Procedural Integrity and Effectiveness Decisions
Key Takeaways
- Effectiveness data and procedural integrity data must be interpreted together.
- Poor outcomes with low integrity do not prove the intervention is ineffective.
- High integrity with poor outcomes suggests the plan may need a functional, procedural, or goal-level revision.
- Data-based decisions should include level, trend, variability, immediacy, risk, and social validity.
Integrity Plus Outcomes
Procedural integrity is the extent to which implementers carry out the intervention as planned. Effectiveness is the extent to which client behavior changes in the desired direction. Domain H questions often require reading both data streams before changing treatment.
If integrity is low, first improve training, materials, prompts for staff, performance feedback, or contextual fit. Do not abandon a sound plan before it has been tested under adequate implementation.
If integrity is high and behavior is not improving, review the assessment hypothesis, reinforcer value, prompting, schedule, criteria, measurement system, and social validity. A plan can be implemented perfectly and still be wrong for the case.
Decision Matrix
| Outcome data | Integrity data | Likely decision |
|---|---|---|
| Improving | High | Continue and plan fading or generalization. |
| Improving | Low | Strengthen integrity before drawing conclusions. |
| Flat or worse | Low | Fix implementation and fit first. |
| Flat or worse | High | Modify the intervention based on assessment and data. |
Exam Rule
Never choose "continue unchanged" only because the written plan looks good. Never choose "change everything" without checking integrity, measurement validity, and risk. The correct action is usually a specific data-based next step.
A graph shows no reduction in self-injury after two weeks, and integrity checks show staff implemented the plan correctly on only 42% of opportunities. What should the BCBA do first?
A replacement communication intervention has 95% procedural integrity for eight sessions, but problem behavior remains stable and the communication response rarely occurs. Which action is most defensible?
Which data pattern most strongly supports continuing an intervention while planning generalization?