Risk-Based Surveillance in Animal Health

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

Key Takeaways

• Risk-based surveillance prioritizes finite resources by directing diagnostic and epidemiological efforts toward populations, places, and pathways with the highest probability of harboring or introducing a hazard, thereby improving detection sensitivity and economic efficiency compared to representative sampling.
• Formal risk assessment is the foundational driver, identifying populations or geographic areas most likely to harbor a hazard based on factors influencing pathogen release (e.g., illegal imports, prevalence in source regions) and exposure pathways (e.g., contact routes).
• Surveillance objectives critically dictate design choices: early detection necessitates oversampling high-risk strata to maximize the probability of identifying an outbreak, while prevalence estimation requires unbiased sampling to ensure accurate population-level inference.
• Population stratification is achieved by identifying subgroups with differing exposure, susceptibility, or detection probabilities, often utilizing production systems (e.g., backyard vs. controlled herds for Trichinella), animal movement networks (e.g., in-degree, ingoing infection chains), or geographic pathways.
• The practical design sequence involves defining the objective and target parameter, identifying and ranking risk factors, stratifying the population, selecting sampling methods within strata (e.g., two-stage cluster sampling), and defining pre-specified decision thresholds and response protocols.
• Common failure modes include misclassification of the target population due to outdated data, selection bias within strata from sampling accessible animals, temporal misalignment of risk factors, and diagnostic error, particularly amplified in high-prevalence strata.

Risk-based surveillance directs finite diagnostic and epidemiological resources toward populations, places, and pathways with the highest probability of harboring or introducing a hazard. This article examines the conceptual foundations, design logic, and operational methods of risk-based surveillance for veterinary researchers and epidemiologists working across production animal, wildlife, and One Health contexts. It addresses how surveillance objectives, risk assessment outputs, and sampling strategies are integrated to improve detection sensitivity and economic efficiency relative to conventional approaches.

The central question is practical: when resources cannot support exhaustive testing, how should surveillance effort be allocated? The answer requires explicit reasoning about where infection is most likely to occur, how it will enter a population, and which sampling strata will maximize the probability of detection. This article provides the frameworks, terminology, and worked examples needed to design, evaluate, and defend such systems.

At a Glance

Defining Risk-Based Surveillance

Risk-based surveillance applies formal risk assessment methods at one or more stages of traditional surveillance design, including hazard selection, population stratification, sampling allocation, and data interpretation. The defining feature is that certain strata of the target population have a higher probability of being sampled because they carry a higher probability of infection or introduction. This contrasts with representative sampling, where every unit has an equal or known chance of selection regardless of risk status.

The rationale is economic as much as epidemiological. Human and financial resources for government veterinary services are constrained in many countries, and issues that present higher risks merit higher priority because investments yield better benefit-cost ratios. The rapid acceptance of this concept has outpaced the development of its theoretical and practical bases, leaving gaps in standardized methodology that later work has sought to fill.

Conceptual Foundations

Risk Assessment as the Design Driver

Risk-based surveillance depends on a prior risk assessment that identifies which populations, production types, or geographic areas are most likely to harbour a hazard. The assessment may be qualitative, semi-quantitative, or fully quantitative, but it must be explicit about the factors that influence release and exposure. For transboundary disease introduction, for example, the likelihood of release depends on the volume and origin of illegal imports, the prevalence of infection in source regions, and the survival of the pathogen in transported products. Exposure depends on the contact pathways between contaminated products and susceptible domestic populations.

A semi-quantitative framework developed for African swine fever introduction into the European Union illustrates the approach. The framework scored countries on factors influencing the likelihood of release of contaminated smuggled meat and subsequent exposure of susceptible pigs, producing a relative risk scale from negligible to very high. This allowed surveillance planners to rank countries and allocate effort accordingly, despite the inherent difficulty of quantifying illegal movement of animal products.

Surveillance Objectives and Design Consequences

The objective of surveillance determines whether risk-based allocation is appropriate and how it should be implemented. When the aim is disease detection and identification of case herds, risk-based approaches increase the sensitivity of the surveillance system. When the aim is estimating prevalence or demonstrating freedom, biased sampling toward high-risk strata complicates inference and requires adjustment.

The distinction matters for evaluation. Surveillance attributes such as sensitivity, specificity, and representativeness must be assessed against the stated objective, and economic efficiency is a recognized evaluation attribute in international surveillance terminology. A system that detects more positives per sample is not automatically better if the objective requires unbiased population estimates.

Stratification and Sampling Logic

Population Subgroups and Risk Factors

Risk stratification begins with identifying subpopulations that differ in exposure, susceptibility, or detection probability. Production system is often the strongest single stratifier. For Trichinella surveillance in pigs, parasites circulate mainly in backyard and free-ranging pigs, while herds under controlled management conditions are the ones routinely tested. Testing the accessible, well-managed herds yields negative results year after year while human cases continue to occur from untested high-risk sources. A risk-based system documents housing conditions and feedstuff sources to identify which herds require testing and which can be exempted.

Network-Based Targeting

Animal movement data provide another stratification basis. Network analysis measures such as in-degree and ingoing infection chain identify herds that are central to pathogen transmission through purchase behavior. In Swedish cattle populations, positive serological results for bovine coronavirus and bovine respiratory syncytial virus were significantly associated with animal purchases, and selecting herds based on network measures increased the number of detected positives compared with random sampling. The practical implication is that movement databases, updated for the relevant time period, offer a straightforward tool for risk-based herd selection.

One Health and Transdisciplinary Surveillance

Risk-based surveillance extends naturally to the human-wild animal interface, where the objective is detecting viruses at their source before spillover into people or food animals. The PREDICT project consortium designed targeted surveillance based not on humans as sentinels but on ecological and behavioral risk factors that predict viral emergence. This transdisciplinary approach required collaboration across veterinary, medical, ecological, and social science sectors, and it demonstrated that risk-based allocation can operate across species boundaries when the hazard is zoonotic.

The One Health framing adds complexity to risk assessment because the relevant populations span domestic animals, wildlife, and humans, each with different exposure pathways and detection opportunities. Surveillance planners must decide which species or environmental samples provide the earliest and most efficient detection signal, and this decision is itself a risk-based allocation problem.

Practical Design Sequence for Risk-Based Surveillance

The operational core of risk-based surveillance is a structured sequence that moves from hazard identification to resource allocation. The sequence begins with defining the surveillance objective in measurable terms, then proceeds through risk factor identification, population stratification, sampling design, and finally data interpretation against pre-specified decision thresholds. Each step constrains the options available at the next, so design errors compound rapidly.

Step 1: Define the Objective and the Target Parameter

The objective determines every subsequent choice. Early detection of an exotic pathogen requires a different design than demonstrating freedom from infection for trade purposes, which in turn differs from estimating prevalence in a defined population. The World Organization for Animal Health distinguishes these purposes in its surveillance standards, and the distinction matters because the statistical targets differ. For freedom from disease, the design must achieve a specified design prevalence and confidence level, typically 95% confidence of detecting infection at a design prevalence of 1% or lower in a defined population. For early detection, the target is timeliness: the system must detect introduction before amplification and spread reach a threshold that makes control impractical. For prevalence estimation, the target is precision, expressed as a confidence interval width.

The objective also determines whether the unit of interest is the individual animal, the herd, the region, or the production sector. Herd-level sensitivity and individual-level sensitivity are different quantities, and the design must specify which one the sampling strategy is optimized to achieve.

Step 2: Identify and Rank Risk Factors

Risk factor identification draws on published literature, historical surveillance data, expert elicitation, and pathway analysis. The factors that matter vary by pathogen, species, and production system. For endemic infections spread by animal movement, network measures derived from movement records identify herds at elevated risk. In Swedish cattle populations, the number of incoming animal movements and the size of the ingoing infection chain were significantly associated with seropositivity for bovine coronavirus and bovine respiratory syncytial virus, and selecting herds on these network measures increased the number of detected positives compared with random sampling. For pathogens introduced through trade, the risk pathway may run through illegal importation of contaminated products instead of live animal movements, as demonstrated in risk assessments for African swine fever introduction into the European Union.

Risk factors should be ranked by the strength of their association with the outcome, the prevalence of the factor in the population, and the feasibility of using the factor for stratification in practice. A strong risk factor that cannot be measured reliably at the population level is of limited design value.

Step 3: Stratify the Population

Stratification divides the target population into groups with different expected probabilities of infection or different consequences of infection. The strata must be mutually exclusive, exhaustive, and defined by variables that are measurable at the point of sampling. Common stratification variables include production type, herd size, geographic region, movement history, biosecurity status, and previous test results.

The allocation of samples across strata follows one of two logics. Proportionate allocation assigns samples in proportion to stratum size and maximizes precision for prevalence estimation. Disproportionate allocation deliberately oversamples high-risk strata and maximizes the probability of detecting infection when the objective is early detection or freedom from disease. The choice between these allocations is a direct expression of the surveillance objective.

Step 4: Select Sampling Methods Within Strata

Within each stratum, the sampling method determines whether the design can support statistical inference. Simple random sampling provides unbiased estimates but is often logistically impractical in animal populations. Two-stage cluster sampling, with herds selected at the first stage and animals within herds at the second, is the standard approach for production animal populations. Convenience sampling at slaughterhouses or diagnostic laboratories is frequently used but introduces selection bias that must be acknowledged in interpretation.

The critical design decision is the number of herds and the number of animals per herd. For a fixed total sample size, sampling more herds with fewer animals per herd increases herd-level sensitivity, while sampling fewer herds with more animals per herd increases the confidence of detecting infection within selected herds. The optimal balance depends on the expected within-herd prevalence and the between-herd distribution of infection.

Step 5: Define Decision Thresholds and Response Protocols

A surveillance system has no operational value unless the data trigger predefined actions. Decision thresholds should be specified before data collection begins. For freedom from disease, the threshold is the number of positive results that would overturn the freedom declaration. For early detection, the threshold is the number of suspect cases or the detection of a pathogen with epidemic potential that triggers a confirmed outbreak investigation.

The response protocol should specify who is notified, what additional diagnostic testing is performed, what movement restrictions apply, and what epidemiological investigation is initiated. These protocols are often defined by national veterinary authorities and international standards, and the surveillance design must be compatible with the reporting obligations of the jurisdiction in which it operates.

Documenting the Design and Its Assumptions

Transparency in surveillance design is a professional obligation, not an administrative formality. The design document should record the objective, the target parameter and its statistical specification, the risk factors used for stratification and the evidence supporting each, the sampling frame and its limitations, the sample allocation, the laboratory tests and their diagnostic sensitivity and specificity, and the decision thresholds. The document should also state the assumptions that were made where data were incomplete.

The value of this documentation becomes apparent when surveillance results are negative. A negative result is only interpretable if the design sensitivity of the system can be calculated and reported. Design sensitivity is the probability that the surveillance system would detect infection if it were present at the design prevalence, and it depends on the sampling strategy, the diagnostic test performance, and the risk-based weighting of the sampled strata. Reporting design sensitivity alongside surveillance results allows trading partners and policymakers to assess the evidential weight of a negative finding.

Species and Production System Modifications

The correct design choices differ substantially across species and production systems. In intensive pig production, the herd is the natural epidemiological unit, and movement records are often complete and accessible. In backyard and free-ranging pig populations, the epidemiological unit is less well defined, movement records are absent, and the risk profile is different. Testing programs that sample only controlled housing conditions while the parasite burden circulates in free-ranging and backyard animals will produce misleading reassurance, as demonstrated for Trichinella surveillance. The design must match the population structure in which the pathogen actually circulates, not the population structure that is easiest to sample.

In wildlife populations, the sampling frame is rarely known, and convenience sampling at hunter harvest or live capture is often the only feasible approach. The design sensitivity must be interpreted with explicit acknowledgement of the unknown sampling frame. In companion animal populations, the sampling frame is the population of animals presented to veterinary clinics, which is a biased subset of the total population, and the direction and magnitude of that bias must be considered in interpretation.

Monitoring Parameters and System Evaluation

Surveillance systems require ongoing evaluation, also initial design. The attributes that should be monitored include sensitivity, timeliness, representativeness, and cost. Sensitivity can be assessed through the detection of known positive cases, through comparison with alternative data sources, or through simulation studies. Timeliness is measured as the interval between infection, sample collection, laboratory confirmation, and notification. Representativeness is assessed by comparing the characteriztics of sampled animals with the target population.

The evaluation should be scheduled at defined intervals and should feed back into design revision. The World Organization for Animal Health surveillance standards describe the attributes that should be used for this evaluation, and the standards provide a common vocabulary for describing surveillance activities across countries. Economic efficiency is a central evaluation attribute, because the justification for risk-based surveillance is the achievement of higher benefit-cost ratios with existing or reduced resources.

The following table summarizes the key design decisions and the factors that should drive each choice.

Uncertainty and Evidence Gaps

The evidence base for risk-based surveillance is uneven. The theoretical framework for applying risk assessment methods to surveillance design is well established, but the empirical validation of specific risk-based approaches remains limited. The rapid acceptance of the core concept has outpaced the development of its theoretical and practical bases, and this gap is most visible in the scarcity of field studies that compare risk-based designs with random sampling under operational conditions. Network-based targeting has stronger empirical support in cattle populations, where movement data are systematically recorded, than in other species or production systems.

The choice of risk factors and the weighting applied to them should be revisited as new data accumulate. A risk factor that was significant in one population or time period may not transfer to another context, and the surveillance design should include mechanisms for updating the risk model as surveillance data and movement data accrue. Where the evidence base is contested, the design document should state the competing positions and the basis for the choice made.

Recognized Complications and Failure Modes

Risk-based surveillance fails in characteriztic patterns, most of which trace back to a mismatch between the design assumptions and the field reality. The most common failure is misclassification of the target population. When the risk strata are built on outdated demographic data, sampling frames no longer reflect the current distribution of high-risk animals. This is detected early by comparing predicted stratum sizes against observed enrollment rates at each sampling point. A persistent shortfall in one stratum signals either a sampling logistics problem or a shift in the underlying population structure.

Selection bias within strata is a second frequent failure. Field teams tend to sample accessible animals, which in many production systems means the calmest, youngest, or most recently handled individuals. These animals may not represent the within-stratum risk distribution. The discriminating check is to record refusal rates and substitution patterns. If more than a small fraction of sampled animals are replaced because the originally selected animal was unavailable, the design's statistical basis is compromised.

A third failure mode is temporal misalignment. Risk factors change seasonally, and a design built on winter housing conditions loses validity by mid-summer. This is detected by monitoring the positivity rate over time within each stratum. A gradual convergence of detection rates across strata suggests that the stratification variables no longer separate risk groups effectively.

The fourth recognized complication is diagnostic error, either from imperfect test sensitivity or from sample degradation during transport. Risk-based designs concentrate sampling in high-prevalence strata, which amplifies the effect of false negatives. Early detection requires the inclusion of internal quality controls, such as split samples or periodic retesting of a random subset at a reference laboratory.

Common Errors in Design and Execution

Less experienced practitioners frequently confuse risk-based sampling with convenience sampling. The distinction is that risk-based designs assign explicit inclusion probabilities derived from a documented risk assessment, whereas convenience sampling selects whatever is easiest to obtain. The corrective action is to write the inclusion probability for each stratum into the sampling protocol before field work begins.

A second common error is over-stratification. Designers create many narrow strata, each with a small sample size, which destroys the statistical power the risk-based approach was intended to create. The corrective action is to collapse strata that share similar risk profiles and to verify that each remaining stratum can support the minimum sample size required for the detection target.

A third error is the failure to update the risk assessment when the system changes. A surveillance program designed for a closed herd is invalid after the operation begins purchasing replacement stock. The corrective action is to schedule a formal design review at fixed intervals and after any major change in production practices, movement patterns, or disease status in the surrounding region.

Limitations of the Current Evidence

The evidence base for risk-based surveillance is strongest for production species with well-characterized movement networks and for diseases with known risk factor profiles. It is considerably weaker for wildlife, where population sizes, movement patterns, and diagnostic test performance are often poorly characterized. The PREDICT project demonstrated that targeted surveillance at the human-wild animal interface can detect viruses before spillover, but the generalizability of that approach to routine surveillance programs remains uncertain PREDICT consortium findings on transdisciplinary surveillance.

Expert opinion still differs on the appropriate role of risk-based methods for proving freedom from disease. Some authorities argue that risk-based sampling can support freedom claims when the risk assessment is well documented, while others maintain that demonstration of freedom requires probabilistic sampling that is not biased by risk stratification. The proposed terms and concepts for describing and evaluating animal-health surveillance systems acknowledge these inconsistencies in terminology and approach, and the field has not yet reached consensus on this point.

A further limitation is the scarcity of validation studies that compare risk-based designs against comprehensive census sampling for the same population and disease. Most published examples, such as the Swedish cattle movement study, demonstrate that risk-based approaches detect more positives than random sampling, but they do not establish the sensitivity of the risk-based approach relative to a true gold standard network analysis parameters in risk-based surveillance.

Referral, Consultation, and Escalation

Referral to specialist expertise is warranted when the risk assessment requires quantitative methods beyond the local team's capacity. This includes formal import risk assessment, complex network analysis, or spatial modeling. The WOAH terrestrial animal health code provides the international framework for surveillance design and reporting, and consultation with the national veterinary authority is appropriate when a design must support trade claims.

Laboratory involvement is required when diagnostic test performance is uncertain for the target species or when samples originate from a new geographic region. Reference laboratory confirmation should be built into the design for any positive result that will trigger regulatory action.

Regulatory reporting is mandatory when a surveillance finding meets the case definition for a notifiable disease. The WOAH animal health surveillance standards describe the reporting obligations and the information that must accompany a notification. The decision to escalate should be made on the basis of the case definition, not on the strength of the risk-based design. A positive finding in a low-risk stratum carries the same regulatory weight as one in a high-risk stratum.

Frequently Asked Questions

How Do I Prioritize Risk-Based Surveillance When Budgets Are Severely Constrained?

Start by ranking hazards according to their public health, economic, and trade consequences, since these consequences should drive selection of diseases or hazards in any risk-based design. Focus resources on the highest-ranked strata where detection probability is greatest, instead of attempting broad coverage. For example, testing only controlled-housing herds for Trichinella while free-ranging pigs carry the parasite wastes resources and misses human risk, so redirect sampling toward documented high-risk production systems. Document the assumptions behind each prioritization decision so that when funds change, the logic can be revisited. Even a semi-quantitative risk ranking, as used for African swine fever introduction through illegal pork imports, can justify targeted surveillance where fully quantitative data are unavailable.

What Can I Do When the Ideal Sampling Frame or Laboratory Capacity Is Unavailable?

Use proxy indicators that correlate with the risk factor of interest. If individual animal identification is incomplete, target premises based on network measures such as in-degree or ingoing infection chain, which have been shown to increase detection of seropositive herds compared with random sampling. If laboratory throughput is limited, pool samples within strata and test strategically, but validate pooling protocols against the diagnostic sensitivity required for your surveillance objective. Where field laboratories are absent, consider whether samples can be transported to a central facility without compromising analyte stability. Document every deviation from the ideal design and assess how it affects the sensitivity of the system, since transparency about limitations is essential for interpreting negative results.

How Does Risk-Based Surveillance Differ for Wildlife Compared With Domestic Livestock?

Wildlife populations lack the movement records and premises-level management data that make livestock surveillance straightforward. The PREDICT project demonstrated that targeting high-risk interfaces, such as hunting, trade, and habitat encroachment, can detect viruses at their source before spillover into people or food animals. Sampling must account for species ecology, home range, and seasonal aggregation, which influence both risk and detectability. Capture methods introduce welfare and safety constraints that do not apply to livestock. Diagnostic test validation is often weaker in wildlife species, so interpret results with caution. Collaboration with wildlife agencies and conservation groups is usually necessary, and the transdisciplinary structure of the surveillance team becomes a design element instead of an administrative detail.

What Records Must I Keep to Defend a Risk-Based Surveillance Decision Later?

Record the stated objective, the target parameter, and the risk factors used for stratification, with the evidence or expert opinion supporting each factor. Document the population size, stratum boundaries, sampling fractions, and the statistical method used to calculate sample size. Keep the raw data, laboratory results, and any deviations from the protocol, including reasons for each deviation. Record the decision thresholds and the response protocol triggered by a positive result. The WOAH surveillance standards emphasize transparency and comparability, so your documentation should allow another epidemiologist to reconstruct the design and evaluate its sensitivity. If the system is evaluated later, these records become the basis for assessing whether the surveillance met its stated purpose.

How Do I Explain a Risk-Based Approach to a Producer Who Expects Every Animal to Be Tested?

Explain that testing every animal is often less protective than testing the right animals. Use a concrete example, such as Trichinella testing in pigs, where millions of controlled-housing pigs are tested while the parasite circulates in backyard and free-ranging herds, meaning the wrong animals are being sampled. Frame the approach around the producer's own risk: if their herd is low-risk, they may accept a smaller sampling fraction with confidence, and if they are high-risk, they benefit from more intensive scrutiny. Emphasize that the goal is not to reduce testing but to place testing where it detects disease earliest. Provide the written protocol and the risk assessment behind it, and invite questions about how their operation was classified.

When Should I Abandon a Risk-Based Design and Return to Representative Sampling?

Abandon the risk-based design when the risk factors used for stratification are no longer supported by current evidence or when the population structure has changed so that the strata no longer reflect true risk. If the surveillance objective shifts from detection of infection to estimation of prevalence for trade or certification, representative sampling is required because risk-based sampling introduces selection bias that invalidates unbiased prevalence estimates. Also reconsider the design if evaluation shows the system is missing cases in low-risk strata that subsequently prove to be epidemiologically important. The choice between risk-based and representative sampling depends on the objective, and the two approaches answer different questions, so revisit the decision whenever the objective changes.

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

• Concepts for risk-based surveillance in the field of veterinary medicine and veterinary public health: review of current approaches.. 2006.
• One Health proof of concept: Bringing a transdisciplinary approach to surveillance for zoonotic viruses at the human-wild animal interface.. 2017.
• Proposed terms and concepts for describing and evaluating animal-health surveillance systems.. 2013.
• Searching for Trichinella: not all pigs are created equal.. 2014.
• Introduction of African swine fever into the European Union through illegal importation of pork and pork products.. 2013.
• Application of network analysis parameters in risk-based surveillance - examples based on cattle trade data and bovine infections in Sweden.. 2012.
• 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

• Risk Assessment Frameworks for Veterinary Public Health
• One Health Surveillance: Integrating Human, Animal, and Environmental Data
• Syndromic Surveillance in Veterinary Practice
• Risk Factor Analysis for Disease in Animal Populations
• Time Series Analysis for Veterinary Disease Surveillance

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.