To handle partially observed sessions in IOA, identify the window both observers validly shared, calculate agreement only for matched units inside it, and report coverage separately. Retain minutes, trials, intervals, and events observed by only one person. A missing score, lost view, or invalid interval needs its own state because treating it as agreement inflates confidence.

Separate exposure states

Use planned, A-only, B-only, shared-valid, shared-invalid, and missing-record states.

Apply unit-level rules

For interval or trial data, mark each unit. For count or duration data, preserve start and stop boundaries.

Report coverage beside agreement

Show observed and eligible numerators and denominators. Strong agreement across limited coverage carries limited scope.

Resample important gaps

Schedule another observation when the missing portion contains a risk, setting, staff member, support, or response form central to the decision.

Build an exposure-state timeline

Create mutually exclusive states for planned, shared-valid, A-only, B-only, shared-invalid, and missing-record exposure. Record the start and stop boundary, condition, reason, and source for every state. For interval data, classify each interval. For events or duration, retain the actual observer windows so events near a boundary can be reviewed consistently.

Elena's timeline reconciles exactly: 31 shared-valid + 8 before Observer B joined + 6 with one observer's view lost = 45 planned minutes. Matched coverage is 31 of 45, or 68.9%. The other 14 of 45 minutes, or 31.1%, remain outside the matched calculation. Keep the 8-minute and 6-minute reasons separate because they may call for different repairs.

Check whether missing exposure is selective

Compare activities, staff, supports, response opportunities, and health or access conditions in the matched and unmatched portions. If lost visibility occurs mainly during transitions, community activity, high-rate responding, or AAC use, the shared window may represent an easier slice of the session. Report that limitation and obtain another sample that covers the missing condition.

Agreement within the 31 shared minutes describes only that window. It does not recover the 14 unpaired minutes or establish that those minutes would have produced the same result.

Apply the method inside the shared window

Align interval boundaries and trial IDs before calculating interval or trial agreement. For event data, use a prospective event-matching rule within the shared window. For count or duration data, preserve exact start and stop times and avoid comparing totals drawn from different exposure lengths. Invalid or missing units remain separate from disagreement.

Choose the next action from the decision. A low-stakes description may tolerate the limitation with a clear caveat. A treatment change, safety decision, or staff-performance conclusion may require a new observation with stronger coverage. The qualified clinician owns that judgment.

Apply the field control to one defined question

The partial-exposure reconciliation for Elena begins with the exact decision, the responsible qualified role, and the evidence needed. Preserve original observer records, matched exposure, the selected calculation, raw numerator and denominator, unrounded result, and every excluded or unresolved unit.

Work the example for Elena

Elena's planned session lasts 45 minutes. Both observers share valid exposure for 31 minutes; 8 minutes occur before Observer B joins and 6 minutes lose one observer's view. Coverage is 31 of 45, or 68.9%. The other 14 minutes remain visible outside the IOA calculation.

Audit the denominator for Elena

The denominator for Elena is the eligible cohort defined before review. Report planned, due, observed, matched, invalid, missing, and open units separately. A percentage gains meaning from the unit, coverage, dates, conditions, and disagreement pattern.

Protect access and safety during the sample for Elena

The observation for Elena keeps AAC, interpreters, mobility, food, water, bathroom use, prescribed care, health supports, rest, relationships, and emergency help available. Observer procedure never delays immediate safety or mandated action. Record changes in access, health, assent, distress, visibility, staff behavior, or routine.

Use current sources within their scope for Elena

For Elena, the BACB ethics hub points to current ethics materials. The CASP public summary supplies high-level autism-treatment scope. The BACB Ethics Code addresses competence, client involvement, consent and assent when applicable, documentation, supervision, risk, and evaluation for covered behavior analysts. The BCBA Test Content Outline includes measurement, reliability, sampling, and data-based decisions as examination content.

For Elena and the partial IOA sample handling question, Vollmer, Sloman, and St. Peter Pipkin discuss practical implications of data reliability and treatment-integrity monitoring. Essig, Rotta, and Poling review IOA and fidelity reporting in an identified research corpus. Reed and Azulay describe agreement calculations and a calculation tool. ASHA says AAC users should always have access to their tools or devices. The responsible clinician still selects a case-specific measurement and review plan.

Close the IOA review for Elena

Review the partial-exposure reconciliation with Elena through accessible communication and with the responsible clinician. Record the question, sample, method, arithmetic, disagreement types, limits, repair, owner, next observation, and review date. Reopen the review when the definition, observer, setting, system, exposure, access, prevalence, procedure, or decision changes.

Before release, confirm that exposure states sum to the complete planned sample, the agreement denominator contains only shared-valid units, and unmatched units remain visible by reason. Record whether those gaps could bias the decision and whether another observation is due.

Related resources

Sources