Childhood immunization survey: pre-data collection checklist

It fails earlier: when the target cohort is fuzzy, the household list is stale, the questionnaire cannot distinguish a card from caregiver recall, or supervisors arrive in a cluster without a route plan.

That failure is expensive. It produces a coverage number that looks precise, gets discussed at district review meetings, and then sends vaccines, outreach teams, and corrective action toward the wrong bottleneck. In child survival work, bad measurement is not a paperwork problem. It is a supply-chain problem wearing a spreadsheet.

India already has multiple immunization data streams. A district team may see high administrative coverage in routine reporting while a household survey finds a very different picture. Neither number is automatically fraudulent. They may simply be counting different populations, using different denominators, covering different periods, and relying on different evidence. The job before fieldwork begins is to make those differences visible and controlled.

This childhood immunization survey preparation checklist is built for that job: locking the design before teams start knocking on doors.

1. Define the question before defining the sample

“Immunization coverage” is not a single indicator. Treating it as one is the first operational error.

A survey team needs a written indicator specification that answers, in plain language:

• Which children are eligible? State the age range precisely. A survey of children aged 12–23 months is not interchangeable with a survey of children aged 9–35 months or children under five.
• Which geography is being estimated? District-wide? Rural blocks only? Urban slums? Tribal hamlets? A district immunization coverage assessment becomes unreliable when it quietly excludes hard-to-reach settlements but presents the result as district coverage.
• Which vaccines and dose schedule count? Use the applicable National Immunization Schedule and write the exact antigen-dose combination into the protocol.
• What is the reference period? Coverage by the first birthday, coverage at the date of interview, or doses received in the previous 12 months are different measures.
• What is the evidence hierarchy? Card only, card plus caregiver recall, or card plus registry verification if permissions and access are in place.
• What decision will the estimate drive? More fixed-site sessions, outreach microplanning, cold-chain allocation, defaulter tracing, or a deeper investigation of missed doses.

For India’s Universal Immunization Programme, a child requires seven contacts with a health facility by age five to complete the schedule. The commonly used first-year “fully immunized” definition includes one BCG dose, three OPV doses, three pentavalent doses, and one measles-rubella dose. But do not drop the phrase “fully immunized” into a report as if it explains itself.

Write the definition in full. Specify the cohort. Specify whether the estimate refers to vaccination by a milestone age or at interview. State whether birth-dose polio is included or excluded. This is not pedantry. It is the difference between an operational indicator and a headline.

A coverage estimate is only as honest as its denominator, dose definition, and evidence rule.

Build an indicator sheet that field teams can actually use

Before finalizing the protocol, produce a one-page indicator sheet for supervisors and enumerators. It should show:

If the primary goal is to examine dlhs child immunization indicators over time, preserve comparability deliberately. Do not assume that a new questionnaire, a new age cohort, or a new evidence rule can be laid beside older household-survey results without adjustment. India’s reported full-immunization estimates have moved from 53.5% in DLHS-3 (2007–08) to 62.0% in NFHS-4 (2015–16) and 76.4% in NFHS-5 (2019–2021). Those trend points are useful, but only when the underlying definitions and methods are read alongside the percentages.

2. Fix the sampling frame before the field team fixes it for you

Ad hoc household selection is the fastest route to a biased survey.

The pattern is familiar. A team reaches a selected village late. The map is incomplete. The local guide points them to houses near the road. A few locked homes are replaced with neighboring homes. The supervisor calls it practical. By the time the dataset reaches analysis, the probability sample has become a convenience sample with a formal-looking cluster ID.

Do not allow field pressure to redesign the survey.

A defensible immunization survey planning guide needs a probability-based selection process at every stage: selected areas, listed households, and eligible children. The household-selection method must be recorded because those records are needed later for survey weights. If the selection pathway cannot be reconstructed, neither can the estimate.

WHO survey guidance recommends that household line lists and maps for selected enumeration areas should ideally be no more than six months old. That is a practical threshold, not an Indian statutory rule. But the logic is hard to argue with. New settlements, seasonal migration, apartment construction, informal lanes, and displacement can make an old list structurally wrong.

The sampling and mapping control list

1. Define the sampling unit. Identify whether the first-stage unit is a village, census enumeration area, urban ward segment, or another operationally defensible unit. Do not mix units without documenting the design.

2. Confirm the frame’s date and coverage. Record when the map and household list were produced, who produced them, and which settlements may be missing. “Available list from local office” is not enough.

3. List households where the frame is stale or absent. If a recent household list or usable map does not exist, conduct a household-listing exercise in every selected area before selecting households. Listing is not glamorous. It is where representativeness is built.

4. Use simple or systematic household selection. Define the procedure in the field manual: starting point, interval where relevant, treatment of multi-household structures, and how teams identify eligible children.

5. Document every selection and replacement event. Better still, design the survey so replacement is not casually permitted. A locked home is not an invitation to interview the nearest available household.

6. Plan revisits. Hard-to-reach households are often absent during daytime visits because adults are working, migrating, or collecting water. Set a revisit window and a maximum number of attempts before fieldwork begins.

7. Retain the records for weighting. Selection probabilities, household counts, completed interviews, non-response, and protocol deviations must survive beyond the field period. The analysis team cannot repair missing selection records with statistical enthusiasm.

A typical cluster may take one to two days, depending on terrain, household availability, and travel conditions. Treat that as a planning estimate, not a rule. A dense urban settlement with locked rental units can consume more time than a compact rural village. A remote hamlet may have fewer households but lose half a day to transport. Build the route plan from physical reality, not from a spreadsheet that assumes every cluster is flat and accessible.

3. Obtain ethics clearance before schedules, hiring, and travel become sunk costs

Ethics review is often treated as a gate at the end of planning. That is backward. It should begin early because delays in approval can freeze hiring, translation, pilot work, and field deployment all at once.

A childhood vaccination survey data collection protocol should clearly state:

• how informed consent will be obtained;
• which caregivers may provide consent;
• what information will be collected from vaccination cards;
• whether names, phone numbers, addresses, or GPS coordinates are required;
• how personally identifying information will be separated from analytical data;
• who can access identifiable records;
• where devices and paper forms will be stored;
• how long identifiable data will be retained;
• what happens when an interviewer identifies a child with no documented doses, a severe access barrier, or an immediate health concern.

The last point needs discipline. A survey is not automatically a service-delivery campaign, but teams cannot pretend they did not see a child who may have missed care. Build a referral protocol with district health authorities before launch. Give enumerators a simple, accurate referral pathway: nearest session site, facility contact point, or local frontline worker channel. Do not ask them to make clinical promises they cannot fulfill.

Consent has to work in the local language, not only in the protocol language

Translation is not a formatting exercise. If the consent script says “confidentiality,” but the local-language explanation implies that data will be shared with village officials, the consent is defective in practice.

Training and tools should preferably operate in the local interviewing language. Where the training language differs, translate key concepts and back-translate them into the language of instruction. Test comprehension during role-play. If an interviewer cannot explain why a vaccination card is being photographed, copied, or transcribed—and what will happen to that information—the field procedure is not ready.

4. Design the tool around evidence, not around convenient answers

The vaccination card is the primary evidence source. The questionnaire must make that hierarchy visible.

Too many tools begin with caregiver recall because it is faster. That decision creates a predictable bias: a caregiver may remember that a child “received all injections” without recalling the antigen, dose number, timing, or whether the visit was recorded for another child. Recall has a role. It is not the same thing as documented vaccination history.

Field teams should actively encourage caregivers to locate a misplaced card. That means allowing time, asking whether the card is stored with other health records, and returning later if feasible. A rushed interviewer who records “card unavailable” after a 20-second question is manufacturing missing data.

When a card is available, dates must be transcribed exactly as written—even when they look implausible. Do not let enumerators “correct” a date in the household. A strange sequence or an impossible-looking date may indicate a recording error, but that is an analysis and validation problem. Preserve the source record first.

Fieldwork captures evidence. Analysis decides how to handle inconsistencies. Mixing those jobs corrupts both.

Put the evidence hierarchy into the questionnaire logic

The tool should have separate, visible fields for:

• card seen and transcribed;
• card reported but not seen;
• no card available after reasonable search;
• caregiver-reported dose history;
• date recorded on card;
• vaccine name or mark as it appears on the card;
• interviewer notes on illegibility, missing pages, or multiple cards;
• referral or follow-up action, if the protocol permits it.

Do not merge “yes, vaccinated” into one answer regardless of source. That destroys the ability to report card-only coverage versus card-plus-recall coverage later.

If the survey uses paper-assisted personal interviewing (PAPI), build legible transcription spaces and mandatory supervisor review. If it uses computer-assisted personal interviewing (CAPI), program range checks, skip logic, and daily synchronization procedures—but do not confuse an electronic form with quality assurance. A tablet can enforce a required field. It cannot tell whether the worker transcribed the right child’s card.

A proper pretest must stress the failure points:

• cards with handwritten dates and corrections;
• households with more than one eligible child;
• caregivers who need time to find records;
• absent caregivers;
• conflicting recall and card information;
• households at the edge of a cluster map;
• local terminology for vaccines and session sites.

The tool is ready when these cases produce consistent results across interviewers. Not when the form looks polished.

5. Train teams for the actual field environment

Five days is the minimum recommended duration for basic interviewer training, including at least one day of community-based practice. Again, that is methodological guidance, not a legal compliance number. But trying to compress consent, eligibility screening, vaccination schedules, card transcription, device use, household selection, and field safety into a two-day classroom sprint is how error gets industrialized.

A workable classroom group is generally 20 to 50 participants. Beyond that, role-play becomes theater: a few people perform, everyone else watches, and the first real interview becomes the first real test.

Training needs to be operational, not presentation-heavy. The field team should leave with muscle memory for the questions that create the most damage when handled casually.

Non-negotiable training modules

• Household and cluster identification: finding the correct selected area, using maps, identifying boundaries, and documenting deviations.
• Eligibility screening: applying dates of birth and household residence rules consistently.
• Consent and confidentiality: explaining participation, handling refusal, and protecting paper and digital records.
• Immunization schedule literacy: recognizing the doses relevant to the target cohort without inventing an interpretation from memory.
• Vaccination-card transcription: copying vaccine entries and dates exactly as recorded.
• Caregiver recall: asking neutral prompts without leading the respondent toward a socially desirable answer.
• Data-quality checks: reviewing incomplete forms, contradictory fields, duplicate child records, and missing evidence-source fields.
• Daily field routine: team check-in, device charging, secure storage, sync status, travel plan, and supervisor review.

Role-play must include more than a cooperative caregiver with a clean card. Run the difficult scenarios: no card, card in another household, a grandmother respondent, a parent who believes every injection is “polio,” a child with multiple health records, and an interviewer arriving at the wrong hamlet boundary.

Community-based practice is where the design meets infrastructure decay. Maps fail. Mobile networks disappear. Families are unavailable. Local names for settlements do not match administrative names. The pilot day should expose these problems while there is still time to repair the route plan and field manual.

6. Build quality control into the daily operating cycle

Endline cleaning cannot rescue fieldwork that was never controlled.

The supervisor needs a daily quality dashboard, whether the tool is paper or digital. It does not need to be elaborate. It needs to show the signals that identify a broken process while teams are still in the cluster.

Track, at minimum:

Supervisors should review forms every day, not at the end of the week. They should inspect a sample of vaccination-card transcriptions against the original card where consent procedures allow. They should conduct structured back-checks for household identity, child eligibility, card availability, and a limited number of core responses.

The point is not to punish enumerators for every discrepancy. The point is to identify whether the error is isolated, training-related, tool-related, or systemic. A single wrong date is a correction. Ten teams selecting the first houses on a lane is a sampling failure.

7. Do not confuse HMIS performance with household-survey coverage

India’s administrative and household-survey figures answer different questions. Treating them as interchangeable is a category error.

The Ministry of Health and Family Welfare has reported 94% full immunization in HMIS administrative reporting for financial year 2022–23. NFHS-5 reported 76.4% for 2019–2021 based on household-survey measurement. The gap cannot be read as a simple contradiction, nor should one figure be used to “validate” the other without a method review.

Administrative systems typically depend on service reports and target population denominators. Household surveys depend on sampled households, respondent availability, vaccination cards, recall, field procedures, and weighting. The reporting periods differ. The populations and definitions may differ. Denominator quality may differ. A district team needs both kinds of intelligence, but it must not blend them into one performance narrative.

Use the systems for what they can do:

• HMIS administrative data can flag reporting trends, service volumes, and possible disruptions quickly. It is useful for operational monitoring.
• Household surveys can measure documented and reported vaccination history in a sampled population, reveal equity gaps, and test whether routine reporting aligns with household-level evidence.
• Facility and cold-chain reviews explain the machinery behind both numbers: stock availability, session cancellation, outreach reach, staffing gaps, and last-mile transport.

A high district average can hide a failure corridor: riverine villages, migrant settlements, urban informal colonies, remote tribal hamlets, or communities where cards are routinely lost. Design the sample and strata to find those gaps if finding them is the point. A district-wide average alone rarely tells a program manager where the repair crew should go first.

The fieldwork gate: do not launch until the system is closed

Before the first household visit, the survey manager should be able to answer yes to every item below:

1. The target cohort, vaccine definition, reference period, and primary indicator are written and approved.

2. The sampling frame is current enough to defend, or household listing has been completed in selected areas.

3. Household selection is probability-based, documented, and protected from casual replacement.

4. Ethics approval, consent procedures, confidentiality controls, and referral pathways are active.

5. The questionnaire separates card evidence from caregiver recall and preserves dates exactly as recorded.

6. The tool has been piloted in local conditions, not merely reviewed in a meeting room.

7. Interviewers have completed structured training, role-play, and community practice.

8. Supervisors have a daily review process, back-check plan, and escalation route for protocol failures.

9. Transport, devices, charging, paper backups where needed, maps, contact lists, and cluster schedules are physically in place.

10. The reporting plan states clearly that household-survey estimates and HMIS administrative figures are different measurements, not rival slogans.

This is the scalable fix: not another dashboard, not a grand announcement, not an after-the-fact workshop on data quality. Build a survey operation where the sample is traceable, the evidence source is explicit, the staff know the protocol, and supervisors can stop a failure before it spreads across twenty clusters.

That is how childhood immunization data becomes useful enough to protect children—not just polished enough to publish.