ABA Missing Data Review Worksheet for Clinicians

An ABA missing data review worksheet should keep a true observed zero separate from no opportunity, not observed, missing, and invalid. It should preserve the planned denominator, original record, reason for each state, display decision, and follow-up owner. That distinction prevents a blank cell from being graphed as an event that was observed not to occur.

Clinicians & ABA Professionals / Data, Outcomes and Clinical Decision-Making.

The worksheet does not choose a clinical measure, set a universal completeness threshold, or authorize imputation. A qualified clinician remains responsible for the measurement plan and interpretation, with methods, documentation, privacy, payer, and legal review when applicable.

Define every state before reviewing the week

The current BCBA Test Content Outline, Sixth Edition includes operational definitions, measurement procedures, validity, reliability, representative measurement, data display, and interpretation. The BACB's test-content outline page identifies the current outline. The outline establishes neither a missing-data codebook nor a minimum acceptable coverage percentage.

Anchor the review to the measurement contract then in force. Identify the observation unit, eligible opportunity, recording method, observer, window, and rules for every state. If the rules were not defined prospectively, document that limitation rather than inventing a clean retrospective distinction.

Complete the measurement and review contract:

Client or authorized case code.
Goal, response, or process measured.
Operational definition and version.
Measurement unit and observation window.
Planned windows or opportunities and eligibility rule.
Expected observer and recording route.
Rules for true zero and no opportunity.
Rules for not observed, missing, and invalid.
Display and exclusion rule.
Review owner and escalation route.
Privacy, access, and retention controls.

For a detailed field-level distinction, see Distinguish Zero, No Opportunity, Missing, Invalid, and Not Observed in ABA Records. The present page turns those distinctions into a weekly or review-period working artifact.

A state dictionary prevents blank-cell assumptions

StateMinimum meaningIncluded in the measured value?What remains visibleValid observed valueObservation occurred under the stated contract and produced a usable valueYesValue, unit, observer, time, and definition versionTrue observed zeroEligible observation occurred and the measured count was zeroYesExplicit zero and evidence that the observation occurredNo opportunityObservation occurred or context was known, but no eligible opportunity aroseUsually no; follow the defined measureOpportunity rule and contextNot observedA relevant period existed, but the designated observation did not occurNoWhy it was not observed and who knewMissingA record expected under the plan is absent or its disposition is unknownNoExpected item, discovery, search, and unresolved statusInvalidA record exists but fails a predefined usability ruleNo unless a qualified plan says otherwiseOriginal value, failure reason, and preservation route

Use different symbols for zero and missing. Preserve an invalid value even when it is excluded. The record, exclusion reason, and reviewer should remain traceable.

The BACB's Ethics Codes page and Ethics Code for Behavior Analysts address accurate work, documentation, confidentiality, and data-based services within their scope. They do not make this dictionary a universal clinical or legal rule.

Log each planned window before summarizing

Use one row for every planned window, even when no usable value results. This keeps the denominator auditable. If the plan changes, preserve the earlier plan and open a dated version.

WindowPlanned context and durationEligible opportunity known?Observation completed?Raw valueState codeReason or evidenceDisplay dispositionFollow-up owner1Yes / no / unknownYes / noPlot / gap / annotate / exclude2Yes / no / unknownYes / noPlot / gap / annotate / exclude3Yes / no / unknownYes / noPlot / gap / annotate / exclude4Yes / no / unknownYes / noPlot / gap / annotate / exclude5Yes / no / unknownYes / noPlot / gap / annotate / exclude6Yes / no / unknownYes / noPlot / gap / annotate / exclude

CASP's ABA Practice Guidelines Version 3 access page supplies the public version, audience, and licensing context. This worksheet is an original artifact; it neither reproduces licensed guideline text nor claims CASP endorsement.

Reconcile counts before calculating coverage

Use the ABA missing data review worksheet to summarize mutually exclusive states. The state count should equal the planned denominator, or the reviewer should explain the difference.

Coverage summaryCountCalculation or notePlanned windowsPrespecified denominatorValid observed records, including true zerosNo-opportunity windowsInvalid recordsNot-observed windowsMissing or unresolved recordsState-count reconciliationSum of mutually exclusive states / planned windowsValid measurement coverageValid observed / planned × 100Resolved-state coverageValid + no opportunity + invalid + other resolved states / planned × 100

“Resolved” means the state is known, not that the row is valid measurement. A device-failure row can be resolved and still excluded. Keep that distinction visible.

A fictional 12-window review

The following example is synthetic. It is not a client record, a recommended schedule, a completeness threshold, or a basis for a treatment decision.

A fictional plan calls for 12 equal-length windows. Seven contain valid values, including three true zeros. Two have no opportunity. One is invalid because a device clock failed the timing rule. Two remain missing or not observed.

WindowRaw valueStateReview note10Valid observed zeroObservation completed; eligible context present21Valid observed valueObservation completed32Valid observed valueObservation completed4No valueNo opportunityContext documented; no eligible opportunity arose50Valid observed zeroObservation completed; eligible context present6No valueInvalidDevice clock failure prevents use under the predefined rule71Valid observed valueObservation completed8No valueMissingPlanned record absent; disposition unresolved90Valid observed zeroObservation completed; eligible context present10No valueNo opportunityContext documented; no eligible opportunity arose113Valid observed valueObservation completed12No valueNot observedPlanned window occurred without the designated observation

The state counts reconcile: 7 valid + 2 no opportunity + 1 invalid + 1 missing + 1 not observed = 12 planned windows.

Valid measurement coverage: 7 valid / 12 planned × 100 = 58.3%.

Resolved-state coverage: (7 valid + 2 no opportunity + 1 invalid) / 12 planned × 100 = 10 / 12 × 100 = 83.3%.

The seven valid windows contain 0, 1, 2, 0, 1, 0, and 3 occurrences. Their sum is 7, and their arithmetic mean is 7 / 7 = 1.0 occurrence per valid window. That descriptive value is not a complete weekly estimate. No-opportunity, invalid, missing, and not-observed rows are not zeros and do not enter the numerator.

Fictional calculation checkValuePlanned windows12Valid values7No opportunity2Invalid1Missing1Not observed1Reconciliation7 + 2 + 1 + 1 + 1 = 12Valid measurement coverage7 / 12 = 58.3%Resolved-state coverage10 / 12 = 83.3%Sum among valid values0 + 1 + 2 + 0 + 1 + 0 + 3 = 7Mean among valid equal-length windows7 / 7 = 1.0

Neither percentage is labeled acceptable or unacceptable. Pattern, reason, representativeness, decision stakes, and the measurement plan govern interpretation.

Ask whether missingness is patterned

The count can hide a pattern. Missing records may cluster by time, setting, staff, response intensity, access need, device, or schedule. Clustering can change which part of the client's experience is visible.

For each lens, note the evidence, whether a pattern appears, and any interpretation limit or action:

Time, day, setting, or activity.
Observer, team, or workflow.
Client access, assent, health, or communication context.
Response level or clinical condition.
Device, network, or integration.
Definition or measurement-version change.
Authorization, schedule, or service interruption.

A review of visual-analysis methods in applied behavior analysis notes that incomplete data can limit interpretation in the literature it examined. A discussion of visual analysis with sufficient data illustrates why amount and arrangement matter. Neither article supplies a universal threshold.

Keep display and decision dispositions explicit

For graph mechanics, use How to Display Missing and Invalid Data on ABA Graphs. Keep unknown spans and display limitations visible. Plot no opportunity at zero only when zero is genuinely the observed value under the defined measure.

Output or decisionData usedMissingness visible?Current dispositionOwner and review dateRaw-data displayYes / noUse / annotate / hold / reviseClinical graphYes / noUse / annotate / hold / reviseProgress summaryYes / noUse / qualify / hold / reviseTreatment or supervision reviewYes / noProceed / seek more data / otherCaregiver or client communicationYes / noAuthorization, payer, or quality reportYes / no

When coverage cannot support the question, collect more representative observations, narrow the claim, examine the workflow, or hold the decision. Record the reasoning; a completeness percentage cannot substitute for clinical judgment.

Treat imputation as a separate analysis decision

The research article indexed as PubMed 38374608 discusses missing data and imputation in a single-case research context. PubMed 30805863 examines an expectation-maximization approach in another research setting. Neither source authorizes silent clinical-record imputation or makes one method appropriate for every design.

If a qualified plan uses imputation, keep the observed record unchanged and label derived values as estimated. Record the method, assumptions, implementation, analyst, version, sensitivity analysis, and affected outputs. Never backfill the raw record.

Derived-value review, if applicableEntryClinical or analytic questionObserved-data set and frozen versionMissingness description and assumptionsQualified method reviewerMethod and implementation versionSensitivity or alternative analysisVisual label for estimated valuesDecisions that may use the estimateStop condition

Leaving this table blank is appropriate when no imputation is planned. A spreadsheet should never invent values simply to remove gaps from a graph.

Protect client rights and record security

The BACB Ethics Code resources support professional review within their scope. A coverage problem may also signal access barriers, inequitable scheduling, unclear roles, or technology failures. Include client and caregiver perspectives without pressuring anyone to accept more intrusive recording.

HHS Security Rule risk-analysis guidance provides context for regulated entities. The guidance does not approve this worksheet or any storage channel. Use authorized systems, minimum necessary information, role-based access, and approved retention. Recording, consent, assent, site rules, and state law require separate review.

Turn the review into proportionate follow-up

Follow-up itemOwnerDue dateEvidence of completionProspective effective dateClarify definition or state codeRepair device or integration issueRestore accessible collection routeCollect additional representative dataAnnotate graph or summaryReview affected decision or communicationEscalate privacy, safety, payer, legal, or incident concernSchedule a later drift or coverage check

Prospective changes should not retroactively reclassify old rows collected under a different contract. If an earlier row is actually wrong, route it through the separate correction process and preserve the original audit trail.

Closure review

Do the mutually exclusive state counts reconcile to the planned denominator?
Are true observed zeros distinguishable from every unknown or ineligible state?
Are invalid records preserved with reasons rather than deleted?
Are missingness patterns and access barriers visible?
Are graphs and summaries explicit about gaps, exclusions, and versions?
Are any derived or imputed values separately labeled and reviewed?
Are decision limits recorded in ordinary language?
Does each follow-up have an owner without assigning blame from incomplete evidence?
Are privacy, recording, payer, billing, licensure, and legal questions routed appropriately?

A completed worksheet documents what is known, what remains unknown, and the next action. Completion does not certify measurement validity, treatment success, medical necessity, or compliance.

ABA Missing, Invalid, and No-Opportunity Data Review Worksheet