Vaccination coverage is a fraction with a moving denominator

#Question

Why can a vaccination coverage estimate be precise to one decimal place and still be epidemiologically uncertain?

#Answer in brief

Administrative vaccination coverage is usually a ratio: doses recorded for a target group divided by an estimate of the target population. WHO explicitly notes that both parts can be wrong. Numerators can be incomplete or inflated, while denominators can be distorted by population movement, census projections or competing population sources.^[1] WHO and UNICEF therefore assess coverage using multiple data sources rather than treating the administrative percentage as ground truth.^[2]

The apparent precision of the percentage does not remove uncertainty about who should have been counted.

#The familiar fraction

A simplified formula is:

coverage = vaccinated eligible people / eligible target population

The arithmetic is easy.

The epidemiology is not.

The denominator requires a definition of who belongs to the population, where they belong, and during which period. The numerator requires a definition of which dose, which person, which record and which eligibility state count.

If those definitions drift, the fraction can change without any true change in immunization behaviour.

#Denominator error is not a footnote

WHO's administrative-method guidance states directly that denominator inaccuracies can arise from population movement, inaccurate census estimates or projections, and multiple denominator sources.^[1]

That has practical consequences.

Suppose a district administers 9,500 first doses and its estimated target population is 10,000. Coverage is reported as 95%.

If the true target population is 9,000 because of out-migration, the observed numerator exceeds the nominal population that would have been expected under the estimate. If the true population is 11,000 because of rapid growth not captured in projections, the programme reaches fewer eligible people than the headline suggests.

A coverage value above 100% is often a warning that the numerator and denominator do not describe the same population process, not evidence that more than everyone was vaccinated.

WHO's 2022 guidance on target population estimates explicitly uses implausibly high coverage, erratic fluctuations and outbreaks despite apparently high coverage as signals that denominator accuracy deserves investigation.^[3]

#Numerators can move independently too

A vaccination numerator can be affected by:

• incomplete reporting from public or private providers;
• late uploads;
• duplicate records;
• people vaccinated outside their area of residence;
• doses entered under the wrong cohort;
• catch-up doses mixed with routine on-time doses;
• data reconciliation that occurs after the reporting period.

WHO notes that administrative numerators can be underestimated when reporting is incomplete or some providers are excluded, and overestimated when doses for other target groups are included.^[1]

A clean percentage therefore requires alignment between person, dose, place, cohort and time.

#Eligibility is part of the denominator

A target population is not always "everyone aged X."

Eligibility can depend on:

• birth cohort;
• age at a reference date;
• previous doses;
• risk condition;
• pregnancy;
• occupation;
• residence;
• campaign eligibility;
• interval since an earlier dose.

When eligibility changes, the denominator changes.

Comparing coverage before and after a schedule change without reconstructing eligibility can create a false trend.

#Residence and service location are different concepts

Vaccination systems often record where a dose was delivered. Public-health monitoring often wants to know coverage among residents.

Those are not the same geography.

A person may:

• live in one municipality and vaccinate in another;
• move during the year;
• receive a dose while travelling;
• have an old residence in the registry;
• be temporarily present but not part of the resident denominator.

A service-delivery count answers, "How many doses did this provider or area administer?"

A resident coverage estimate asks, "What proportion of the eligible resident population received the dose?"

Mixing the two creates a numerator-denominator mismatch.

#Time turns a static fraction into a moving target

Coverage can be measured:

• at a fixed age;
• by the end of a calendar year;
• for a birth cohort;
• by a programme deadline;
• at a cross-sectional extraction date.

Those choices change the meaning.

A child vaccinated two months late may count as vaccinated in an annual cross-section but was unprotected during the interval when the schedule intended protection.

This is why the companion article on vaccination trajectories treats timeliness separately from final status.

#Administrative coverage and survey coverage answer related but different questions

WHO/UNICEF coverage estimation uses administrative, official and survey information. The WHO explanation notes that administrative estimates depend on reporting and target-population accuracy, while surveys have their own biases, including exclusion of hard-to-reach populations and reliance on recall when documented records are unavailable.^[2]

No data source is magically bias-free.

The value of triangulation is that different sources fail differently.

For example:

• administrative data may have good temporal granularity but weak denominator estimates;
• household surveys may better estimate population coverage but be infrequent;
• registries may support individual trajectories but depend on completeness and identity resolution.

#A synthetic local example

A municipality reports 97% coverage for a second vaccine dose among two-year-olds.

The following year coverage falls to 91%.

Before concluding that acceptance declined, analysts discover:

• the denominator was updated after a new population estimate;
• private-provider uploads became more complete;
• 400 duplicate resident records were removed;
• some children previously counted in the numerator had moved outside the municipality;
• the vaccination schedule itself did not change.

The new estimate may actually be more accurate even though it looks worse.

This is one reason trend breaks after registry cleaning should be documented rather than cosmetically smoothed.

#A denominator audit

Before interpreting a coverage percentage, ask:

#Population definition

Who is eligible, and why?

#Reference time

At what date or age is eligibility assessed?

#Geography

Residence, registration, service location, or another rule?

#Population source

Census, registry, projection, health-system enrolment, school list, or hybrid?

#Mobility

How are births, deaths, immigration, emigration and internal movement handled?

#Numerator matching

Do recorded doses belong to the same population definition?

#Identity

How are duplicates, uncertain identities and cross-provider records reconciled?

#Delay

Can late data entry change already published periods?

#Versioning

Can the analyst reconstruct which denominator was used for a historical estimate?

If those questions are undocumented, the percentage is difficult to audit.

#Coverage precision should reflect data precision

Reporting 93.7% instead of 94% can be useful when the numerator and denominator support that level of precision.

But a decimal place should not imply certainty that is absent from the population model.

When denominator uncertainty is material, the report should say so. Practical options include:

• showing alternative plausible denominators;
• flagging estimates over 100%;
• documenting population-source changes;
• separating resident and provider-based counts;
• retaining version history;
• comparing administrative estimates with independent sources.

WHO's target-population guidance exists because denominator quality can become the limiting factor once programmes reach high coverage.^[3]

#The real object is a population model

A vaccination coverage estimate is not merely a division operation.

It is the output of a model of:

• who exists;
• who is eligible;
• who received which dose;
• when it was received;
• where that person belongs;
• whether all relevant providers reported;
• how uncertain identities were resolved.

Making that model explicit improves both monitoring and communication.

#Limitations and uncertainty

This article addresses measurement principles, not a specific national immunization schedule or registry. Eligibility rules, legal population definitions, available identifiers and reporting systems differ by jurisdiction. WHO's administrative and target-population guidance is global and should be adapted to current national procedures. Survey, registry and administrative sources each have distinct biases; disagreement between them should trigger investigation rather than automatic preference for one estimate.^[2]