An ABA treatment integrity observation form should let a second qualified reader reconstruct what was supposed to happen, what the observer could actually see, and how every reported percentage was obtained. The form below is a copyable starting point for one version-controlled procedure. It deliberately keeps implementation, observation coverage, client outcomes, and follow-up decisions in separate fields.

Clinicians & ABA Professionals / Data, Outcomes and Clinical Decision-Making.

The worksheet is educational, not a universal clinical or employment standard. Adapt it to the client, procedure, setting, service agreement, payer rules, applicable law, and the practice's approved quality process. A percentage cannot establish that a procedure is appropriate, effective, safe, accessible, or acceptable to the client.

What this form measures and what it leaves open

Treatment integrity describes how closely an observed implementation matched a defined procedure. The peer-reviewed review on the impact of treatment integrity on intervention effectiveness explains why implementation of an independent variable matters when interpreting change. A separate practice-focused paper on data reliability and treatment-integrity monitoring discusses efficient collection and the need to connect measurement with feedback. Neither source turns one fidelity score into a causal conclusion.

Use the form to answer a narrow question: during the sampled opportunities, which observable procedure steps were implemented as currently defined? Review client outcomes, adverse or unwanted effects, social validity, dignity, assent or dissent, feasibility, and access in their own evidence streams. A well-implemented plan can still be ineffective or inappropriate. A helpful plan can appear to have low integrity when the checklist is ambiguous, the wrong version is used, or the observer misses key events.

The BACB Ethics Code hub identifies the current code resources. The linked Ethics Code for Behavior Analysts addresses competence, accurate representation, confidentiality, client and stakeholder responsibilities, documentation, supervision, assessment and behavior-change programs. Apply the actual standards and any other governing requirements to the people and service at issue; this worksheet does not interpret an ethics case.

Define the observation before collecting scores

Complete the header before an observer marks a step. If the underlying plan changes, create a new version instead of silently changing the old scoring key.

Observation fieldEntryClient or case codeProcedure name and versionClinical owner and approval dateImplementer roleObserver role and relationshipDate, setting and service contextScheduled window / observed windowOpportunity definitionPlanned opportunity sampleRecording method and authorized accessClient assent/dissent signals and responseSafety or stop rulePrespecified review route

An opportunity definition needs a beginning and an end. “Every trial” is incomplete if one person counts a repeated prompt as a new trial and another does not. State whether naturally occurring opportunities, planned teaching opportunities, interrupted sequences, caregiver-led moments, and client-declined opportunities enter the denominator.

The observer also needs enough independence to report what occurred. If the observer designed the procedure, trains the implementer, controls employment action, or is scoring their own performance, record that relationship. It does not automatically invalidate an observation, but it changes how the evidence should be interpreted and reviewed.

Build the step key

Write one observable action or omission per row. Avoid labels such as “used good judgment” or “implemented correctly.” Those labels hide the scoring rule inside the observer's opinion.

StepObservable definitionWhen applicableCorrectIncorrectCritical?Evidence limitsAYes / NoBYes / NoCYes / NoDYes / NoEYes / No

Use five response codes:

  • C, correct: the defined step was applicable, visible, and met the definition.
  • I, incorrect: the step was applicable and visible, but did not meet the definition or a prohibited action occurred.
  • NA, not applicable: the procedure's prespecified condition did not occur. Exclude it from the step's integrity denominator.
  • NO, not observed: the condition may have occurred, but the observer could not see or hear enough to score it. Do not convert this to correct.
  • IE, insufficient evidence: the record cannot support C, I, or NA because the definition, version, timing, or source evidence is inadequate. Stop and resolve the defect before calculating a formal result.

Keep “critical” as a clinical flag, not an arithmetic weight. A critical safety or consent step can require immediate review even when the overall percentage is high. Weighting a critical step two or three times changes the estimand and should never happen as an undocumented spreadsheet shortcut.

Copyable opportunity log

OpportunityTime / contextABCDEClient response or context, not an integrity scoreObserver note123456789101112

Add rows rather than squeezing several events into one cell. If observation is continuous rather than opportunity based, define the time or interval unit and use a matching denominator. Do not mix opportunity, interval, duration, and permanent-product scores in one percentage.

Calculate integrity and coverage separately

For a step or the entire observed form:

integrity percentage = correct scores / (correct scores + incorrect scores) × 100

Report the fraction with the percentage. NA, NO, and IE do not become correct. NA is excluded because the step truly did not apply; NO and IE should be reported as evidence gaps and may prevent interpretation.

For planned opportunity sampling:

opportunity coverage = observed eligible opportunities / planned eligible opportunities × 100

Coverage tells the reader how much of the planned sample was captured. It does not repair a biased sample. Also report distribution across people, settings, routines, times, risk levels, and procedure variants when those dimensions matter.

If a second observer independently scores a sample, use the agreement method selected before review. For exact cell agreement in this template:

exact agreement = matching step scores / independently double-scored step cells × 100

Agreement is not accuracy. Two observers can agree on a flawed definition. The review of treatment-integrity reporting in behavior-analytic practice is useful context for reporting fidelity methods and limitations, not a mandate for one universal formula.

Fictional worked example

This example is synthetic and is not a recommendation for a real client. A fictional team observes version 3 of a five-step choice-and-transition routine. Step B is “present the agreed cue only after the client-orientation condition in the plan is visible.” Step C is “wait the defined five-second interval before the first prompt unless the safety exception applies.” Both are marked critical. The team planned to observe 15 eligible opportunities but the client used a break option and the session ended after 12. In opportunity 6, Step C was NA because the client completed the response before a wait or prompt sequence could apply.

OpportunityAB*C*DECorrectIncorrectNA1CCCCC5002CCICC4103CCCIC4104CCCCC5005CICCC4106CCNACC4017CCCCI4108CCCCC5009ICCCC41010CCICC41011CCCCC50012CCCCC500Total11/1211/129/1111/1211/125361

The overall result is 53 / (53 + 6) × 100 = 89.8%. Opportunity coverage is 12 / 15 × 100 = 80.0%. Step B is 91.7%; Step C is 81.8%. The critical-step record preserves one Step B error and two Step C errors rather than letting the 89.8% total hide them.

A second observer independently rescored opportunities 3, 8, and 12. Fourteen of 15 step cells matched, so exact cell agreement is 14 / 15 × 100 = 93.3%. The single disagreement stays visible and prompts a definition check. This small convenience sample does not prove observer accuracy or adequate agreement across the full procedure.

The appropriate follow-up is not “the implementer passed” or “the treatment failed.” The team reviews whether the Step C definition was practical, whether the missed wait changed client experience or outcomes, whether the observation sample represented ordinary use, and whether instruction, modeling, rehearsal, feedback, environmental support, procedure revision, or another clinical response is warranted.

Record follow-up without rewriting the observation

Review fieldEntryWhat the data supportWhat remains unknownCritical-step reviewClient outcome / adverse-effect evidence reviewed separatelyClient and caregiver perspectiveFeasibility or accessibility barrierImmediate safety action, if anyCoaching, environment or plan actionOwner and due dateReobservation planProcedure version retained / superseded

Do not alter an old observation to match a new definition. Retain the old key, raw codes, denominator, notes, reviewer decision, and change history according to the applicable record policy.

Privacy, recording, and respectful use

Use the least identifying data needed for the purpose. Confirm who may observe, receive client information, create a recording, and access the completed form. The HHS HIPAA Privacy Rule overview describes federal privacy protections for covered entities and business associates; the HHS Security Rule risk-analysis guidance emphasizes assessing electronic protected health information across the organization. Those sources do not decide whether a particular observation or recording is permitted. Apply contracts, consent, organizational policy and other law as well.

Treatment-integrity observation should support safer, more accurate care. It should not become surveillance detached from a clinical purpose, a hidden employment score, or a substitute for speaking with the implementer and client. Separate clinical coaching from disciplinary processes and give people a fair route to correct an inaccurate record.

The CASP ABA Practice Guidelines access page identifies Version 3.0 and its licensing terms. Reference the guidelines within the applicable license rather than copying protected content into this worksheet. The form remains an original educational artifact and does not claim CASP endorsement.

Stop conditions

Pause scoring or interpretation when the procedure version cannot be identified; a step is not observable; applicable and not-applicable conditions are undefined; the observer lacks enough access; recording or disclosure authority is unresolved; observation would create risk or override assent/dissent; the sample omits the context under review; the denominator cannot be reconstructed; a critical event needs immediate clinical or safety escalation; or someone proposes using the result outside the approved clinical, quality, credentialing, payer, legal, or employment process.

Related resources

Sources