What Happens When ABA Data Are Missing?

Classify missing ABA data before interpreting it

A blank can mean no session occurred, the target had no eligible opportunity, the client declined, the device failed, the observer missed the event, a form was lost, the entry was delayed, or the definition was unclear. Those states are not equivalent.

Use a small set of reason codes plus a note when needed. Preserve planned sessions, delivered sessions, valid opportunities, completed observations, and missing entries as separate counts.

Keep the denominator honest

If 12 observations were due and data exist for 9, completeness is 9 of 12, not 100% of available records. Excluding every missing or inconvenient entry can make performance look stronger and hide a system problem. When an opportunity truly did not occur, explain the eligibility rule and report that state separately.

The BCBA Test Content Outline covers measurement, data display and interpretation, treatment integrity, and assessment concepts. It is examination content, not a universal missing-data protocol.

Reconstruction has strict limits

A schedule, note, video, device log, or caregiver report may help establish that a session or event occurred. It may not support the exact response count, prompt level, latency, or duration. Mark reconstructed information by source and confidence. Do not copy a neighboring value or treat memory as direct observation.

The RBT Ethics Code requires accurate documentation. The BACB Ethics Code addresses accuracy, documentation, data evaluation, and correction for covered behavior analysts.

Separate clinical data from billing evidence

A missing clinical measure, missing session note, missing time record, and missing claim artifact create different problems. One source should not be invented to repair another. Ask which service occurred, which documentation is required, whether a claim was submitted, and what each responsible role is correcting.

If missingness affects safety, health, consent, or a restrictive event, escalate promptly.

A practical example

A goal has 15 planned observation periods. Eleven contain valid data, two sessions were canceled before service, one device file is corrupt, and one delivered session lacks the goal measure. Report 11 of 13 delivered observation periods complete and preserve both cancellations outside that delivered denominator. The two missing delivered records remain open with reason and owner.

Build a missing-data map

Start with the complete cohort that should be accounted for. For a weekly goal review, that might be every scheduled visit, every delivered session, or every eligible opportunity during a defined period. Then assign each item one status.

A practical map can include complete, canceled before service, no opportunity, declined, interrupted, observer unavailable, device or file failure, entry overdue, invalid measurement, and unknown. Keep the status date, source, owner, and next action. A blank field should be treated as unresolved rather than assumed to be one of these reasons.

Decide which conclusions are exposed

Missing data do not affect every decision equally. Ask which graph, baseline, mastery criterion, safety review, progress report, authorization request, supervision check, or discharge discussion used the affected period. A missing value far from a decision threshold may have limited impact. A concentrated gap in the only community observations may undermine a generalization conclusion.

The qualified clinician can perform a sensitivity check. If five of 20 opportunities are missing and 12 observed opportunities contain the response, the possible overall result ranges from 12 of 20 to 17 of 20. This range does not replace clinical judgment. It shows whether plausible missing values could change the numerical conclusion.

Recover only what the evidence supports

Create a hierarchy of contemporaneous sources. A signed paper form may support an exact count. A video may support rescoring if lawful, available, and appropriate. A session note might establish that an activity occurred while lacking enough detail for the target measure. A schedule only shows that service was planned.

Record the source and who made the reconstruction. If the exact value cannot be recovered, mark it unavailable and collect new evidence when needed. Avoid using an average of nearby sessions, the last recorded value, or a staff estimate as though it were observed fact.

Fix the workflow that produced the gap

Missingness can come from an inaccessible form, unstable device, unclear ownership, insufficient documentation time, offline sync failure, staffing transition, or a definition that observers cannot use. The corrective action should match the cause.

A prevention plan might add required status codes, an offline backup, an overdue queue, supervisor review, device testing, or protected completion time. Track future completeness across all due records. A reminder alone is weak when the form or workload makes accurate entry impractical.

Communicate uncertainty plainly

A family update can state: “Fourteen observations were due. Eleven are complete, one session was canceled, one device file is unrecoverable, and one entry remains under review. The clinician is holding the mastery decision until a new community probe is completed.”

This explanation identifies the cohort, states, impact, and next step. It is more useful than a graph with gaps and no reason. Families can ask to be notified if the missingness changes a report or decision already shared.

A second example with concentrated missingness

Maya has 30 school and home opportunities. Data are present for 24. All six missing opportunities occurred at home because the mobile form failed offline. The observed total may look adequate, but the gap is concentrated in one setting. The team should avoid claiming performance across both settings from the available data.

It can repair the app, validate the fix, obtain new home observations, and keep the school evidence separate. The missingness pattern matters more than the overall 80% completeness rate.

Close the review with accountable states

For each missing item, record recovered, replaced by new observation, confirmed no opportunity, formally invalid, or still unresolved. Recalculate affected summaries and ask the clinician whether conclusions change. A missing-data review is complete when every expected record has a disposition, not when the visible blanks disappear.

Keep a copy of the final reconciliation with the affected graph or report. It should show the original cohort, disposition totals, any recalculated result, the clinician's impact decision, and the prevention control. At the next review, verify that new due records are complete and that the same failure has not returned.

Preserve any remaining uncertainty explicitly.

Questions families can use

Ask what was expected, what occurred, which values are missing, why, whether the reason was known before review, which denominator is used, whether reconstruction is possible, who interprets the gap, which decision is held, and what control will prevent recurrence.