock, movement records and holding registers provide denominators, but these registers may be incomplete in some regions, and the World Organization for Animal Health surveillance standards emphasize that reporting completeness directly affects the interpretability of any spatial summary.

Step 2: Data Cleaning and Georeferencing

Assign coordinates to each record. Global positioning system readings from farm gates or pasture access points are usually accurate to a few metres. Address geocoding introduces positional error that varies with the quality of the street network and the rural address density. In rural areas, a postcode or postal address may locate a holding to within several kilometres, which can exceed the scale of the disease process under study. For example, airborne dispersal of Coxiella burnetii from ruminant holdings operates over distances of up to 18 km under gale force winds, but the highest infection risk in rural areas occurs within 5 km of the source. A geocoding error of 2 km will not change the broad conclusion in that setting, but it would invalidate a fine-scale analysis of within-herd transmission.

Deduplicate records by animal identifier and date. Check for coordinate swaps, records placed in water bodies, and implausible distances between consecutive locations for the same animal. Record the cleaning steps in a log file. The CDC principles of epidemiology describe standard data quality checks that apply directly to spatial datasets, including range checks, consistency checks, and cross-tabulation of variables that should agree.

Step 3: Exploratory Mapping and Visualization

Produce dot maps of cases, choropleth maps of incidence rates, and smoothed maps before any formal cluster detection. Dot maps show the raw distribution but are misleading when the population density varies, because areas with more animals simply show more dots. Choropleth maps of raw rates are unstable for small populations, as a single case in a sparsely populated unit produces a very high rate. Smoothed maps, using empirical Bayes or kernel smoothing, stabilize rates by borrowing information from neighbouring units. The review of spatial approaches for infectious disease management identifies visualization as the necessary first step before formal clustering or regression, because it reveals data errors, missing regions, and the general spatial pattern that later methods should respect.

Step 4: Cluster Detection and Hypothesis Generation

Select the cluster detection method based on the data type and the question. The table below summarizes the main choices.

The spatial scan statistic is the most widely used for veterinary outbreak investigation because it adjusts for the multiple testing inherent in scanning many possible windows and reports a likelihood ratio for each cluster. It requires a case count and a population count for each unit, or a point process model for individual-level data. When population data are missing or unreliable, the scan statistic cannot be applied, and the analyst must either obtain a denominator or switch to a method that models the case distribution directly.

Interpret clusters as hypothesis-generating, not etiologic. A cluster of fasciolosis in dairy herds, for example, may reflect rainfall, soil type, or pasture management, and the linear regression models developed for Fasciola hepatica in England and Wales explained over 70% of the variation in exposure prevalence using climatic and environmental predictors. The cluster identifies where to look, not why the disease is there.

Step 5: Risk Factor Modeling

Formal regression follows cluster detection. The outcome is the case count or rate per spatial unit, and the predictors are environmental, climatic, demographic, or management variables. Spatial regression methods that adjust for neighbour effects or spatial heterogeneity are preferred over ordinary regression, because residuals from ordinary models are usually spatially autocorrelated, violating the independence assumption. The spatial epidemiology of cutaneous leishmaniasis in Colombia illustrates the value of linking identified clusters to socioeconomic and demographic covariates, finding that deforestation, livestock density, and urbanisation were associated with the spatial distribution of cases.

Model selection should be guided by the causal framework, not purely by stepwise algorithms. Include variables that plausibly affect transmission, such as vector habitat, host density, or climate, and report the spatial autocorrelation of the residuals after fitting. A model that leaves strong residual autocorrelation has missed a spatial process and should not be interpreted as complete.

Software Tools for Spatial Analysis

The choice of software depends on the analyst's programming skill, the data volume, and the need for reproducibility. The following tools are in common use in veterinary spatial epidemiology.

For a veterinary researcher starting a spatial analysis, R with the sf package for vector data and the spatstat package for point patterns provides the best balance of flexibility and reproducibility. SaTScan is the recommended tool for formal cluster detection because it implements the scan statistic correctly and is freely available. QGIS is sufficient for data cleaning, visualization, and map production when statistical modeling is not required.

The correct choice changes with the task. A practitioner investigating a single farm outbreak needs only QGIS to plot cases and identify a possible source. A research group analyzing national surveillance data needs R or Python for reproducible modeling and SaTScan for cluster detection. A laboratory without programming support may rely on ArcGIS with its spatial statistics toolbox. The MSD Veterinary Manual and the American Veterinary Medical Association practice resources provide additional guidance on data management and professional standards, though neither offers specific software recommendations.

Documentation and Reporting

Every spatial analysis should be documented to the standard expected of a laboratory procedure. Record the data sources, the cleaning steps, the spatial unit, the coordinate reference system, the software versions, and the parameter choices for each method. The coordinate reference system matters more than is commonly appreciated: a projected coordinate system in metres is required for distance-based methods such as the scan statistic and kernel density estimation, while a geographic coordinate system in degrees will produce incorrect distances. State the CRS explicitly in any report.

Report cluster statistics with their p-values or likelihood ratios, the cluster radius, and the observed and expected case counts. For regression models, report the coefficients, confidence intervals, and the spatial autocorrelation of residuals. Maps should include a scale bar, north arrow, and legend, and should use a color scheme that is perceptible to color-blind readers.

The international standards for animal health surveillance require that surveillance outputs, including spatial analyzes, be documented in a way that supports trade decisions and disease control planning. A spatial analysis that cannot be reproduced from its documentation has limited value beyond the immediate study.

Recognized Complications and Failure Modes

Spatial analyzes fail in characteriztic ways, and most failures trace to decisions made before any statistical test is run. The most common complication is the ecological fallacy, where associations observed at the aggregate level are attributed to individuals within those aggregates. A herd-level risk factor for fasciolosis identified in a postcode-level regression, for example, does not establish that every animal in a high-risk postcode is exposed. The discriminating check is to compare the spatial unit of analysis against the unit of the biological process under study. When they differ, the analysis should be framed explicitly as generating hypotheses about group-level processes, not individual risk.

A second failure mode is the modifiable areal unit problem. Changing the boundaries of the spatial units, for instance moving from county to postcode or from grid cells to watersheds, can change both the magnitude and the direction of observed associations. Detection of this problem requires running the same analysis on at least two alternative spatial aggregations and comparing the stability of the results. If the conclusions shift materially, the choice of boundary must be treated as a study limitation instead of a fixed feature of the data.

Edge effects arise when cases near the study boundary have fewer neighbours than cases in the interior, biasing cluster statistics. The standard corrective is to analyze a buffer zone around the study area or to use methods that explicitly account for boundary geometry. A related problem is the small number problem, where sparse counts in small areas produce unstable rates. Empirical Bayes smoothing or related shrinkage estimators are the usual remedy, and their use should be reported alongside the raw counts.

Common Errors and Corrective Actions

Less experienced analysts frequently confuse incidence with prevalence when interpreting mapped data. A map of seroprevalence reflects cumulative exposure, not current transmission, and the two can have very different spatial patterns. The corrective is to state the measure being mapped in the figure title and in the methods, and to choose the measure that matches the question being asked.

A second common error is treating the absence of reported cases as the absence of disease. Surveillance gaps, diagnostic submission patterns, and reporting delays all shape the observed spatial distribution. The CDC principles of epidemiology in public health practice emphasize that surveillance data reflect the system that produced them as much as the disease itself. The corrective is to map the underlying population at risk and, where possible, the intensity of surveillance effort alongside the case map.

A third error is overinterpreting the output of a single cluster detection method. Different methods have different null hypotheses and different sensitivities to the shape and size of clusters. The corrective is to run at least two complementary methods and to report concordance. A cluster identified by only one method should be treated as provisional.

Limitations of the Current Evidence

The evidence base for spatial methods in veterinary medicine is uneven. Much of the methodological literature derives from human infectious disease epidemiology, and the transfer to animal populations is not always direct. Livestock populations are managed, moved, and culled in ways that human populations are not, and these management decisions can dominate the spatial signal. The review of spatial approaches in infectious disease management notes that spatial dimensions include shared environments and infectious animal species, but the relative contribution of each varies by production system and pathogen.

Expert opinion still differs on several points. The appropriate distance for exclusion zones around infected holdings is one such point. The systematic review of Q fever dispersal found that infection risk in rural areas is highest within 5 km of sources but that Coxiella burnetii can travel up to 18 km on gale force winds, which suggests that fixed-radius zones may be inadequate in some conditions. Whether to use contemporary weather data or long-term averages in risk mapping is another area of disagreement. The linear regression work on Fasciola hepatica in dairy herds found that models using five-year average data performed only slightly better than models using contemporary data, with rainfall a consistent predictor in all models. The choice may therefore depend on data availability more than on model performance.

The integration of phylogeographic and spatial methods is promising but methodologically demanding. The H5N1 diffusion study in Egypt demonstrated that combining phylogeography with spatial epidemiology can identify supported predictors of viral spread, but the computational burden and the need for paired genetic and spatial data limit routine application.

Referral, Consultation, and Regulatory Reporting

Spatial analyzes that inform disease control decisions should not be conducted in isolation. When a cluster is detected in a notifiable disease, the responsible authority must be informed according to the reporting frameworks set out in the WOAH terrestrial animal health standards. The WOAH animal health surveillance standards describe the international notification obligations, and national authorities will specify the local reporting pathway.

Specialist consultation is warranted when the analysis requires methods beyond the analyst's training, when the data structure is complex, or when the results will be used in legal or regulatory proceedings. A veterinary epidemiologist with spatial training should review the analysis before it informs control policy. Laboratory involvement is required when case confirmation depends on diagnostic testing, and the spatial analysis should incorporate the laboratory's case definition and testing protocols.

The threshold for regulatory reporting is not a statistical one. A cluster that is statistically significant but involves a non-notifiable disease may require no regulatory action, while a single case of a notifiable disease triggers reporting regardless of any spatial analysis. The analyst should establish the notifiable status of the disease before beginning the analysis, not after a cluster is found. Where the evidence base is contested, as with exclusion zone distances, the analysis should present the range of plausible interpretations and flag the uncertainty for decision makers.

Frequently Asked Questions

How much does a spatial epidemiology study cost, and what resources are actually required?

Costs scale with the spatial unit and data source. Aggregate data from national surveillance systems, such as postcode-level exposure records, are inexpensive to obtain and can be analyzed with open-source software on standard hardware. Individual-level data with high-resolution georeferencing, field validation, and repeated sampling are substantially more expensive. The largest recurring cost is staff time for data cleaning and georeferencing, not software licensing. If commercial GIS licences are unaffordable, open-source packages provide equivalent functionality for cluster detection and regression. Veterinary diagnostic laboratories and government agencies often hold existing datasets that can be repurposed, reducing primary collection costs. Budget for at least one round of data validation, because positional errors in farm coordinates propagate directly into cluster boundaries and risk estimates.

What should I do when farm coordinates are imprecise or missing?

Imprecision is common in veterinary data. Postcode centroids, village centroids, or administrative district polygons are acceptable substitutes when exact farm locations are unavailable, provided the spatial unit is stated explicitly in all outputs. Assign each record to the smallest reliable geographic unit and analyze at that scale. Do not interpolate missing coordinates from neighbouring farms, as this artificially inflates clustering. Sensitivity analysis is essential: repeat the cluster detection using coarser spatial units and confirm that identified clusters persist. If clusters disappear at a coarser scale, report them as scale-dependent findings. The CDC principles of epidemiology emphasize that the spatial resolution of data determines which hypotheses can be tested, and this constraint should be documented in the methods section.

How do spatial analysis approaches differ between livestock and wildlife populations?

Livestock populations have defined ownership, known movement records, and stable spatial units such as holdings or postcode areas, which simplifies denominator construction and cluster interpretation. Wildlife populations lack these structures. Home ranges, seasonal migration, and variable detection probability mean that apparent clusters may reflect sampling effort instead of true disease aggregation. For wildlife, use detection-corrected prevalence estimates and interpret clusters as indicators of ecological drivers instead of fixed geographic boundaries. Cross-species transmission complicates both settings, as demonstrated in analyzes of H5N1 diffusion where avian and human population densities jointly shaped viral spread (phylogeographic and spatial predictors of H5N1 diffusion). Consult species-specific reference material from the MSD Veterinary Manual when defining the epidemiological unit and natural history parameters.

What records must I keep to make a spatial analysis reproducible and defensible?

Maintain a versioned dataset with the original georeferenced records, the cleaning log, and the final analytical file. Document every coordinate transformation, any records excluded and why, and the spatial unit used at each analytical step. Record the software version, package versions, and parameter settings for each cluster detection and regression run. Keep the random seed if stochastic methods are used. This documentation supports peer review and regulatory scrutiny. The WOAH animal health surveillance standards require traceable data management for internationally reported diseases, and the same standard strengthens any local investigation. Store the data in a format that does not depend on proprietary software, and archive it with a date-stamped file name.

How do I explain spatial cluster findings to a producer or practice owner?

Frame the explanation around practical decisions, not statistical mechanics. State that the analysis identified areas where disease occurs more frequently than expected, and that this pattern may reflect shared risk factors such as drainage, stocking density, or wildlife contact. Use a printed map with the cluster boundary overlaid on familiar landmarks, and avoid presenting confidence intervals or p-values. Explain that a cluster is a hypothesis-generating finding, not proof of causation, and that farm-level investigation is needed before changing management. Reference the AVMA practice resources for guidance on communicating epidemiological findings to clients. Offer concrete next steps, such as targeted testing or biosecurity review, and clarify that the cluster boundary does not imply that farms outside it are risk-free.

When is it appropriate to use a simple visual map instead of formal cluster detection?

Visual mapping is appropriate for initial exploration, small datasets, or when the spatial signal is strong and obvious. It is also the correct choice when the research question is descriptive, such as documenting the geographic range of a newly detected pathogen. Formal cluster detection is required when the goal is to compare observed case counts against expected counts, to identify statistically significant aggregations, or to prioritize areas for intervention. Visual inspection alone cannot distinguish true clustering from random variation, particularly with small case numbers or uneven population density. The distinction between visualization and formal clustering methods is a central theme in spatial epidemiology reviews of infectious disease transmission. When in doubt, run both and report the formal results as primary, with the map as supporting material.

Related Clinical & Scientific Guides

• Evaluating Veterinary Surveillance System Attributes
• Network Analysis for Infectious Disease Spread in Animal Populations
• Randomized Controlled Trials in Veterinary Field Settings

References and Further Reading

• How Spatial Epidemiology Helps Understand Infectious Human Disease Transmission.. 2022.
• Emerging virus diseases: can we ever expect the unexpected?. 2012.
• Spatial epidemiology of cutaneous leishmaniasis in Colombia: socioeconomic and demographic factors associated with a growing epidemic.. 2019.
• Airborne geographical dispersal of Q fever from livestock holdings to human communities: a systematic review and critical appraisal of evidence.. 2018.
• The development of linear regression models using environmental variables to explain the spatial distribution of Fasciola hepatica infection in dairy herds in England and Wales.. 2010.
• Combining phylogeography and spatial epidemiology to uncover predictors of H5N1 influenza A virus diffusion.. 2015.
• WOAH Animal Health Surveillance Standards. WOAH.
• CDC Principles of Epidemiology in Public Health Practice. CDC.
• MSD Veterinary Manual, Professional Edition. MSD Veterinary Manual.

Related Articles

• Bayesian Hierarchical Models for Veterinary Disease Mapping
• Spatial Cluster Detection Methods in Veterinary Epidemiology
• Participatory Epidemiology Methods for Livestock Disease Surveillance
• Using Simulation Models in Veterinary Epidemiology
• Compartmental Models in Veterinary Disease Dynamics

This article is educational professional reference material for veterinary audiences. It is not a substitute for veterinary diagnosis, individual clinical judgment, current product labeling, or applicable regulatory requirements.