To calculate component level treatment integrity, divide correctly implemented eligible component opportunities by eligible component opportunities due, then report raw counts. Keep sequence, timing, critical components, missing observations, and inapplicable steps visible. Calculate per-component and per-session results before any aggregate. A weighted or global percentage requires a stated rationale and cannot erase repeated failure of a safety or communication step.

Build the denominator from steps due

Count a component only when its defined cue and required conditions occurred. Mark conditional steps inapplicable when they were never due.

Create one record per eligible component opportunity with the procedure version, cue, context, due status, observed status, and evidence source. Separate not due, unobservable, invalid, adapted, declined, and missing values. Do not place those states in the accuracy denominator without a defined reason. Reconcile component counts to the sessions and routines from which they came.

Calculate more than one view

Report each component's correct count over eligible opportunities, each routine's correct count over steps due, and the pooled result only when pooling is meaningful.

Show raw numerators and denominators beside every percentage and explain weighting. A pooled rate can overrepresent frequent, easy steps and hide rare critical components. Report by implementer, setting, version, and opportunity class only when samples and definitions support those views. Keep small or unmatched cells visible without presenting them as stable ranks.

Flag critical failures

Show communication, safety, consent, health, and active-mechanism steps individually. A strong total cannot offset a recurring critical omission.

Define why a step is critical, what immediate protection or escalation follows a miss, and who reviews the pattern. Keep critical misses visible as counts and dates rather than folding them into one average. A single event may justify prompt safeguarding while remaining insufficient for a broad competence conclusion. Preserve client experience and system conditions around the event.

Handle missing and agreement

Keep unobservable events outside the accuracy numerator and label them missing. Report observer agreement on a separately sampled set using the same definitions.

State observations due and completed, reasons for missingness, and whether absence clusters by setting, person, or critical routine. Calculate agreement at the component or opportunity level and identify the sampled pairs. Agreement does not establish accuracy, but persistent disagreement limits interpretation. Repair visibility or definitions before using disputed values for clinical, coaching, or personnel conclusions.

Version the calculation specification and test it with hand-worked examples, conditional steps, zero-opportunity sessions, partial observations, and valid adaptations. Reconcile automated output to raw records on a sample and investigate discrepancies. Preserve original values and correction history. Release the result only with scope, missingness, critical-component findings, and the decision it is designed to support.

Put the component-level calculation into practice

Farid's routine has five possible components, but conditional steps are due only when their cues occur. Across eight observed routines, 31 component opportunities are eligible and 24 are correct. AAC access is a critical component and is reported separately. Keep two unobservable steps in the missing state.

Compare measurement options for Farid

Farid's team compares direct observation, component scoring, opportunity sampling, duration, permanent product, structured interview, system record, and direct client feedback only when each option can answer the component-level integrity calculation's decision. Record validity, visibility, reactivity, effort, privacy, access, timeliness, and what each source cannot establish.

Pilot Farid's component-level integrity calculation

Test the definition, form, observer instructions, technology, access, timing, missing-data states, and review workflow in representative conditions. Ask Farid and the implementer whether the process changes the routine or adds burden. Revise the measurement system before using it for coaching, treatment change, supervision, or performance decisions.

Protect access and clinical responsibility for Farid

Farid's review keeps AAC, interpreters, mobility, food, water, bathroom use, prescribed care, health support, rest, relationships, and emergency help available. Obtain required consent and assent when applicable and follow the governing response to withdrawal or distress. A qualified clinician retains responsibility for clinical interpretation and treatment decisions; a fidelity score does not transfer that responsibility.

Keep procedure versions and evidence states separate

For Farid, record the active procedure version, date, author, authorized adaptations, temporary responses, observed deviations, missing evidence, and unresolved classifications. Preserve the original record. Add a dated explanation whenever a correction or later review changes the procedure version used to judge an observation.

Ask six integrity questions for Farid

Use these questions in the component-level integrity calculation:

  • Which decision, procedure version, component, cue, and eligible opportunity apply?
  • Which client, implementer, setting, time, and observation conditions were sampled?
  • Which access, safety, health, consent, assent, and communication supports were present?
  • Which integrity, exposure, outcome, experience, adverse-effect, and agreement series stay separate?
  • Which missing evidence, adaptation, drift, burden, reactivity, or concurrent change limits interpretation?
  • Which owner, repair, stop rule, next evidence step, and review date follow?

Keep every unresolved item visible with a state, owner, age, and next action.

A fictional integrity example for Farid

Farid is fictional and involved in a community purchase routine with five conditional steps. Reviewers freeze 34 component, opportunity, sequence, timing, critical-step, missing, inapplicable, agreement, and summary fields and complete 24 of 34 by the checkpoint. Every missing, unobservable, inapplicable, adapted, drifted, failed, or pending field keeps its defined state.

The component-level integrity calculation measures evidence completeness. It does not establish effectiveness, safety, acceptability, medical necessity, authorization, payment, compliance, generalization, maintenance, or causation. Report the applicable raw counts and conditions before a percentage.

Use current sources within their scope for Farid

For Farid's component-level integrity calculation, the CASP public summary supplies high-level individualized assessment, implementation, and evaluation scope for ABA treatment of autistic people. The BACB Ethics Code addresses competence, client involvement, consent and assent when applicable, risk, documentation, supervision, and evaluation for covered behavior analysts. The BCBA Test Content Outline includes procedural fidelity as examination content; it is not a case protocol or universal threshold.

For Farid's component-level integrity calculation, the Ferguson practitioner guide supports observable component and eligible-opportunity design. The Essig, Rotta, and Poling review supports caution about the consistency and interpretation of fidelity reporting in the literature it examined. ASHA's AAC guidance says AAC users should always have access to their communication tools or devices. None of these sources makes one sampling rate, percentage, or repair rule valid for every case.

Close Farid's review

Ask Farid and each implementer to review the component-level integrity calculation through accessible communication. Record the selected decision, raw evidence, procedure version, direct feedback, limitations, responsible role, repair, next observation, and review date. Reopen the analysis when the goal, procedure, client preference, access, health, setting, implementer, risk, or outcome changes.

Related resources

Sources