To monitor treatment integrity after an ABA plan change, define the current procedure version and score both revised and stable steps across relevant people, settings, and conditions. Train and calibrate observers, preserve invalid opportunities, and report agreement separately from fidelity. Link errors to access, materials, training, workflow, or plan clarity before coaching. Continue sampling long enough to verify that repairs persist and that correct implementation remains safe and acceptable to the person.

Lock the procedure version

Identify component, effective date, revised steps, stable steps, permitted adaptations, access conditions, stop rules, training cohort, and superseded materials.

Score the procedure that was authorized for the observed setting, not a generic master description. Put a version identifier on the scoring tool and link each item to the corresponding plan step. Include stable steps because unchanged access or safety requirements can still fail during a revision. Retire obsolete forms while retaining them for historical interpretation, and verify that observers can distinguish the current version from a familiar prior sequence.

Define scoring opportunities

For every step, specify eligible opportunity, correct performance, error, prompt, not observed, invalid event, client withdrawal, environmental failure, and exclusion reason.

Define the denominator before sampling begins. A step cannot be scored incorrect when its opportunity never occurred, and an invalid event should not disappear simply because it cannot enter the percentage. State how partially observed sequences, client requests to stop, device failures, and permitted adaptations will be coded. This makes the integrity result reproducible and protects client or system conditions from being mislabeled as implementer performance.

Sample representative conditions

Cover implementers, supervisors, settings, shifts, routines, client states, materials, devices, and early versus later use without selecting only convenient sessions.

Create a sampling map from the rollout rather than relying on whoever is easiest to observe. Include newly trained and experienced staff, lower-support shifts, community or home contexts, and the devices or materials actually used there. If risk or feasibility requires oversampling one condition, say so. Report the coverage beside the result so a high clinic score is not generalized silently to unobserved settings.

Calibrate observers

Train on the same examples, assess agreement across relevant conditions, report agreement separately, and revisit ambiguous definitions. Consensus by itself does not establish validity.

Use examples and nonexamples that include the likely edge cases, such as a pause request, a failed device, or a permitted change in response mode. Compare observers independently before discussion so agreement is not manufactured through real-time consensus. When disagreement clusters around one item, repair the definition or plan language and document the change. High agreement can show consistent scoring, but it cannot prove that the underlying procedure is appropriate.

Diagnose error patterns

Trace errors to unclear plan language, missing access, materials, scheduling, training, competing workflow, supervision, feasibility, or an unsuitable procedure before choosing repair.

Look for patterns by step, person, setting, time, and prerequisite. Repeated errors across trained staff may point to confusing instructions or an impractical workflow rather than individual noncompliance. A failure that occurs only when AAC is unavailable is also an access and release problem. Preserve multiple contributing factors where they overlap, then assign the clinical, operational, training, or system owner who has authority to address each one.

Verify repair persistence

Set coaching, environmental repair, resampling, stable-performance window, client-experience check, and escalation or plan-revision rule with explicit owners and dates.

Verify the repair in the condition where the problem occurred, then sample again after enough time to show that performance persisted beyond immediate coaching. Do not close integrity monitoring on percentage alone. Review Emmett's experience, unwanted effects, feasibility, and whether correct implementation produced the intended opportunity. A qualified clinician should decide when repeated implementation difficulty suggests that the plan itself needs revision rather than another round of staff correction.

Build Emmett's post-change integrity plan

Create a versioned post-change integrity plan for this post-review implementation question. Preserve direct client input, actual version exposure, health and communication access, qualified authority, staff and setting readiness, procedure integrity, data quality, client experience, unwanted effects, exceptions, rollback, tasks, owners, due dates, communications, correction routes, and later evaluation. Another qualified reviewer should be able to reconstruct what was released, what actually occurred, and how the evidence changed the next decision.

Work through Emmett's example

Emmett's eight-step procedure changes two steps. Across twelve observations, the two revised steps are correct in 19 of 24 scored opportunities, while the six stable steps are correct in 68 of 72. Report 79.2% and 94.4% separately. A combined 87 of 96 would hide the concentrated revision problem. Keep every planned and actual event, version, numerator, denominator, overlap, open state, and unavailable item visible. This fictional three-implementer skill program example illustrates one post-review control and supplies no universal release rule, integrity threshold, clinical recommendation, legal conclusion, payer result, or outcome guarantee.

Address Emmett's main implementation risk

A total integrity score can let many stable steps dilute errors in the changed steps. Emmett's review segments the exact components that need repair. Treat approval, release readiness, actual exposure, integrity, client experience, unwanted effects, and outcomes as separate evidence. Technical completion can occur before clinically acceptable use is established.

Choose Emmett's next action

The supervisor repairs the material and coaching problem, then samples the revised steps across all three implementers before reducing monitoring. Record the responsible role, authority, affected component and scope, immediate control, due date, evidence needed for closure, accessible communication, correction route, and next review. Software may coordinate workflow. Qualified professionals make case-specific decisions within scope.

Protect Emmett's access and choice

Keep Emmett's AAC, interpreters, mobility, food, water, bathroom use, prescribed care, pain communication, health support, movement, rest, relationships, and emergency help available. Offer private and accessible ways to ask, disagree, decline, pause, withdraw when applicable, and correct the record. Proxy and professional input may inform review while Emmett's own experience remains distinct.

Apply current sources to Emmett's implementation

The BACB ethics hub, current Ethics Code, CASP public summary, and BCBA Test Content Outline provide scoped professional, client-involvement, documentation, evaluation, and training context.

An evidence-based ABA framework and treatment-integrity practitioner guide, Essig review, impact study, and reporting review support contextual decisions, explicit implementation evidence, observer quality, and bounded conclusions.

ASHA supports continuous AAC access.

Rehearse Emmett's implementation path

Test the post-change integrity plan with a stale copy, untrained shift, missing AAC, inaccessible material, health change, incorrect step, mixed-version setting, absent observer, client dissent, adverse effect, unauthorized deviation, rollback, overdue task, and reopened decision. Confirm that access, attribution, authority, version state, evidence, and follow-up remain intact.

Close Emmett's implementation record

Review the post-change integrity plan with Emmett, the responsible clinician, affected implementers, and the specialists named by the manifest. Preserve client input, actual exposure, evidence, decisions, disagreement, versions, tasks, limits, and open findings. Keep the page draft and noindex until required clinical, client or family, AAC, access, medical, educational, payer, records, privacy, employment, and legal reviews are complete.

Related resources

Sources