THE GOLD STANDARD OF INTELLIGENCE · SPECIALIST STUDY
Respond to an RDS fieldwork deviation without confusing staff support, participant remedy, data validity and disciplinary judgement
A supervisor may see unusual duration, altered wording, consent uncertainty, incentive confusion, coupon handling error or unexplained degree values. The first task is not to decide guilt from a dashboard. It is to protect participants, preserve evidence, stop any unsafe recurrence and route the case by consequence and uncertainty. Retraining, blinded re-review, data quarantine and estimator sensitivity answer different questions and should remain visibly separate.
This guide belongs to the Intelligence subject root, the Gold Standard of Education, and the How X Works hub. It begins with a concrete decision, then tests whether the method still holds when the context changes.
Choose a reading route
Recognise and contain · chapters 1, 2, 3
Triage evidence fairly · chapters 4, 5
Repair field practice · chapters 6, 7, 8
Protect participants and estimates · chapters 9, 10, 11
Close, learn and transfer · chapters 12, 13
The whole problem at a glance
| Signal | Immediate question | Unsafe inference |
|---|---|---|
| Unusually short cases | Which modules and conditions changed? | Fabrication |
| Consent correction | Was understanding or voluntariness affected? | Clerical only |
| Coupon mismatch | Can recruitment lineage be restored? | Delete the participant |
| Degree heaping | Question, language or interviewer effect? | Edit values silently |
Separate participant, protocol and analytical consequences.
CHAPTER 1 OF 13 · Recognise and contain
1. Define a deviation without prejudging intent
A deviation is a departure from the approved protocol or authorised instruction. It can arise from ambiguity, workload, translation, device design, unsafe pressure or deliberate conduct; intent and consequence require separate evidence.
Evidence record: Version protocol, amendment, training material, case state, observed signal, reporter, time, language, device and the exact step expected. Preserve timestamps, versions, denominators, excluded cases, failures and rejected alternatives so another reviewer can reconstruct the decision rather than inherit a conclusion.
Decision rule: Classify the departure first; route personnel conclusions only through the authorised process with independent evidence. Name the authorised owner, trigger, stop condition, remedy and verification window. A control is incomplete until a changed-context case tests the mechanism it is meant to protect.
Evidence discipline: record the observation, version or time boundary that supports this chapter’s conclusion. Preserve the rejected alternative as well as the selected action. That makes a later review able to distinguish a changed world from an undocumented change of mind.
Changed-context transfer: A form update changes the expected step mid-week; separate old and new cases before judging compliance. The method transfers only if the decision question and measurement meaning survive the change. If either changes, rebuild the comparison rather than copying the earlier threshold or weight.
Chapter 1 check · what should survive an audit
- State the decision unit for “Define a deviation without prejudging intent”.
- Bind the claim to versions, denominators, missing cases and time boundaries.
- Connect evidence to an owner, bounded action and later verification.
CHAPTER 2 OF 13 · Recognise and contain
2. Contain immediate participant or data risk
Some signals require pausing an interview, incentive channel, coupon batch or interviewer assignment before full review. Containment must be narrow enough to avoid unnecessary harm or stigma.
Evidence record: Keep trigger, cases, participant contact status, coupons, payments, devices, access, data copies, temporary owner and review deadline. Preserve timestamps, versions, denominators, excluded cases, failures and rejected alternatives so another reviewer can reconstruct the decision rather than inherit a conclusion.
Decision rule: Use the least disruptive control that prevents plausible recurrence while preserving evidence and essential participant support. Name the authorised owner, trigger, stop condition, remedy and verification window. A control is incomplete until a changed-context case tests the mechanism it is meant to protect.
Evidence discipline: record the observation, version or time boundary that supports this chapter’s conclusion. Preserve the rejected alternative as well as the selected action. That makes a later review able to distinguish a changed world from an undocumented change of mind.
Changed-context transfer: A payment problem is isolated to one vendor; protect entitlement without suspending unrelated interviews. The method transfers only if the decision question and measurement meaning survive the change. If either changes, rebuild the comparison rather than copying the earlier threshold or weight.
Chapter 2 check · what should survive an audit
- State the decision unit for “Contain immediate participant or data risk”.
- Bind the claim to versions, denominators, missing cases and time boundaries.
- Connect evidence to an owner, bounded action and later verification.
CHAPTER 3 OF 13 · Recognise and contain
3. Preserve evidence and alternative explanations
Paradata, audio where authorised, notes, forms and device logs can support review, but each has missingness and privacy limits. Preserve originals and record transformations.
Evidence record: Retain source, consent authority, access, extraction, checksum, clock, missing intervals, annotations, translations and reviewer actions. Preserve timestamps, versions, denominators, excluded cases, failures and rejected alternatives so another reviewer can reconstruct the decision rather than inherit a conclusion.
Decision rule: Never reconstruct or edit source evidence in place; label inferred timelines and unresolved contradictions. Name the authorised owner, trigger, stop condition, remedy and verification window. A control is incomplete until a changed-context case tests the mechanism it is meant to protect.
Evidence discipline: record the observation, version or time boundary that supports this chapter’s conclusion. Preserve the rejected alternative as well as the selected action. That makes a later review able to distinguish a changed world from an undocumented change of mind.
Changed-context transfer: A tablet clock is wrong; rebuild timing from multiple sources and preserve uncertainty. The method transfers only if the decision question and measurement meaning survive the change. If either changes, rebuild the comparison rather than copying the earlier threshold or weight.
Chapter 3 check · what should survive an audit
- State the decision unit for “Preserve evidence and alternative explanations”.
- Bind the claim to versions, denominators, missing cases and time boundaries.
- Connect evidence to an owner, bounded action and later verification.
CHAPTER 4 OF 13 · Triage evidence fairly
4. Triage by consequence, recurrence and uncertainty
A harmless documentation slip differs from compromised consent, repeated leading prompts or recruitment-tree corruption. Severity should reflect affected rights, protocol purpose, scope, reversibility and analytical impact.
Evidence record: Record consequence domains, affected count, duration, recurrence, detectability, reversibility, participant vulnerability and uncertainty with rationale. Preserve timestamps, versions, denominators, excluded cases, failures and rejected alternatives so another reviewer can reconstruct the decision rather than inherit a conclusion.
Decision rule: Escalate mandatory rights or safety concerns independently of a composite score; do not average them away. Name the authorised owner, trigger, stop condition, remedy and verification window. A control is incomplete until a changed-context case tests the mechanism it is meant to protect.
Evidence discipline: record the observation, version or time boundary that supports this chapter’s conclusion. Preserve the rejected alternative as well as the selected action. That makes a later review able to distinguish a changed world from an undocumented change of mind.
Changed-context transfer: One low-frequency case affects a vulnerable participant; apply the rights floor despite small count. The method transfers only if the decision question and measurement meaning survive the change. If either changes, rebuild the comparison rather than copying the earlier threshold or weight.
Chapter 4 check · what should survive an audit
- State the decision unit for “Triage by consequence, recurrence and uncertainty”.
- Bind the claim to versions, denominators, missing cases and time boundaries.
- Connect evidence to an owner, bounded action and later verification.
CHAPTER 5 OF 13 · Triage evidence fairly
5. Use blinded re-review where it answers a real question
A reviewer can be blinded to interviewer identity, original disposition or outcome to reduce expectation effects, but full blinding may be impossible and can remove necessary context.
Evidence record: Specify material, redactions, blinding target, randomisation, reviewer qualification, rubric, disagreements, unblinding and adjudication. Preserve timestamps, versions, denominators, excluded cases, failures and rejected alternatives so another reviewer can reconstruct the decision rather than inherit a conclusion.
Decision rule: Use blinded review to estimate classification reliability or re-score evidence, not to conceal facts needed for participant protection. Name the authorised owner, trigger, stop condition, remedy and verification window. A control is incomplete until a changed-context case tests the mechanism it is meant to protect.
Evidence discipline: record the observation, version or time boundary that supports this chapter’s conclusion. Preserve the rejected alternative as well as the selected action. That makes a later review able to distinguish a changed world from an undocumented change of mind.
Changed-context transfer: A language nuance requires local context; unblind that field through a documented exception. The method transfers only if the decision question and measurement meaning survive the change. If either changes, rebuild the comparison rather than copying the earlier threshold or weight.
Chapter 5 check · what should survive an audit
- State the decision unit for “Use blinded re-review where it answers a real question”.
- Bind the claim to versions, denominators, missing cases and time boundaries.
- Connect evidence to an owner, bounded action and later verification.
Worked deviation triage
| Case | Potential consequence | First control | Analytical status |
|---|---|---|---|
| D17 | Missing consent confirmation | Pause and verify authority | Quarantine |
| D18 | Late incentive explanation | Participant contact review | Flagged |
| D19 | Coupon code transposed | Reconcile source records | Provisional |
| D20 | Long accessibility interview | None without other evidence | Include |
Actual authority and participant contact rules are study-specific.
CHAPTER 6 OF 13 · Repair field practice
6. Separate coaching, retraining and formal investigation
Coaching repairs a narrow performance gap, retraining rebuilds protocol competence and a formal investigation determines accountability under separate authority. Combining them can deter honest reporting.
Evidence record: Keep referral basis, learning objective, trainer, observed practice, assessment, support needs, personnel boundary and independent investigator where applicable. Preserve timestamps, versions, denominators, excluded cases, failures and rejected alternatives so another reviewer can reconstruct the decision rather than inherit a conclusion.
Decision rule: Choose the route by evidence and consequence; never use retraining as proof of fault or as a substitute for mandatory investigation. Name the authorised owner, trigger, stop condition, remedy and verification window. A control is incomplete until a changed-context case tests the mechanism it is meant to protect.
Evidence discipline: record the observation, version or time boundary that supports this chapter’s conclusion. Preserve the rejected alternative as well as the selected action. That makes a later review able to distinguish a changed world from an undocumented change of mind.
Changed-context transfer: Several staff misread one instruction; correct the system and train everyone without singling out one person. The method transfers only if the decision question and measurement meaning survive the change. If either changes, rebuild the comparison rather than copying the earlier threshold or weight.
Chapter 6 check · what should survive an audit
- State the decision unit for “Separate coaching, retraining and formal investigation”.
- Bind the claim to versions, denominators, missing cases and time boundaries.
- Connect evidence to an owner, bounded action and later verification.
CHAPTER 7 OF 13 · Repair field practice
7. Rehearse the protocol through teach-back and observation
Completion certificates do not prove usable competence. The interviewer should explain the purpose, demonstrate the step, handle a changed-context case and receive bounded supervised practice.
Evidence record: Record scenario, rubric, errors, prompts, accessibility and language conditions, assessor, retry, live observation and release decision. Preserve timestamps, versions, denominators, excluded cases, failures and rejected alternatives so another reviewer can reconstruct the decision rather than inherit a conclusion.
Decision rule: Return duties only after the relevant skill passes in realistic conditions and monitoring is proportionate. Name the authorised owner, trigger, stop condition, remedy and verification window. A control is incomplete until a changed-context case tests the mechanism it is meant to protect.
Evidence discipline: record the observation, version or time boundary that supports this chapter’s conclusion. Preserve the rejected alternative as well as the selected action. That makes a later review able to distinguish a changed world from an undocumented change of mind.
Changed-context transfer: A scripted case passes but a refusal scenario fails; keep refusal handling supervised. The method transfers only if the decision question and measurement meaning survive the change. If either changes, rebuild the comparison rather than copying the earlier threshold or weight.
Chapter 7 check · what should survive an audit
- State the decision unit for “Rehearse the protocol through teach-back and observation”.
- Bind the claim to versions, denominators, missing cases and time boundaries.
- Connect evidence to an owner, bounded action and later verification.
CHAPTER 8 OF 13 · Repair field practice
8. Quarantine data without erasing provenance
Quarantine means keeping affected records separate from routine analysis while review continues. Deleting or silently correcting them destroys evidence and can distort recruitment relationships.
Evidence record: Store case IDs, fields, reason, source version, related coupons, recruiter links, payments, access, permitted uses, review state and release or exclusion receipt. Preserve timestamps, versions, denominators, excluded cases, failures and rejected alternatives so another reviewer can reconstruct the decision rather than inherit a conclusion.
Decision rule: Allow only declared review uses until validity and participant obligations are resolved; preserve original and corrected representations. Name the authorised owner, trigger, stop condition, remedy and verification window. A control is incomplete until a changed-context case tests the mechanism it is meant to protect.
Evidence discipline: record the observation, version or time boundary that supports this chapter’s conclusion. Preserve the rejected alternative as well as the selected action. That makes a later review able to distinguish a changed world from an undocumented change of mind.
Changed-context transfer: A coupon code is repairable but degree data is not; release fields separately with lineage. The method transfers only if the decision question and measurement meaning survive the change. If either changes, rebuild the comparison rather than copying the earlier threshold or weight.
Chapter 8 check · what should survive an audit
- State the decision unit for “Quarantine data without erasing provenance”.
- Bind the claim to versions, denominators, missing cases and time boundaries.
- Connect evidence to an owner, bounded action and later verification.
CHAPTER 9 OF 13 · Protect participants and estimates
9. Provide participant remedy through authorised routes
A protocol deviation can affect consent, privacy, payment, burden or access. Remedy may include explanation, payment correction, withdrawal support, reconsent or incident response, depending on ethics and study authority.
Evidence record: Keep impact, contact permission, language, urgency, remedy options, chosen action, completion, refusal, escalation and no-retaliation safeguards. Preserve timestamps, versions, denominators, excluded cases, failures and rejected alternatives so another reviewer can reconstruct the decision rather than inherit a conclusion.
Decision rule: Do not recontact when unsafe or unauthorised; route uncertainty to the ethics, privacy or study authority rather than improvising. Name the authorised owner, trigger, stop condition, remedy and verification window. A control is incomplete until a changed-context case tests the mechanism it is meant to protect.
Evidence discipline: record the observation, version or time boundary that supports this chapter’s conclusion. Preserve the rejected alternative as well as the selected action. That makes a later review able to distinguish a changed world from an undocumented change of mind.
Changed-context transfer: A participant cannot be reached safely; document the protected alternative and unresolved obligation. The method transfers only if the decision question and measurement meaning survive the change. If either changes, rebuild the comparison rather than copying the earlier threshold or weight.
Chapter 9 check · what should survive an audit
- State the decision unit for “Provide participant remedy through authorised routes”.
- Bind the claim to versions, denominators, missing cases and time boundaries.
- Connect evidence to an owner, bounded action and later verification.
Re-entry gate after retraining
| Gate | Evidence | If absent |
|---|---|---|
| Knowledge | Observed protocol teach-back | Continue coaching |
| Practice | Supervised mock and live check | Limit duties |
| Participants | Remedies completed or routed | Do not close |
| Data | Quarantine and sensitivity resolved | Keep restricted |
Each consequence needs its own evidence.
CHAPTER 10 OF 13 · Protect participants and estimates
10. Repair recruitment and payment lineages carefully
RDS cases connect participants, coupons, recruiters and secondary incentives. Correcting one identifier can alter trees and weights; privacy-preserving linkage still needs an audit trail.
Evidence record: Record original and proposed links, evidence, confidence, affected payments, duplicate risk, privacy transformation, approver and downstream recomputation. Preserve timestamps, versions, denominators, excluded cases, failures and rejected alternatives so another reviewer can reconstruct the decision rather than inherit a conclusion.
Decision rule: Do not force a unique recruiter when evidence supports several possibilities; carry uncertainty into sensitivity analysis. Name the authorised owner, trigger, stop condition, remedy and verification window. A control is incomplete until a changed-context case tests the mechanism it is meant to protect.
Evidence discipline: record the observation, version or time boundary that supports this chapter’s conclusion. Preserve the rejected alternative as well as the selected action. That makes a later review able to distinguish a changed world from an undocumented change of mind.
Changed-context transfer: Two coupons share a transposed digit; compare timing and issue logs without exposing identities broadly. The method transfers only if the decision question and measurement meaning survive the change. If either changes, rebuild the comparison rather than copying the earlier threshold or weight.
Chapter 10 check · what should survive an audit
- State the decision unit for “Repair recruitment and payment lineages carefully”.
- Bind the claim to versions, denominators, missing cases and time boundaries.
- Connect evidence to an owner, bounded action and later verification.
CHAPTER 11 OF 13 · Protect participants and estimates
11. Measure estimator sensitivity to plausible resolutions
Affected interviews can be included, excluded, reclassified or have uncertain recruiter or degree information. One preferred clean-up hides the range of defensible estimates.
Evidence record: Define estimand, estimator, seeds, network size, exclusions, repaired links, imputation, uncertainty sets and outcome differences across scenarios. Preserve timestamps, versions, denominators, excluded cases, failures and rejected alternatives so another reviewer can reconstruct the decision rather than inherit a conclusion.
Decision rule: Report whether substantive conclusions change and which unresolved field drives them; do not treat stability as proof that participant harm was harmless. Name the authorised owner, trigger, stop condition, remedy and verification window. A control is incomplete until a changed-context case tests the mechanism it is meant to protect.
Evidence discipline: record the observation, version or time boundary that supports this chapter’s conclusion. Preserve the rejected alternative as well as the selected action. That makes a later review able to distinguish a changed world from an undocumented change of mind.
Changed-context transfer: A prevalence estimate is stable but a subgroup estimate moves; preserve both conclusions. The method transfers only if the decision question and measurement meaning survive the change. If either changes, rebuild the comparison rather than copying the earlier threshold or weight.
Chapter 11 check · what should survive an audit
- State the decision unit for “Measure estimator sensitivity to plausible resolutions”.
- Bind the claim to versions, denominators, missing cases and time boundaries.
- Connect evidence to an owner, bounded action and later verification.
CHAPTER 12 OF 13 · Close, learn and transfer
12. Monitor re-entry without creating punitive surveillance
After intervention, focused checks can verify repair. Excessive hidden monitoring can change behaviour, undermine trust and burden participants.
Evidence record: Declare duration, sampling, authorised reviewers, privacy, feedback, success, recurrence, escalation, sunset and deletion of extra monitoring data. Preserve timestamps, versions, denominators, excluded cases, failures and rejected alternatives so another reviewer can reconstruct the decision rather than inherit a conclusion.
Decision rule: End enhanced monitoring when predeclared evidence supports it; reopen only on a new justified trigger. Name the authorised owner, trigger, stop condition, remedy and verification window. A control is incomplete until a changed-context case tests the mechanism it is meant to protect.
Evidence discipline: record the observation, version or time boundary that supports this chapter’s conclusion. Preserve the rejected alternative as well as the selected action. That makes a later review able to distinguish a changed world from an undocumented change of mind.
Changed-context transfer: A supervisor changes; transfer the monitoring rationale and sunset rather than restarting indefinitely. The method transfers only if the decision question and measurement meaning survive the change. If either changes, rebuild the comparison rather than copying the earlier threshold or weight.
Chapter 12 check · what should survive an audit
- State the decision unit for “Monitor re-entry without creating punitive surveillance”.
- Bind the claim to versions, denominators, missing cases and time boundaries.
- Connect evidence to an owner, bounded action and later verification.
CHAPTER 13 OF 13 · Close, learn and transfer
13. Close each consequence and learn at system level
Case closure requires separate evidence for participant remedy, data disposition, staff readiness, recruitment lineage, payments and required reporting. Repeated deviations may reveal workload or design problems.
Evidence record: Publish a restricted closure map with open items, owners, decisions, appeals, estimator impact, systemic causes, control changes and recurrence review. Preserve timestamps, versions, denominators, excluded cases, failures and rejected alternatives so another reviewer can reconstruct the decision rather than inherit a conclusion.
Decision rule: Close only when mandatory obligations are resolved or explicitly transferred to an accountable owner. Name the authorised owner, trigger, stop condition, remedy and verification window. A control is incomplete until a changed-context case tests the mechanism it is meant to protect.
Evidence discipline: record the observation, version or time boundary that supports this chapter’s conclusion. Preserve the rejected alternative as well as the selected action. That makes a later review able to distinguish a changed world from an undocumented change of mind.
Changed-context transfer: Transfer the method to a school research survey while replacing RDS coupons and estimator fields. The method transfers only if the decision question and measurement meaning survive the change. If either changes, rebuild the comparison rather than copying the earlier threshold or weight.
Chapter 13 check · what should survive an audit
- State the decision unit for “Close each consequence and learn at system level”.
- Bind the claim to versions, denominators, missing cases and time boundaries.
- Connect evidence to an owner, bounded action and later verification.
Contents · Previous chapter · Return to the Intelligence guide
Verification lab: work the decision twice
Triage six fictional deviations across consent, wording, coupon, payment, timing and accessibility. Separate immediate containment from later judgement.
Run a blinded re-review with a written rubric, one justified unblinding and disagreement adjudication.
Estimate an outcome under include, exclude, repaired-link and uncertain-degree scenarios, then design participant and staff re-entry actions.
Limits and misconceptions
Paradata is not proof of misconduct
It is incomplete evidence that must be read with protocol and context.
Retraining is not a universal remedy
Rights, data validity, payments and reporting may need separate action.
Quarantine is not deletion
Preserve provenance and permit only declared review uses.
Stable estimates do not erase participant harm
Analytical and ethical consequences remain separate.
Questions readers ask
What is a protocol deviation?
- A departure from the approved protocol or authorised instruction, without assuming intent.
When should a supervisor pause fieldwork?
- When continued activity could plausibly repeat participant, privacy, payment or data harm.
Why use blinded re-review?
- To reduce some expectation effects when classifying evidence, while documenting any necessary unblinding.
What does data quarantine mean?
- Separating affected records from routine analysis while preserving originals, lineage and review access.
Can retraining close the case?
- Only the competence part; participant, analytical, payment and reporting consequences need their own closure.
Why run estimator sensitivity?
- To show whether defensible resolutions of uncertain cases change the study conclusion.
Practise, verify and continue
Design a supervisor response to short interviews, consent uncertainty and coupon mismatch. Include containment, evidence, blinded review, retraining, quarantine, participant remedy and estimator sensitivity.
Move the method to a longitudinal school survey. Preserve rights floors, evidence integrity, bounded review, data quarantine and sensitivity analysis while replacing RDS recruitment structures.
Practical evidence check
Before implementation, write a one-page evidence map with the decision, unit of analysis, version or time boundary, inputs, outcomes, uncertainty, authority and stop rule. Attach one worked case that passes and one that fails. A reviewer should be able to reproduce the reasoning without relying on the memory of the person who built the system.
Run one changed-context rehearsal before approval. Alter a mechanism rather than a cosmetic detail, then identify which calculation, safeguard or governance gate must change. Record any conclusion that does not transfer. This short rehearsal is often the fastest way to expose a hidden assumption before it becomes an operational dependency.
Related eduKate reading
RDS interviewer workload and quality thresholds
RDS interviewer changeover and coaching drift
RDS adaptive fieldwork governance
RDS missing recruitment links and tree repair
Authoritative sources and scope
WHO World Health Survey Plus interviewer manual
CDC NHANES 2025–2027 procedure manuals
CDC NHANES interviewer procedures manual
CDC NSFG paradata user guide
WHO and CDC sources support interviewer training, supervision, protocol adherence, quality review and cautious use of field evidence. They do not prescribe this RDS triage, quarantine or estimator-sensitivity framework. Ethics approval, participant safety, employment process, privacy law and study-specific authority govern real cases.