Procedural fidelity

Fidelity connects the written plan with delivered practice

Outcome data alone may show improvement, little change, or deterioration. Fidelity data add a second question: what parts of the planned procedure were actually available and carried out?

A strong outcome with low measured fidelity may mean the scored steps were unnecessary, the definition was weak, or another factor helped. A weak outcome with high fidelity may prompt reassessment of fit, goals, risks, or intervention design.

Define critical steps and eligible opportunities

List observable actions in the order and conditions that matter. Identify critical safety, communication, assent, reinforcement, response, documentation, and stop steps. State which actor owns each action.

The denominator should contain steps that were due in valid opportunities. If preparing materials is an assigned step, missing materials count as a miss for that step. If a planned opportunity never occurs for a reason outside the observed person’s responsibility, record the environmental failure separately and follow the predeclared exclusion rule.

Do not exclude difficult trials merely because the procedure was harder to implement. Report all exclusions with reasons.

Calculate the basic percentage carefully

A common calculation is:

correctly completed steps ÷ applicable steps due × 100.

Suppose 20 critical steps are due across an observation and 18 are completed as defined. Procedural fidelity is 18 ÷ 20 × 100 = 90%.

That percentage needs its unit, observer, setting, date, procedure version, and context. Ten trials with two steps each differ from one trial with twenty steps. A global score can also hide a missed safety step, so show critical components separately.

Predeclare the review rule and avoid treating it as a universal clinical cutoff. A 90% score may still require immediate action when the missed item protects breathing, communication, assent, medication, or another essential safeguard. Repeated misses on a low-risk administrative step may call for a different response.

A fictional case separates implementation from access

Lena is a fictional clinician supporting Omar’s chosen goal of using speech or AAC to ask for clarification during a community class. The procedure includes making AAC available, presenting an accessible instruction, waiting, recognizing either message form, answering clearly, and honoring a pause request.

Across four observed opportunities, 20 critical steps are due. Staff complete 18, giving 90% fidelity. One missed step is a late partner response. The other is a missing AAC backup at arrival.

The team also checks environment readiness across eight class arrivals. All required materials and communication supports are ready in 7 of 8, or 87.5%. The scores remain separate because step implementation and system readiness have different owners and remedies.

Omar communicates in three of the four observed opportunities, and partners honor all three messages. These outcome counts describe the observed opportunities. They do not prove that 90% fidelity caused the result.

Measure agreement on fidelity observations

When fidelity scores drive supervision, mastery, safety, or plan changes, train observers on the same definitions and examples. Sample independent agreement across relevant conditions and report interobserver agreement separately.

High observer agreement cannot repair an ambiguous step or a biased opportunity sample. Review disagreements at the raw-step level before averaging scores.

Use patterns to improve the system

Look for steps repeatedly missed across people or settings. The cause may be unclear wording, inadequate training, inaccessible technology, unrealistic timing, workload, missing materials, conflicting payer instructions, or a procedure that does not fit the person.

Respond with clarification, modeling, rehearsal, feedback, environmental changes, clinical reassessment, or plan revision as appropriate. Avoid tying access to care, employment dignity, family approval, or clinician worth to a single percentage.

Keep client outcomes and experience visible

Fidelity is an implementation measure. It does not establish benefit, clinical significance, consent, assent, safety, generalization, maintenance, or acceptability.

Pair it with the person’s direct report, outcome measures, access measures, unwanted effects, burden, and contextual changes. A perfectly implemented plan that is distressing, ineffective, or poorly fitted needs qualified review.

Document the decision and its follow-up date.

The CASP ABA Practice Guidelines Version 3.0 public summary places planning, implementation, and evaluation within ABA treatment for people diagnosed with autism. The BACB BCBA Test Content Outline, 6th edition includes measurement and implementation concepts as examination content. CASP’s resources page offers organizational resources. These sources support careful measurement while leaving case decisions to qualified roles.