To calculate duration per occurrence IOA, match the same events across observers, divide the smaller duration by the larger duration for each matched event, multiply by 100, then average those event percentages. Preserve start and stop times. Unmatched, partially observed, or differently segmented events require investigation rather than a forced duration pair.

Match events first

Use event IDs, timestamps, context, and sequence. Similar durations do not prove the observers measured the same event.

Calculate each event

Keep each smaller-to-larger ratio unrounded until the final mean.

Retain unmatched events

Report events seen by only one observer, events outside shared exposure, and segmentation disagreements separately.

Inspect start and stop differences

A similar duration can result from equally shifted boundaries. Review timestamps when timing matters clinically.

Define the event-matching rule prospectively

Specify how close start times, context, sequence, and response form must be for two records to represent the same occurrence. A matching algorithm can suggest pairs, but a qualified reviewer should resolve ambiguous, split, and merged events under the declared rule. Keep unmatched events visible in a separate count.

Segmenting one long event into two events can change both the pairing and the mean. Preserve original timestamps, any correction, the software version, and the final match identifier. Report matched-event coverage as matched pairs divided by all events due for matching under the chosen rule.

Reproduce each ratio before taking the mean

Pia's three unrounded ratios are 30/33 = 0.90909..., 20/25 = 0.80, and 45/50 = 0.90. Their sum is 2.60909..., and dividing by three gives 0.86969..., displayed as 87.0%. Calculate with the original seconds and round only the final result.

Pooling duration first would answer a different question. The smaller observer total is 95 seconds and the larger total is 108 seconds, giving 95/108 = 88.0% after rounding. The duration-per-occurrence mean gives each matched event equal weight; the pooled ratio gives longer events more influence. Label either result by its actual method.

Use the formula and units for Pia

For Pia's duration-per-occurrence IOA worksheet: duration-per-occurrence IOA = average of smaller duration / larger duration × 100 for each matched event. Keep observer records independent until both are locked. Preserve the matched exposure, raw observations, numerator, denominator, unrounded calculation, and final rounding.

Work the arithmetic for Pia

Pia's observers record matched durations of 30 and 33 seconds, 20 and 25 seconds, and 45 and 50 seconds. Event agreements are 90.9%, 80%, and 90%. Their mean is 260.9 / 3 = 87.0% after rounding.

Apply matched-event duration agreement to the decision

Pia's qualified clinician identifies the decision before choosing an agreement method or threshold. Compare a stricter or complementary method when prevalence, totals, or aggregation could hide relevant disagreement. Repair definitions, visibility, training, sampling, or matching before judging the implementer or client data.

Audit observer and sample conditions for Pia

Pia's record identifies observers, independence, training, recording mode, shared exposure, response definition, procedure version, observation coverage, missing pairs, reactivity, and disagreements. Consent, privacy, recording, employment, supervision, and payer rules remain separate requirements. Keep every open pair visible with an owner and next action.

Protect client access during Pia's observation

Pia's observation keeps AAC, interpreters, mobility, food, water, bathroom use, prescribed care, health support, rest, relationships, and emergency help available. Do not create extra distress or remove ordinary supports to obtain an agreement sample. Record when access, health, withdrawal, or observer presence changes the condition.

Ask five IOA questions for Pia

Use these questions in the duration-per-occurrence IOA worksheet:

  • Did both observers independently score the same exposure, unit, definition, and condition?
  • Which calculation, numerator, denominator, prevalence, and rounding reproduce the result?
  • Which disagreements, missing pairs, invalid units, reactivity, or matching problems remain?
  • Which validity, treatment-integrity, client-outcome, experience, and access questions stay separate?
  • Which repair, complementary method, next sample, owner, and review date follow?

Keep raw observer records and disagreement types available.

Use current sources within their scope for Pia

For Pia's duration-per-occurrence IOA worksheet, the BACB ethics hub points to current professional ethics materials. The CASP public summary provides high-level autism-treatment scope without licensed detailed procedures. The BACB Ethics Code addresses competence, client involvement, consent and assent when applicable, documentation, risk, supervision, and evaluation for covered behavior analysts. The BCBA Test Content Outline includes measurement, reliability, sampling, and data-based decisions as examination content rather than a case protocol.

For Pia, 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 ABA research corpus. Reed and Azulay describe a calculation tool and agreement methods. ASHA says AAC users should always have access to their tools or devices. These sources do not create one universal IOA method or threshold.

Close Pia's agreement review

Ask Pia and the responsible clinician to review the duration-per-occurrence IOA worksheet through accessible communication. Record the method, arithmetic, sample, disagreements, limits, measurement repair, selected decision, next observation, and review date. Reopen the analysis when the definition, observer, system, exposure, prevalence, client access, procedure, or decision changes.

Clinician release checklist

  • Is every event pair linked by a prospective matching rule and stable identifier?
  • Can another reviewer reproduce 90.9%, 80%, 90%, and the 87.0% mean from raw seconds?
  • Are unmatched, split, merged, missing, and invalid events reported separately?
  • Are start-time and stop-time differences reviewed in addition to duration similarity?
  • Is the 88.0% pooled-duration view labeled as a different weighting method if reported?

Related resources

Sources