The Gold Standard of Zoonoses | Separating Host-Level Shedding Prevalence from Sample-Level Positivity

A host, visit, specimen and assay result occupy different rows in the evidence chain. Host-level questions require hosts in the denominator; specimen-level questions require eligible specimens. Repeated specimens can improve temporal coverage, but they do not create new animals.

Begin with a measurement ledger. Name the quantity, unit, observational unit, time window and transformation rule. Mark every input as observed, stipulated, estimated or assumed. Read the ledger as a sentence before calculating; if the sentence changes when the denominator changes, the arithmetic belongs to a different claim. Keep the raw rows beside every derived column so another reader can reproduce the step. Record exclusions and missing values rather than letting software silently shorten the denominator. If two sources use the same label differently, preserve both definitions and postpone the comparison until they can be reconciled.

Ten fictional hosts each provide three scheduled specimens. Thirty assay results exist, but the host denominator remains ten. A report that says n=30 without naming specimens invites a reader to mistake repeated observations for independent animals.

Before moving on, record the observation that would distinguish the two most plausible explanations. If no feasible observation exists, narrow the claim instead of pretending the ambiguity has disappeared.

A changed-context transfer check is to collapse the table to one row per host and identify which information is lost State what would change the result and what the example cannot establish. That limit belongs inside the answer.

Contents · Next chapter

Construct both summaries from the same ledger instead of arguing over one preferred percentage. The two results can both be correct because their numerators and denominators are defined differently.

Keep the mechanism beside the number. Ask what physical, biological, statistical or protocol relationship connects the observation to the conclusion, and which conditions could break that link. Draw one arrow for each inferential step, then challenge every arrow with a plausible changed setting. A mechanism sketch should identify where selection, dependence, timing or an external constraint enters. It should also show the point at which the available evidence stops. This makes the next measurement purposeful: it tests one uncertain link instead of collecting more convenient data from a link that is already well described.

Across thirty specimens, twelve are positive, so observed specimen positivity is 12/30=40%. Six of ten hosts have one or more positive specimens, so host-level ever-observed positivity is 6/10=60%. The second result depends on the three-visit observation window.

Before moving on, record the observation that would distinguish the two most plausible explanations. If no feasible observation exists, narrow the claim instead of pretending the ambiguity has disappeared.

A changed-context transfer check is to add a fourth visit and show why the ever-positive numerator can rise even if the biology is unchanged State what would change the result and what the example cannot establish. That limit belongs inside the answer.

Contents · Previous chapter · Next chapter

The sequence of results distinguishes persistent observed positivity, intermittent detection and isolated findings. A single pooled proportion erases this pattern and can conceal dependence within hosts.

Use an independent check that follows a different route. Reconstruct totals from components, compare a bound with a point estimate, inspect a distribution behind an average, or solve a changed case from the original definitions. Disagreement is useful because it locates the definition, unit or boundary that needs repair. Preserve the first result while investigating rather than overwriting it. The audit trail should show which input or rule resolved the difference. Agreement between two calculations is strongest when they do not merely repeat the same assumption in different notation.

Host A has +++; Host B has +–; Host C has -+-; four other hosts contain positives and three are —. Their contributions to 40% specimen positivity differ, and their biological interpretations remain uncertain without method-specific evidence.

Before moving on, record the observation that would distinguish the two most plausible explanations. If no feasible observation exists, narrow the claim instead of pretending the ambiguity has disappeared.

A changed-context transfer check is to shuffle visit order and state which summaries remain identical and which temporal claim disappears State what would change the result and what the example cannot establish. That limit belongs inside the answer.

Contents · Previous chapter · Next chapter

A positive result is an assay-defined observation from a stated specimen. It may support evidence of target material or viable agent according to the method, yet the word shedding must match specimen type, route, analytical target and validation.

Match the claim to the evidence. A tidy decimal cannot repair selective observation, dependence or a missing protocol condition. Report the result, the conditions that make it interpretable and the next observation needed for the actual decision. Keep the numerical conclusion separate from the practical conclusion. Add a brief boundary statement naming the population, representation, system or interval to which the result applies. Then name one tempting stronger claim that the evidence does not support. This pair of sentences prevents a precise local answer from becoming an unsupported general rule.

A respiratory swab and a faecal specimen from the same host are not interchangeable route measurements. Pooling them as two generic shedding tests could answer no coherent route-specific question even when both assays are technically valid.

Before moving on, record the observation that would distinguish the two most plausible explanations. If no feasible observation exists, narrow the claim instead of pretending the ambiguity has disappeared.

A changed-context transfer check is to hold the host fixed but change specimen type and rewrite the endpoint State what would change the result and what the example cannot establish. That limit belongs inside the answer.

Contents · Previous chapter · Next chapter

Ever-positive during three visits is a cumulative observation over a window. Point positivity at visit two is a cross-sectional result. Longer follow-up usually offers more opportunities to observe at least one positive result.

Begin with a measurement ledger. Name the quantity, unit, observational unit, time window and transformation rule. Mark every input as observed, stipulated, estimated or assumed. Read the ledger as a sentence before calculating; if the sentence changes when the denominator changes, the arithmetic belongs to a different claim. Keep the raw rows beside every derived column so another reader can reproduce the step. Record exclusions and missing values rather than letting software silently shorten the denominator. If two sources use the same label differently, preserve both definitions and postpone the comparison until they can be reconciled.

At visit two, three of ten hosts are positive, giving 30% visit-specific host positivity. Across all three visits, six hosts are ever positive, giving 60%. The difference does not prove that prevalence doubled.

Before moving on, record the observation that would distinguish the two most plausible explanations. If no feasible observation exists, narrow the claim instead of pretending the ambiguity has disappeared.

A changed-context transfer check is to shorten follow-up to one visit and identify why direct comparison becomes unfair State what would change the result and what the example cannot establish. That limit belongs inside the answer.

Contents · Previous chapter · Next chapter

The planned denominator, collected denominator and valid-result denominator can differ. Missing or invalid specimens need reasons because their absence may relate to host condition, capture difficulty or specimen quality.

Keep the mechanism beside the number. Ask what physical, biological, statistical or protocol relationship connects the observation to the conclusion, and which conditions could break that link. Draw one arrow for each inferential step, then challenge every arrow with a plausible changed setting. A mechanism sketch should identify where selection, dependence, timing or an external constraint enters. It should also show the point at which the available evidence stops. This makes the next measurement purposeful: it tests one uncertain link instead of collecting more convenient data from a link that is already well described.

Thirty specimens are scheduled, twenty-eight collected and twenty-six yield valid results; eleven valid results are positive. The transparent proportion is 11/26 among valid specimens, accompanied by the two missing and two invalid outcomes.

Before moving on, record the observation that would distinguish the two most plausible explanations. If no feasible observation exists, narrow the claim instead of pretending the ambiguity has disappeared.

A changed-context transfer check is to treat every missing result as negative and explain the unsupported assumption State what would change the result and what the example cannot establish. That limit belongs inside the answer.

Contents · Previous chapter · Next chapter

When hosts contribute unequal numbers of specimens, a pooled specimen proportion weights frequent contributors more heavily. A host-level summary or model can answer a different population question.

Use an independent check that follows a different route. Reconstruct totals from components, compare a bound with a point estimate, inspect a distribution behind an average, or solve a changed case from the original definitions. Disagreement is useful because it locates the definition, unit or boundary that needs repair. Preserve the first result while investigating rather than overwriting it. The audit trail should show which input or rule resolved the difference. Agreement between two calculations is strongest when they do not merely repeat the same assumption in different notation.

Host A contributes ten specimens with eight positives; nine other hosts contribute one each with one positive in total. Pooled positivity is 9/19, while two of ten hosts are ever positive. Neither denominator should be hidden.

Before moving on, record the observation that would distinguish the two most plausible explanations. If no feasible observation exists, narrow the claim instead of pretending the ambiguity has disappeared.

A changed-context transfer check is to cap every host at one randomly selected specimen and describe the resulting estimand State what would change the result and what the example cannot establish. That limit belongs inside the answer.

Contents · Previous chapter · Next chapter

Results from the same host can be correlated through biology, environment, collection and assay conditions. Treating them as independent can understate uncertainty and exaggerate effective sample size.

Match the claim to the evidence. A tidy decimal cannot repair selective observation, dependence or a missing protocol condition. Report the result, the conditions that make it interpretable and the next observation needed for the actual decision. Keep the numerical conclusion separate from the practical conclusion. Add a brief boundary statement naming the population, representation, system or interval to which the result applies. Then name one tempting stronger claim that the evidence does not support. This pair of sentences prevents a precise local answer from becoming an unsupported general rule.

Three specimens from each of ten hosts produce thirty binary rows. A naive calculation that treats all thirty as unrelated observations ignores shared host effects. A host-clustered analysis preserves the grouping even when reporting specimen-level positivity.

Before moving on, record the observation that would distinguish the two most plausible explanations. If no feasible observation exists, narrow the claim instead of pretending the ambiguity has disappeared.

A changed-context transfer check is to increase specimens per host from three to thirty without adding hosts and explain why population coverage barely changes State what would change the result and what the example cannot establish. That limit belongs inside the answer.

Contents · Previous chapter · Next chapter

Several consecutive positives can belong to one shedding episode, while positives separated by negatives may reflect intermittent shedding, assay variability or separate episodes. An episode rule must be predeclared and biologically justified.

Begin with a measurement ledger. Name the quantity, unit, observational unit, time window and transformation rule. Mark every input as observed, stipulated, estimated or assumed. Read the ledger as a sentence before calculating; if the sentence changes when the denominator changes, the arithmetic belongs to a different claim. Keep the raw rows beside every derived column so another reader can reproduce the step. Record exclusions and missing values rather than letting software silently shorten the denominator. If two sources use the same label differently, preserve both definitions and postpone the comparison until they can be reconciled.

A host has results +,+,-,+ over four visits. Counting three positive specimens is simple; counting one or two episodes depends on the gap rule and visit spacing. The ledger should retain the raw sequence beside any derived episode count.

Before moving on, record the observation that would distinguish the two most plausible explanations. If no feasible observation exists, narrow the claim instead of pretending the ambiguity has disappeared.

A changed-context transfer check is to double the gap between visits and reconsider the episode rule State what would change the result and what the example cannot establish. That limit belongs inside the answer.

Contents · Previous chapter · Next chapter

The start and end of shedding may lie between visits. Scheduled sampling brackets transitions but rarely observes the exact biological time, so duration calculations need interval-aware language.

Keep the mechanism beside the number. Ask what physical, biological, statistical or protocol relationship connects the observation to the conclusion, and which conditions could break that link. Draw one arrow for each inferential step, then challenge every arrow with a plausible changed setting. A mechanism sketch should identify where selection, dependence, timing or an external constraint enters. It should also show the point at which the available evidence stops. This makes the next measurement purposeful: it tests one uncertain link instead of collecting more convenient data from a link that is already well described.

A host is negative on day 0, positive on day 7 and negative on day 14. The observed positive visit does not prove shedding began exactly on day 7 or lasted seven days. It bounds events under stated assay assumptions.

Before moving on, record the observation that would distinguish the two most plausible explanations. If no feasible observation exists, narrow the claim instead of pretending the ambiguity has disappeared.

A changed-context transfer check is to sample daily and state which uncertainty shrinks and which assay uncertainty remains State what would change the result and what the example cannot establish. That limit belongs inside the answer.

Contents · Previous chapter · Next chapter

A repeated-measures study can describe sampled hosts under its design. Population prevalence requires a target population, sampling frame and inclusion probabilities; repeated testing alone does not repair convenience selection.

Use an independent check that follows a different route. Reconstruct totals from components, compare a bound with a point estimate, inspect a distribution behind an average, or solve a changed case from the original definitions. Disagreement is useful because it locates the definition, unit or boundary that needs repair. Preserve the first result while investigating rather than overwriting it. The audit trail should show which input or rule resolved the difference. Agreement between two calculations is strongest when they do not merely repeat the same assumption in different notation.

Ten rescued symptomatic animals are sampled repeatedly. Six ever-positive animals describe that cohort, not every animal in the region. More swabs from the same cohort do not turn it into a representative population sample.

Before moving on, record the observation that would distinguish the two most plausible explanations. If no feasible observation exists, narrow the claim instead of pretending the ambiguity has disappeared.

A changed-context transfer check is to replace convenience recruitment with probability sampling and list the new evidence needed State what would change the result and what the example cannot establish. That limit belongs inside the answer.

Contents · Previous chapter · Next chapter

The related article Why Reservoir Prevalence Does Not Equal Spillover Risk keeps host density, shedding, contact and recipient susceptibility distinct. Repeated shedding results refine one upstream step only.

Match the claim to the evidence. A tidy decimal cannot repair selective observation, dependence or a missing protocol condition. Report the result, the conditions that make it interpretable and the next observation needed for the actual decision. Keep the numerical conclusion separate from the practical conclusion. Add a brief boundary statement naming the population, representation, system or interval to which the result applies. Then name one tempting stronger claim that the evidence does not support. This pair of sentences prevents a precise local answer from becoming an unsupported general rule.

A higher ever-positive fraction can still coexist with little route-relevant contact. Conversely, brief shedding during intense contact may matter more than a longer episode outside the interface period. The study cannot supply missing contact evidence.

Before moving on, record the observation that would distinguish the two most plausible explanations. If no feasible observation exists, narrow the claim instead of pretending the ambiguity has disappeared.

A changed-context transfer check is to hold shedding observations fixed while moving the contact season State what would change the result and what the example cannot establish. That limit belongs inside the answer.

Contents · Previous chapter · Next chapter

Observed positive and negative results depend on assay performance in the relevant specimen and setting. Repetition can change the chance of at least one positive, including the chance of at least one false positive.

Begin with a measurement ledger. Name the quantity, unit, observational unit, time window and transformation rule. Mark every input as observed, stipulated, estimated or assumed. Read the ledger as a sentence before calculating; if the sentence changes when the denominator changes, the arithmetic belongs to a different claim. Keep the raw rows beside every derived column so another reader can reproduce the step. Record exclusions and missing values rather than letting software silently shorten the denominator. If two sources use the same label differently, preserve both definitions and postpone the comparison until they can be reconciled.

If specificity is imperfect, three opportunities per host can raise the probability of at least one false-positive observation. An ever-positive rule therefore needs assay context rather than being treated as an error-free infection label.

Before moving on, record the observation that would distinguish the two most plausible explanations. If no feasible observation exists, narrow the claim instead of pretending the ambiguity has disappeared.

A changed-context transfer check is to raise the number of repeated tests and compare the qualitative effect on false-positive opportunity State what would change the result and what the example cannot establish. That limit belongs inside the answer.

Contents · Previous chapter · Next chapter

Pooling combines material from multiple units before testing and changes the link between a result and individual hosts. One positive pool does not identify every contributor as positive.

Keep the mechanism beside the number. Ask what physical, biological, statistical or protocol relationship connects the observation to the conclusion, and which conditions could break that link. Draw one arrow for each inferential step, then challenge every arrow with a plausible changed setting. A mechanism sketch should identify where selection, dependence, timing or an external constraint enters. It should also show the point at which the available evidence stops. This makes the next measurement purposeful: it tests one uncertain link instead of collecting more convenient data from a link that is already well described.

Five hosts contribute to one pool that tests positive. The pool is one tested unit; it cannot be entered as five positive hosts or five positive specimens. Follow-up design determines what can be resolved.

Before moving on, record the observation that would distinguish the two most plausible explanations. If no feasible observation exists, narrow the claim instead of pretending the ambiguity has disappeared.

A changed-context transfer check is to split the pool into individual tests and preserve the chain of custody State what would change the result and what the example cannot establish. That limit belongs inside the answer.

Contents · Previous chapter · Next chapter

Technical replicates estimate analytical repeatability around one specimen; they are not new field specimens or new hosts. Preserve their level in the data hierarchy.

Use an independent check that follows a different route. Reconstruct totals from components, compare a bound with a point estimate, inspect a distribution behind an average, or solve a changed case from the original definitions. Disagreement is useful because it locates the definition, unit or boundary that needs repair. Preserve the first result while investigating rather than overwriting it. The audit trail should show which input or rule resolved the difference. Agreement between two calculations is strongest when they do not merely repeat the same assumption in different notation.

One swab is run three times and yields +,+,-. Reporting two positives out of three technical runs is different from three specimens across time. A decision rule for the specimen must be validated and stated.

Before moving on, record the observation that would distinguish the two most plausible explanations. If no feasible observation exists, narrow the claim instead of pretending the ambiguity has disappeared.

A changed-context transfer check is to replace technical replicates with three independently collected specimens State what would change the result and what the example cannot establish. That limit belongs inside the answer.

Contents · Previous chapter · Next chapter

Sampling hosts again because an earlier result was positive makes observation counts outcome-dependent. Naive pooled proportions then mix biological results with the study’s follow-up rule.

Match the claim to the evidence. A tidy decimal cannot repair selective observation, dependence or a missing protocol condition. Report the result, the conditions that make it interpretable and the next observation needed for the actual decision. Keep the numerical conclusion separate from the practical conclusion. Add a brief boundary statement naming the population, representation, system or interval to which the result applies. Then name one tempting stronger claim that the evidence does not support. This pair of sentences prevents a precise local answer from becoming an unsupported general rule.

Positive hosts receive five later specimens while negative hosts receive none. The resulting specimen pool overrepresents previously positive hosts by design. Report the adaptive rule and analyse the intended estimand.

Before moving on, record the observation that would distinguish the two most plausible explanations. If no feasible observation exists, narrow the claim instead of pretending the ambiguity has disappeared.

A changed-context transfer check is to apply the same visit schedule to all hosts and compare weighting State what would change the result and what the example cannot establish. That limit belongs inside the answer.

Contents · Previous chapter · Next chapter

If hosts enter and leave the study, the set at risk changes. An ever-positive proportion over all enrolled hosts differs from visit-specific positivity among hosts observed at that visit.

Begin with a measurement ledger. Name the quantity, unit, observational unit, time window and transformation rule. Mark every input as observed, stipulated, estimated or assumed. Read the ledger as a sentence before calculating; if the sentence changes when the denominator changes, the arithmetic belongs to a different claim. Keep the raw rows beside every derived column so another reader can reproduce the step. Record exclusions and missing values rather than letting software silently shorten the denominator. If two sources use the same label differently, preserve both definitions and postpone the comparison until they can be reconciled.

Ten hosts start, two leave and three new hosts enter. Visit-three positivity should use the eligible observed hosts then; a cumulative ever-positive measure needs a documented cohort rule.

Before moving on, record the observation that would distinguish the two most plausible explanations. If no feasible observation exists, narrow the claim instead of pretending the ambiguity has disappeared.

A changed-context transfer check is to restrict analysis to the original cohort and describe the change in generalisability State what would change the result and what the example cannot establish. That limit belongs inside the answer.

Contents · Previous chapter · Next chapter

A confidence interval or model should reflect the sampling design, number of hosts, repeated observations and chosen endpoint. Thirty rows do not justify the uncertainty of thirty independent hosts.

Keep the mechanism beside the number. Ask what physical, biological, statistical or protocol relationship connects the observation to the conclusion, and which conditions could break that link. Draw one arrow for each inferential step, then challenge every arrow with a plausible changed setting. A mechanism sketch should identify where selection, dependence, timing or an external constraint enters. It should also show the point at which the available evidence stops. This makes the next measurement purposeful: it tests one uncertain link instead of collecting more convenient data from a link that is already well described.

The 6/10 host summary has only ten independent host units under the simplified design. The 12/30 specimen result has clustered observations. Report both counts and avoid a precision claim borrowed from the larger row count.

Before moving on, record the observation that would distinguish the two most plausible explanations. If no feasible observation exists, narrow the claim instead of pretending the ambiguity has disappeared.

A changed-context transfer check is to compare a cluster bootstrap with a row bootstrap conceptually State what would change the result and what the example cannot establish. That limit belongs inside the answer.

Contents · Previous chapter · Next chapter

Use stable host IDs, visit dates, specimen types, collection status, assay method, result, validity and protocol deviations. Derived host flags should be reproducible from the preserved rows.

Use an independent check that follows a different route. Reconstruct totals from components, compare a bound with a point estimate, inspect a distribution behind an average, or solve a changed case from the original definitions. Disagreement is useful because it locates the definition, unit or boundary that needs repair. Preserve the first result while investigating rather than overwriting it. The audit trail should show which input or rule resolved the difference. Agreement between two calculations is strongest when they do not merely repeat the same assumption in different notation.

A reviewer recomputes the six ever-positive hosts and twelve positive specimens from the table. If a result changes after excluding invalid runs, the exclusion rule and both totals remain visible.

Before moving on, record the observation that would distinguish the two most plausible explanations. If no feasible observation exists, narrow the claim instead of pretending the ambiguity has disappeared.

A changed-context transfer check is to change the validity rule and rerun every derived count State what would change the result and what the example cannot establish. That limit belongs inside the answer.

Contents · Previous chapter · Next chapter

More specimens do not equal more hosts; ever-positive is not point prevalence; a negative visit does not prove a host never shed; and specimen positivity is not spillover risk. Each repair restores a unit or pathway step.

Match the claim to the evidence. A tidy decimal cannot repair selective observation, dependence or a missing protocol condition. Report the result, the conditions that make it interpretable and the next observation needed for the actual decision. Keep the numerical conclusion separate from the practical conclusion. Add a brief boundary statement naming the population, representation, system or interval to which the result applies. Then name one tempting stronger claim that the evidence does not support. This pair of sentences prevents a precise local answer from becoming an unsupported general rule.

When a report says 40% positive, ask: forty percent of hosts, visits, specimens, pools or assay runs? Then ask over what period and with what missing-result rule.

Before moving on, record the observation that would distinguish the two most plausible explanations. If no feasible observation exists, narrow the claim instead of pretending the ambiguity has disappeared.

A changed-context transfer check is to remove host identifiers and show why several questions become permanently unanswerable State what would change the result and what the example cannot establish. That limit belongs inside the answer.

Contents · Previous chapter · Next chapter

A useful conclusion states the sampled population, schedule, valid denominators, endpoint and uncertainty. It can report two complementary summaries without declaring one the true answer to every question.

Begin with a measurement ledger. Name the quantity, unit, observational unit, time window and transformation rule. Mark every input as observed, stipulated, estimated or assumed. Read the ledger as a sentence before calculating; if the sentence changes when the denominator changes, the arithmetic belongs to a different claim. Keep the raw rows beside every derived column so another reader can reproduce the step. Record exclusions and missing values rather than letting software silently shorten the denominator. If two sources use the same label differently, preserve both definitions and postpone the comparison until they can be reconciled.

Write: ‘Among ten sampled hosts observed at three scheduled visits, six were positive at least once; twelve of thirty valid fictional specimens were positive.’ Then add limits on selection, assay interpretation and repeated measures.

Before moving on, record the observation that would distinguish the two most plausible explanations. If no feasible observation exists, narrow the claim instead of pretending the ambiguity has disappeared.

A changed-context transfer check is to replace all counts with percentages and identify the transparency lost State what would change the result and what the example cannot establish. That limit belongs inside the answer.

Contents · Previous chapter · Next chapter

The WOAH wildlife-health programme explains the role of disease surveillance in wild animals. The CDC epidemiology lesson on ratios and proportions shows why numerators and denominators must be named. Return to the Zoonoses root for the wider One Health pathway.

Keep the mechanism beside the number. Ask what physical, biological, statistical or protocol relationship connects the observation to the conclusion, and which conditions could break that link. Draw one arrow for each inferential step, then challenge every arrow with a plausible changed setting. A mechanism sketch should identify where selection, dependence, timing or an external constraint enters. It should also show the point at which the available evidence stops. This makes the next measurement purposeful: it tests one uncertain link instead of collecting more convenient data from a link that is already well described.

The transferable habit is to keep nested observation units visible. The same reasoning applies to pupils with repeated tests, machines with repeated readings and households with repeated visits.

Before moving on, record the observation that would distinguish the two most plausible explanations. If no feasible observation exists, narrow the claim instead of pretending the ambiguity has disappeared.

A changed-context transfer check is to apply the ledger to repeated water samples from the same tap without importing animal-health conclusions State what would change the result and what the example cannot establish. That limit belongs inside the answer.

Contents · Previous chapter · Return to Zoonoses