# Likelihood Ratios in Veterinary Diagnostic Testing


## Key Takeaways

- Likelihood ratios (LRs) quantify how much a diagnostic test result shifts the probability of disease, directly addressing the clinician's question of post-test probability revision, unlike sensitivity and specificity alone.
- Positive likelihood ratios (LR+) are calculated as sensitivity / (1 - specificity), and negative likelihood ratios (LR-) as (1 - sensitivity) / specificity, allowing for calculation of post-test odds by multiplying pre-test odds by the LR.
- Interpretation benchmarks for LRs include values above 10 (LR+) or below 0.1 (LR-) indicating large probability shifts, while values near 1.0 provide minimal diagnostic information, with clinical utility dependent on the pretest probability relative to action thresholds.
- Sequential testing with conditionally independent tests allows for multiplicative application of LRs, but conditional dependence (e.g., two ELISA tests targeting the same viral antigen) can lead to overestimation of diagnostic gain, necessitating caution or use of multivariate LRs.
- Pretest probability estimation is crucial, drawing from prevalence data, signalment, history, and examination findings, and should be individualized; posttest probability is then calculated via odds conversion: (Pretest Probability / (1 - Pretest Probability)) * LR = Posttest Odds, then converted back to probability.
- Published LRs should be used with caution, considering potential population mismatches (e.g., disease spectrum, severity) and the risk of conditional dependence between sequentially applied tests, which can inflate posttest probabilities.

---

Clinical decision-making in veterinary medicine depends on translating test results into a revised probability of disease. Sensitivity and specificity describe test performance in the abstract, but they do not directly answer the question a clinician faces after a result is in hand: how much should this result change my suspicion of disease? Likelihood ratios (LRs) answer that question by quantifying the weight a test result carries for or against a diagnosis. This article explains the derivation, interpretation, and practical application of likelihood ratios across species and clinical contexts, with attention to their advantages over other test performance metrics and their limitations in real-world veterinary settings.

The intended reader is a veterinary researcher or clinician comfortable with epidemiologic terminology and quantitative reasoning. The article assumes familiarity with sensitivity, specificity, and pretest probability, and focuses instead on the logic of likelihood ratios, their calculation from published data, their use in sequential testing, and common pitfalls in their application. The clinical question addressed throughout is direct: given a measured pretest probability and a test result, what is the posttest probability of disease, and how should that probability guide action?

## At a Glance

| Parameter | Definition or decision point |
|---|---|
| Positive likelihood ratio (LR+) | Probability of a positive test in diseased animals divided by probability of a positive test in non-diseased animals, equals sensitivity / (1 minus specificity) |
| Negative likelihood ratio (LR-) | Probability of a negative test in diseased animals divided by probability of a negative test in non-diseased animals, equals (1 minus sensitivity) / specificity |
| Interpretation benchmark | LR+ above 10 and LR- below 0.1 generate large shifts in probability, values near 1.0 carry no diagnostic information |
| Pretest probability | Estimated from prevalence, signalment, history, and examination findings, the starting point for probability revision |
| Posttest probability | Computed by converting pretest probability to odds, multiplying by the likelihood ratio, then converting back to probability |
| Sequential testing | Likelihood ratios from independent tests can be multiplied, but conditional dependence between tests invalidates simple multiplication |
| Reporting standard | Published test evaluations should report likelihood ratios with confidence intervals, also sensitivity and specificity |

## The Probability Revision Framework

Diagnostic testing is an exercise in conditional probability. A test result partitions animals into those with and without disease, but the clinician observes the result, not the disease status. Likelihood ratios invert the conditional relationship: they describe the probability of observing a particular test result given each disease state, and they do so in a form that can be applied directly to an individual patient.

The likelihood ratio for a given test result is the ratio of two conditional probabilities. For a binary test, the LR+ is the probability of a positive result among diseased animals divided by the probability of a positive result among non-diseased animals. The LR- is the probability of a negative result among diseased animals divided by the probability of a negative result among non-diseased animals. These quantities derive entirely from sensitivity and specificity, but they express test performance in a form that is independent of disease prevalence, which makes them portable across populations with different underlying risks.

The power of the likelihood ratio framework lies in its compatibility with Bayes theorem. Pretest probability is converted to pretest odds by dividing probability by its complement. Multiplying pretest odds by the likelihood ratio yields posttest odds, which convert back to posttest probability. This calculation is simple enough to perform at the bedside with a handheld calculator or a Fagan nomogram, and it provides a principled alternative to the informal gestalt that often governs test interpretation.

## Derivation From Sensitivity and Specificity

For a dichotomous test, the derivation is algebraic. If sensitivity is the proportion of diseased animals that test positive, and specificity is the proportion of non-diseased animals that test negative, then:

LR+ = sensitivity / (1 minus specificity)

LR- = (1 minus sensitivity) / specificity

These formulas reveal an important asymmetry. A test with high sensitivity and modest specificity can still produce a useful LR+ if the false-positive rate is low. Conversely, a test with high specificity but poor sensitivity may have a useful LR- if the false-negative rate is low. The likelihood ratio therefore captures the joint contribution of both test characteriztics in a single number, which is one reason it outperforms sensitivity or specificity alone as a guide to clinical interpretation.

The same logic extends to tests with more than two result categories. A quantitative assay can be divided into ordinal strata, and a likelihood ratio computed for each stratum as the proportion of diseased animals in that stratum divided by the proportion of non-diseased animals in that stratum. This approach preserves information that is lost when a continuous result is forced into a binary positive or negative classification. For example, a mildly elevated enzyme activity may carry a small likelihood ratio, while a markedly elevated value carries a much larger one, and collapsing both into a single positive category obscures that distinction.

## Interpretation Thresholds and Clinical Decision Making

Likelihood ratios are interpreted on a continuous scale, but convention provides useful anchors. A likelihood ratio of 1.0 means the test result does not change the probability of disease at all. Values between 1 and 2, or between 0.5 and 1, shift probability only slightly and rarely justify changes in management. Values between 2 and 5, or between 0.2 and 0.5, produce moderate shifts that may be clinically meaningful when the pretest probability is already near a treatment threshold. Values above 10 or below 0.1 generate large shifts that can confirm or exclude a diagnosis in many clinical scenarios.

These thresholds are heuristics, not mathematical laws. The clinical value of a likelihood ratio depends on where the pretest probability sits relative to the action thresholds for the case at hand. A test with an LR+ of 5 may be decisive when the pretest probability is 40 percent, because the posttest probability crosses the treatment threshold, but the same test is unhelpful when the pretest probability is already 95 percent. The likelihood ratio framework makes this dependence explicit and forces the clinician to state the pretest probability before interpreting the result.

## Sequential Testing and Conditional Dependence

One of the most valuable properties of likelihood ratios is their multiplicative nature in sequential testing. When two tests are conditionally independent given disease status, the posttest odds after the first test become the pretest odds for the second, and the likelihood ratios multiply. This allows a clinician to combine results from, for example, a screening test and a confirmatory test in a principled way.

The assumption of conditional independence is rarely met in practice. Two tests that measure related physiologic pathways, such as two inflammatory markers or two imaging modalities that detect the same lesion, are likely to be correlated within both the diseased and non-diseased populations. When tests are positively correlated, simple multiplication of likelihood ratios overestimates the posttest probability, sometimes substantially. The magnitude of this error depends on the degree of correlation, and it is rarely reported in veterinary test evaluation studies. Clinicians should therefore apply sequential likelihood ratio multiplication with caution, particularly when the tests in question are mechanistically similar.

The statistical literature on likelihood ratio testing in related contexts illustrates the importance of careful model specification. Likelihood ratio tests are used throughout biomedical research to compare nested models and to evaluate the contribution of additional predictors, as described in frameworks for identifying susceptibility genes in the presence of epistasis and in real-time vaccine safety surveillance. These applications share a common logic with diagnostic likelihood ratios: the ratio of probabilities under competing hypotheses quantifies the evidence each observation provides. The same principle that guides selection of genetic markers or detection of adverse events guides interpretation of a single diagnostic test result in a clinical setting.

## Practical Application in the Clinical Workup

The clinical value of likelihood ratios emerges when they are integrated into a structured diagnostic sequence. The process begins with a pretest probability, often estimated from signalment, history, and physical examination findings. The likelihood ratio then converts that pretest probability into a posttest probability through the odds form of Bayes theorem. This conversion is the operational core of likelihood ratio use in practice.

### The Odds Conversion Sequence

The calculation proceeds in three steps. First, convert pretest probability to pretest odds: odds equals probability divided by one minus probability. Second, multiply pretest odds by the likelihood ratio to obtain posttest odds. Third, convert posttest odds back to probability: probability equals odds divided by one plus odds.

A 12 percent pretest probability converts to odds of 0.136. A positive likelihood ratio of 10 yields posttest odds of 1.36, which corresponds to a posttest probability of approximately 58 percent. The same pretest probability with a positive likelihood ratio of 2 yields a posttest probability near 21 percent. The arithmetic is straightforward, but the clinical judgment lies in selecting the pretest probability and choosing which likelihood ratio applies.

### Selecting the Pretest Probability

Pretest probability estimation draws on published prevalence data, regional disease patterns, and individual patient factors. For production animal medicine, herd-level prevalence from surveillance programs and [WOAH animal health surveillance standards](https://www.woah.org/en/what-we-do/animal-health-and-welfare/disease-data-collection/) provide a defensible starting point. For companion animal practice, clinic-level prevalence data and published case series inform the estimate.

The pretest probability should reflect the specific patient, not a generic population figure. A young unvaccinated puppy with acute hemorrhagic diarrhea carries a higher pretest probability of parvovirus than a vaccinated adult dog with the same signs. The likelihood ratio then adjusts that individualised estimate. When the pretest probability is uncertain, clinicians should test the robustness of the posttest conclusion across a plausible range of pretest values. If the management decision changes across that range, further testing is warranted before committing to treatment.

## Likelihood Ratios in Sequential Testing

Sequential testing occurs when a second test is performed after a first test result. The posttest probability from the first test becomes the pretest probability for the second. This process is valid only when the two tests are conditionally independent, meaning their errors do not correlate within disease status groups.

Conditional dependence arises when two tests measure the same biological pathway or share a common failure mode. Two serologic tests targeting the same immunoglobulin class may fail together in an immunocompromised patient. Two imaging modalities may both miss a lesion below their shared resolution threshold. In these situations, applying the second test's likelihood ratio as though it were independent overestimates the diagnostic gain. The [CDC principles of epidemiology in public health practice](https://www.cdc.gov/csels/dsepd/ss1978/index.html) describe the underlying assumptions of test combination that apply here.

When conditional dependence is suspected, the clinician has three options. The first is to use tests that measure different biological dimensions, such as pairing an antigen test with a serologic test. The second is to interpret the combined result pattern using published likelihood ratios for the test pair, when such data exist. The third is to treat the sequential posttest probability with caution, acknowledging that the true value lies somewhere between the naive calculation and the first test's posttest probability.

## Likelihood Ratios for Test Panels

Test panels present a special case. When a panel of tests is run simultaneously, the clinician faces the question of how to interpret the pattern of results. The individual likelihood ratios for each test cannot simply be multiplied across the panel, because the tests within a panel are rarely conditionally independent. They often share physiologic pathways or analytic platforms.

A practical approach is to identify the single most informative test in the panel for the clinical question, apply its likelihood ratio, and then use the remaining results to adjust the posttest probability qualitatively. For example, in a feline viral panel, the PCR result for feline leukemia virus proviral DNA carries more weight than the antibody result in a vaccinated cat. The antibody result then modifies the interpretation only when the PCR is negative and the clinical suspicion remains high.

Multivariate likelihood ratios, which describe the diagnostic performance of an entire result pattern, are published for some veterinary test combinations. These are preferable when available because they account for the correlation structure among tests. Their absence from most clinical references means the qualitative adjustment approach remains the standard in practice.

## Species and Setting Modifications

The correct application of likelihood ratios varies with species, production system, and available equipment. In food animal practice, the consequences of a false positive differ from those in companion animal practice. A false positive for a trade-restricting disease triggers quarantine and economic loss, so the threshold for acting on a positive result should be higher. The [WOAH terrestrial animal health code](https://www.woah.org/en/what-we-do/standards/codes-and-manuals/terrestrial-code-online-access/) specifies confirmatory testing requirements that reflect this asymmetry.

In companion animal practice, the cost of a false negative may outweigh the cost of a false positive, particularly for zoonotic diseases or conditions where early treatment changes outcome. The clinician should therefore select likelihood ratio thresholds that reflect the relative consequences of each error type. The [MSD Veterinary Manual](https://www.msdvetmanual.com/) provides species-specific guidance on test interpretation that incorporates these considerations.

Equipment availability also changes the diagnostic pathway. A practice with in-house PCR can apply a high positive likelihood ratio test immediately. A practice that must send samples to an external laboratory faces a delay, during which the pretest probability may shift as the disease progresses. The likelihood ratio remains constant, but the clinical decision point moves.

## Worked Clinical Example

A 7 year old neutered male Labrador Retriever presents with polyuria, polydipsia, and weight loss. Physical examination reveals poor hair coat and mild hepatomegaly. The clinician estimates a pretest probability of hyperadrenocorticism at 40 percent based on signalment and clinical signs.

An ACTH stimulation test is performed. Published data for this test in dogs with hyperadrenocorticism report a positive likelihood ratio of approximately 3.5 and a negative likelihood ratio of approximately 0.2, though these figures vary with the cortisol cutoff used and the reference laboratory.

The pretest odds at 40 percent probability are 0.67. A positive ACTH stimulation test with a likelihood ratio of 3.5 yields posttest odds of 2.33, corresponding to a posttest probability of 70 percent. This supports proceeding with treatment or confirmatory imaging. A negative result with a likelihood ratio of 0.2 yields posttest odds of 0.13, corresponding to a posttest probability of 12 percent. This effectively rules out hyperadrenocorticism and redirects the workup toward other causes.

The same test in a dog with a pretest probability of 10 percent behaves differently. A positive result yields a posttest probability of 28 percent, which does not justify treatment and warrants additional testing. This illustrates why likelihood ratios cannot be interpreted in isolation from the pretest estimate.

## Documentation and Communication

The clinical record should document the pretest probability, the likelihood ratio applied, and the resulting posttest probability. This transparency allows other clinicians to reconstruct the diagnostic reasoning and to challenge the assumptions if new information emerges. It also supports audit of diagnostic accuracy over time.

When communicating results to colleagues or in referral correspondence, state the posttest probability explicitly instead of relying on qualitative terms such as "likely" or "unlikely." The numeric estimate conveys the degree of certainty and the residual uncertainty that remains after testing. This practice aligns with the [AVMA professional practice resources](https://www.avma.org/resources-tools) emphasis on clear clinical communication and defensible medical records.

Table 1 summarizes the interpretation framework for likelihood ratios in veterinary practice.

| Likelihood ratio value | Interpretation | Typical clinical action |
|---|---|---|
| Greater than 10 | Large increase in disease probability | Strong evidence to treat or confirm with gold standard |
| 5 to 10 | Moderate increase in disease probability | Supports proceeding with definitive management |
| 2 to 5 | Small increase in disease probability | Consider additional testing before committing |
| 1 to 2 | Minimal increase in disease probability | Little diagnostic value added |
| 0.5 to 1 | Minimal decrease in disease probability | Little diagnostic value added |
| 0.2 to 0.5 | Small decrease in disease probability | Consider additional testing before excluding |
| 0.1 to 0.2 | Moderate decrease in disease probability | Supports excluding the disease |
| Less than 0.1 | Large decrease in disease probability | Strong evidence to exclude the disease |

The thresholds in Table 1 follow the conventional bands used in clinical epidemiology. They are guides, not rigid rules. The action taken at each band depends on the consequences of misclassification for the specific patient and clinical context.

## Recognized Failure Modes and Early Detection

Likelihood ratio interpretation fails in predictable ways. The most consequential failure is the assumption that sensitivity and specificity are fixed properties of a test. They vary with disease stage, pathogen strain, host species, and laboratory technique. A test validated in dairy cattle with chronic paratuberculosis will not perform identically in young beef calves with acute disease. When the underlying sensitivity and specificity shift, the likelihood ratios derived from them shift as well, and the post-test probability calculation becomes unreliable.

A second failure mode is the uncritical application of published likelihood ratios to a different population. Likelihood ratios are population-dependent in ways that sensitivity and specificity are not always recognized to be. The distribution of disease severity, the prevalence of cross-reacting conditions, and the spectrum of non-diseased animals all influence the ratio. A positive likelihood ratio of 20 reported in a referral hospital population may be 8 in first-opinion practice, where the non-diseased group includes more animals with mild or early disease that still produces positive results.

Conditional dependence between sequentially applied tests is a third failure mode. Two tests that measure the same biological pathway, such as two serological assays targeting the same immunoglobulin class, violate the independence assumption embedded in sequential likelihood ratio multiplication. The error compounds with each additional dependent test, producing post-test probabilities that are overconfident in both directions.

Early detection of these failures requires discipline. Record the source population for every likelihood ratio you use. When the pre-test probability estimate and the test result produce a post-test probability that conflicts with clinical judgment, stop and question the inputs instead of the judgment. Recalculate using a range of plausible pre-test probabilities to see whether the clinical decision changes. If it does, the test result is not robust enough to drive the decision alone.

## Common Errors and Corrective Action

Less experienced clinicians make characteriztic errors. The most common is treating a negative likelihood ratio as the reciprocal of the positive ratio. A test with a positive likelihood ratio of 10 does not automatically have a negative likelihood ratio of 0.1. The negative ratio must be calculated from the false negative rate and true negative rate directly.

A second error is using the likelihood ratio to replace clinical reasoning instead of refine it. The ratio revises a pre-test probability, it does not generate one. A clinician who cannot articulate why the test was performed, in terms of a specific differential diagnosis and a probability estimate, should not be surprised when the ratio produces an unusable answer.

A third error is rounding too early. Likelihood ratios are often reported to one decimal place, but the odds conversion sequence amplifies rounding error. Carrying the calculation to two or three significant figures through the odds conversion and back to probability preserves the discrimination the ratio provides.

The corrective action for all three errors is the same: write the calculation down, state the pre-test probability explicitly, and check the result against the clinical picture. If the post-test probability does not change the management plan, the test was either unnecessary or the ratio was misapplied.

## Limitations of the Evidence Base

The veterinary literature on likelihood ratios is thinner than the human literature. Many published ratios come from single validation studies with modest sample sizes, and few have been replicated across independent populations. The confidence intervals around sensitivity and specificity estimates are often wide, and the likelihood ratios derived from them inherit that uncertainty. A ratio reported as 12 with a 95% confidence interval from 4 to 36 is not the same evidentiary commodity as a ratio of 12 with a tight interval.

Expert opinion still differs on several points. One is whether likelihood ratios should be preferred over predictive values for communicating results to owners. Another is whether the same ratio can be applied across breeds within a species when disease prevalence and severity differ. A third is how to handle tests with likelihood ratios that straddle the clinical decision threshold, where the post-test probability lands in a grey zone that neither rules in nor rules out the diagnosis. In these cases, the honest answer is that the test has not resolved the clinical question, and further testing or empirical treatment is justified.

The WOAH surveillance standards and the CDC epidemiology curriculum both treat likelihood ratios as part of a broader statistical toolkit instead of as standalone decision rules, and that framing is correct for veterinary use as well.

## Escalation and Referral Circumstances

Referral or specialist consultation is warranted when the post-test probability remains ambiguous after sequential testing, when the test result conflicts with a strongly supported clinical diagnosis, or when the clinician lacks confidence in the pre-test probability estimate. Laboratory involvement is appropriate when the likelihood ratio itself is in question, such as when a new test lot, new analyzer, or new laboratory produces results that do not match the published performance characteriztics. The laboratory should be asked for its own validation data, including the population from which sensitivity and specificity were derived.

Regulatory reporting obligations attach to specific diseases, not to likelihood ratio calculations. When a test result, interpreted through a likelihood ratio, raises suspicion of a notifiable disease, the reporting obligation follows the disease suspicion, not the statistical confidence in the result. The WOAH terrestrial animal health code and national veterinary authorities define which diseases require notification, and those obligations apply regardless of whether the post-test probability is 60% or 95%.

| Observation | Likely cause | Discriminating check |
|---|---|---|
| Post-test probability conflicts with clinical picture | Pre-test probability misestimated | Recalculate across a plausible range of pre-test probabilities |
| Sequential tests produce overconfident result | Conditional dependence between tests | Check whether tests share a biological pathway, use panel likelihood ratios instead |
| Published ratio does not reproduce in practice | Population mismatch between validation and clinical setting | Compare disease spectrum and severity between the two populations |
| Negative result does not lower probability as expected | Negative likelihood ratio miscalculated | Recalculate from false negative and true negative rates directly |
| Ratio changes with each new test batch | Laboratory drift or changed test formulation | Request laboratory validation data for the current test lot |

## Frequently Asked Questions

### How Do I Choose Between Likelihood Ratios and Predictive Values When Reporting Results to a Referring Veterinarian?

Predictive values communicate the probability of disease after a test in a single number, which is intuitive for a busy colleague. Likelihood ratios are preferable when the pretest probability in the referring practice differs from yours, because predictive values do not transfer across populations. Report the likelihood ratio with the pretest probability you assumed, then state the resulting posttest probability. This allows the referring veterinarian to recalculate for their own caseload. The [CDC principles of epidemiology in public health practice](https://www.cdc.gov/csels/dsepd/ss1978/index.html) describe this dependence of predictive values on prevalence, which is the central reason likelihood ratios are more portable.

### What Should I Do When the Published Likelihood Ratio Comes From a Population That Differs From My Patients?

Apply the ratio with caution and document your reasoning. Likelihood ratios are derived from sensitivity and specificity estimates, and both can shift with disease severity, stage, and comorbid conditions. If your hospital sees referral cases with advanced disease, the positive likelihood ratio may be higher than published estimates, while the negative ratio may be less informative. Compare your local data if you have a sufficient case volume. For production animal work, consult [WOAH terrestrial animal health standards](https://www.woah.org/en/what-we-do/standards/codes-and-manuals/terrestrial-code-online-access/) when test performance affects movement or trade decisions, because regulatory contexts may require specific test characteriztics regardless of local estimates.

### How Do I Handle Likelihood Ratios When Only a Qualitative Result Is Available?

Use the likelihood ratio for the qualitative category that matches the result. Many point-of-care assays report positive or negative only, and the ratio for that category is the appropriate multiplier. Avoid converting a qualitative result into a quantitative likelihood ratio by interpolation. If the assay provides a numerical readout but validation studies only report binary results, treat the result as binary. For surveillance contexts where test interpretation feeds into population-level decisions, the [WOAH animal health surveillance standards](https://www.woah.org/en/what-we-do/animal-health-and-welfare/disease-data-collection/) provide guidance on how test performance characteriztics should be reported and applied across different surveillance objectives.

### What Are the Practical Limits of Using Likelihood Ratios in a General Practice Setting?

The main constraint is obtaining a defensible pretest probability. Published prevalence figures may not match your local population, and subjective estimates carry uncertainty. A practical approach is to use a range of pretest probabilities and report the corresponding range of posttest probabilities. This is more honest than a single false-precision number. Time pressure is another limit, but the odds conversion sequence takes under a minute with a calculator. When the ideal test is unavailable, choose the test with the most favourable likelihood ratio for the clinical question and state the limitation in the record. The [MSD Veterinary Manual professional edition](https://www.msdvetmanual.com/) provides species-specific guidance on test availability and interpretation that can inform these decisions.

### How Should I Document Likelihood Ratio Calculations in the Medical Record?

Record the pretest probability, the likelihood ratio used, and the resulting posttest probability. Note the source of the likelihood ratio, whether from a published study, local validation, or a reference standard. If you used a range of pretest probabilities, record the range of posttest probabilities. State any assumptions about disease prevalence or spectrum. This documentation supports later review if the case outcome differs from the predicted probability. The [AVMA practice resources](https://www.avma.org/resources-tools) include guidance on medical record standards that apply to diagnostic reasoning documentation, and consistent recording of probability revisions supports both clinical continuity and medicolegal defensibility.

### How Do I Explain Likelihood Ratios to a Client Without Oversimplifying the Uncertainty?

Frame the conversation around what the test result changes about the plan. Explain that the test result makes the suspected condition more or less likely, but does not confirm or exclude it. Use a concrete statement such as "this result increases the chance that your cat has kidney disease from about 30 percent to roughly 60 percent, so we should proceed with the next diagnostic step." Avoid giving a single precise percentage when the pretest estimate is uncertain. Acknowledge that test performance varies with disease stage and that the interpretation may be revised as more information becomes available. This approach preserves the client's understanding of residual uncertainty while giving them a clear rationale for the recommended next step.

## Related Clinical & Scientific Guides

* [Evaluating Veterinary Surveillance System Attributes](/knowledge/veterinary-medicine/veterinary-epidemiology/evaluating-veterinary-surveillance-system-attributes)
* [Network Analysis for Infectious Disease Spread in Animal Populations](/knowledge/veterinary-medicine/veterinary-epidemiology/network-analysis-infectious-disease-spread-animal-populations)
* [Randomized Controlled Trials in Veterinary Field Settings](/knowledge/veterinary-medicine/veterinary-epidemiology/randomized-controlled-trials-veterinary-field-settings)


## References and Further Reading

- [Night Shift Work and Breast Cancer Incidence: Three Prospective Studies and Meta-analysis of Published Studies.](https://pubmed.ncbi.nlm.nih.gov/27758828/). 2016.
- [Real-time surveillance to assess risk of intussusception and other adverse events after pentavalent, bovine-derived rotavirus vaccine.](https://pubmed.ncbi.nlm.nih.gov/19907356/). 2010.
- [Aspirin and nonsteroidal anti-inflammatory drug use and the risk of subsequent colorectal cancer.](https://pubmed.ncbi.nlm.nih.gov/8117171/). 1994.
- [Radiofrequency versus microwave ablation in a hepatic porcine model.](https://pubmed.ncbi.nlm.nih.gov/15987969/). 2005.
- [A testing framework for identifying susceptibility genes in the presence of epistasis.](https://pubmed.ncbi.nlm.nih.gov/16385446/). 2006.
- [Properties of different selection signature statistics and a new strategy for combining them.](https://pubmed.ncbi.nlm.nih.gov/25990878/). 2015.
- [WOAH Animal Health Surveillance Standards](https://www.woah.org/en/what-we-do/animal-health-and-welfare/disease-data-collection/). WOAH.
- [CDC Principles of Epidemiology in Public Health Practice](https://www.cdc.gov/csels/dsepd/ss1978/index.html). CDC.
- [MSD Veterinary Manual, Professional Edition](https://www.msdvetmanual.com/). MSD Veterinary Manual.

## Related Articles

- [Measures of Association in Veterinary Epidemiology: Risk and Odds Ratios](/knowledge/veterinary-medicine/veterinary-epidemiology/measures-association-veterinary-epidemiology-risk-odds-ratios)
- [Basic Reproductive Ratio (R0) in Veterinary Epidemiology](/knowledge/veterinary-medicine/veterinary-epidemiology/basic-reproductive-ratio-r0-veterinary-epidemiology)
- [Bayesian Methods for Diagnostic Test Evaluation in Animals](/knowledge/veterinary-medicine/veterinary-epidemiology/bayesian-methods-diagnostic-test-evaluation-animals)
- [Diagnostic Test Evaluation: Sensitivity and Specificity in Veterinary Medicine](/knowledge/veterinary-medicine/veterinary-epidemiology/diagnostic-test-evaluation-sensitivity-specificity-veterinary-medicine)
- [Statistical Inference in Veterinary Epidemiology: Hypothesis Testing and Confidence Intervals](/knowledge/veterinary-medicine/veterinary-epidemiology/statistical-inference-veterinary-epidemiology-hypothesis-testing-confidence-intervals)

> 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.