EPR cheat sheet

EPR practice is governed by a code of ethics covering integrity, competence, confidentiality, and avoidance of conflicts of interest; breaches can trigger disciplinary or legal consequences.

Informed consent is a precondition for assessment and intervention: the individual (or lawful substitute decision-maker) must receive purpose, risks, benefits, alternatives, and the right to withdraw.

Confidentiality is not absolute: disclosure may be required by law (e.g., imminent risk of serious harm, court order, statutory reporting) and must be limited to the minimum necessary.

Data protection rules require lawful basis for processing, purpose limitation, storage limits, and secure handling of identifiable assessment records.

Scope of competence defines the boundary of practice; working outside training or qualifications without supervision is an ethical violation, not merely a quality issue.

Dual relationships that impair objectivity or exploit power imbalance must be avoided or managed with documented safeguards.

Regulatory bodies set entry-to-practice requirements, continuing competence obligations, and complaints/discipline processes; registration status should be verified.

Ethical decision frameworks typically sequence: identify the dilemma, gather facts, consult code and law, consider stakeholders, choose and document the least harmful option, review outcome.

Record-keeping duties (retention periods, accuracy, access rights, correction of errors) are legal obligations, not administrative preferences.

Evidence-based practice combines best available research, practitioner expertise, and the individual's values and preferences; none of the three alone is sufficient.

Validity = the instrument measures what it claims; reliability = consistency of results across occasions, raters, or items. Reliability is necessary but not sufficient for validity.

Sensitivity = proportion of true cases correctly identified; specificity = proportion of non-cases correctly identified. Raising one threshold typically lowers the other.

Positive predictive value depends on prevalence: the same test yields lower PPV in low-prevalence populations.

Standard scores (z, T, IQ-type) express distance from the mean in SD units; percentile ranks are not equal-interval and must not be averaged.

Norm-referenced comparison requires matching the norm group on relevant variables (age, language, education, culture); mismatched norms bias interpretation.

Baseline is the pre-intervention level of the target behavior; goal setting compares post-intervention performance against this baseline.

Formative evaluation monitors progress during implementation; summative evaluation judges overall outcome after implementation.

Fidelity = degree to which an intervention is delivered as designed; low fidelity weakens any conclusion that the intervention caused the outcome.

Assessment follows a sequence: referral question, record review, standardized and informal measures, integration, interpretation, and feedback.

Multi-method, multi-source assessment (tests, observations, interviews, informant reports) reduces single-source bias and increases confidence in conclusions.

Behavioral observation should specify setting, antecedent, behavior in observable terms, consequence, frequency, duration, and intensity.

A written plan states the target, measurable goal, method, responsible person, timeline, and criteria for success; vague goals cannot be evaluated.

SMART goal criteria: Specific, Measurable, Achievable, Relevant, Time-bound.

Documentation must be contemporaneous, dated, signed, and factual; opinion must be labeled as opinion and distinguished from observed data.

Report writing separates data description from interpretation and states limitations, including any deviations from standardized administration.

Consent, assent where applicable, and language/accessibility accommodations must be documented before testing begins.

Differential consideration requires ruling out plausible alternative explanations (medical, sensory, linguistic, cultural, motivational) before finalizing conclusions.

Intervention selection matches the assessed need, the evidence base, and the individual's context; the least restrictive effective option is preferred.

Task analysis breaks a complex skill into teachable components sequenced from simplest to most complex or in natural performance order.

Prompting hierarchy typically runs from least to most intrusive (natural cue, verbal, gestural, model, physical); prompts are faded to prevent dependency.

Reinforcement increases behavior; punishment decreases it. Negative reinforcement removes an aversive stimulus and therefore increases behavior - it is not punishment.

Extinction removes the maintaining reinforcer; an initial extinction burst (temporary increase in behavior) is expected and should be anticipated.

Generalization is promoted by training across settings, people, and materials, and by reinforcing naturally occurring instances.

Intervention integrity checks include direct observation, session checklists, and self-report compared against the written protocol.

Implementation should include a crisis/risk protocol specifying triggers, escalation steps, and who is notified.

Dosage matters: frequency, duration, and intensity of sessions must be specified and delivered as planned for outcomes to be attributable.

Progress monitoring uses repeated, brief, comparable measures at fixed intervals; the measure must be sensitive enough to detect small change.

Graph data over time and apply a decision rule (e.g., four consecutive data points below the goal line) rather than reacting to single points.

Trend line slope indicates rate of progress; compare obtained slope to the expected or goal slope to decide whether to continue or modify.

Baseline logic (A-B-A-B or multiple baseline) supports causal inference better than pre-post comparison alone.

Quality indicators should be measurable, benchmarked, and reviewed on a defined cycle with documented actions and owners.

Plan-Do-Study-Act cycles test small changes, study results, and scale or abandon based on data.

Evaluation distinguishes process measures (was it delivered as planned?) from outcome measures (did the target change?).

Data quality checks cover completeness, inter-rater agreement, and consistent definitions of the target behavior across time and raters.

Program evaluation reports limitations: attrition, confounding variables, regression to the mean, and measurement reactivity.

Plain-language communication avoids jargon; technical terms are defined on first use and key messages are limited to a few per conversation.

Feedback structure: describe the specific observed behavior, state its impact, invite the recipient's perspective, agree on next steps.

Collaboration requires explicit role clarification (who does what, by when) and a named point of coordination to prevent duplication or gaps.

Stakeholder mapping by influence and interest guides engagement effort; high-influence/high-interest parties need active, frequent involvement.

Interpretation feedback to individuals follows a strengths-first, collaborative model; results are explained in context and checked for understanding (teach-back).

Conflict is de-escalated by separating positions from interests and seeking integrative options rather than win-lose outcomes.

Consent for information sharing requires specifying what is shared, with whom, for what purpose, and for how long; consent is revocable.

Cross-cultural communication accounts for language proficiency, interpretation needs, and differing norms about family involvement in decisions.

Documenting communication (dates, participants, decisions, agreed actions) protects continuity and provides an audit trail.