Risk-benefit analysis

Start with a specific decision

“Is this plan worth it?” is too broad. Name the person, goal, setting, procedure, intensity, duration, decision owner, review date, and evidence available. The analysis for teaching an AAC-based help message differs from the analysis for community travel, feeding, sleep, self-injury, or a restrictive procedure.

List realistic options. Depending on the question, these may include environmental change, communication access, medical or interdisciplinary referral, caregiver or staff training, a lower-intensity approach, another evidence-based procedure, added support, a time-limited pilot, waiting with monitoring, or continuing current supports. Describe what each option actually requires.

The CASP ABA Practice Guidelines Version 3.0 public summary concerns ABA behavioral health treatment for people diagnosed with autism and frames decision-making around medically necessary, efficacious, and cost-effective treatment. The detailed guidelines require a license. This article uses only that public scope and does not present its framework as a CASP procedure.

Compare benefit that matters to the person

Define the proposed benefit in observable, accessible, and personally meaningful terms. Consider magnitude, likelihood, timing, durability, generalization, required supports, and whether the outcome is valued by the person and family. A change in a convenient clinic measure may have little daily benefit.

Ask what the person wants more or less of and how they will communicate willingness, discomfort, pause, or withdrawal. Where someone else has legal decision authority, keep that authority distinct from the person's assent, dissent, preferences, and experience.

AHRQ defines shared decision making as a collaborative process informed by evidence, care-team knowledge and experience, and the patient's values, goals, preferences, and circumstances. Its SHARE Approach, reviewed in February 2026, supports dialogue about benefits, harms, risks, and what matters to the patient. These are general healthcare resources rather than ABA practice rules.

Describe risk across several domains

Consider the severity, likelihood, duration, detectability, reversibility, and distribution of each possible harm or burden. Relevant domains can include:

• physical injury, pain, fatigue, medical risk, or delayed medical care
• fear, distress, trauma, dignity, autonomy, privacy, and communication access
• loss of useful behavior, masking, stigma, social exclusion, or dependence
• time, travel, school, work, family burden, financial cost, and opportunity cost
• staff competence, supervision, fidelity, setting, equipment, and emergency readiness
• cultural, language, disability, and environmental fit
• rare high-severity events that an average score could hide

Separate uncertainty from low risk. Missing evidence, unfamiliar settings, small samples, and unreliable measures make the estimate less certain. They do not make the harm unlikely.

Include waiting and current-support risks

Every comparison needs a credible reference option. Continuing current supports may preserve stability while leaving a chosen goal unmet. Delaying a decision may allow more assessment while prolonging pain, exclusion, caregiver burden, safety exposure, or loss of access. A “no new intervention” option can still require communication, environmental, medical, or safety support.

Avoid presenting treatment as risk-free or delay as neutral. Record what is expected under each option, how long the comparison applies, and what evidence would change the conclusion.

Use safeguards, stop rules, and reversible steps

Prefer an option with meaningful benefit, lower burden, stronger fit, and reasonable evidence. When uncertainty remains, a smaller and more reversible step may allow learning while limiting exposure. Reversibility is one factor; it cannot rescue an otherwise unacceptable risk.

Before starting, specify:

• required qualifications, training, supervision, setting, equipment, and communication access
• baseline health and safety information plus needed medical or interdisciplinary input
• consent and assent process, accessible refusal or pause response, and partner action
• measures for intended benefit, side effects, burden, fidelity, and client experience
• absolute stop conditions, pause-and-review thresholds, escalation owners, and emergency route
• review frequency, maximum pilot duration, documentation, and decision authority

Immediate danger follows the applicable emergency process. A routine review meeting or payer call should never delay emergency help.

Ethics standards set minimum responsibilities

The BACB ethics hub identifies the current Ethics Code for Behavior Analysts. It applies to BCBA and BCaBA certificants and people who completed an application. Standards 2.14 and 2.15 address risks, benefits, side effects, preferences, context, positive reinforcement, and minimizing risk. Standard 2.18 calls for continual evaluation and timely modification when data indicate that risk exceeds benefit or the intervention is ineffective.

The Code does not give BACB separate jurisdiction over organizations or corporations. A practice needs role-specific policy, escalation, supervision, documentation, and review. Clinical judgment remains with an appropriately qualified clinician; software and administrative staff can surface evidence, missing gates, and due dates.

A fictional staged decision

Nia is a fictional fourteen-year-old who wants to ride a community bus to the library. Nia uses AAC and has twice moved toward the curb before the bus stopped. The team compares current family transport, immediate full-route practice, and staged practice at a quiet stop with a familiar support person.

Nia selects the staged option. Before each practice, adults verify charged or backup AAC, an agreed stop message, route conditions, a safe waiting position, and defined adult roles. All five gates are ready in 6 of 6 scheduled practices. Nia chooses to proceed in five and declines once; the decline is honored, so participation is 5 of 6, with no penalty.

Across the five completed practices, Nia uses the agreed stop or help message in 4 of 5 defined opportunities, and the partner follows the planned response in 4 of 4 messages. No curb entry or injury occurs in this small sample. The observations support the next review; they cannot establish future safety, causation, or superiority over every option. Nia's report, travel value, burden, near misses, ordinary supports, and family experience remain part of the decision.

Review when facts or preferences change

Reopen the analysis after an adverse event, new health information, withdrawal of assent, goal change, poor fidelity, weak benefit, new setting, staffing change, extended interruption, or better alternative. Record which evidence changed, who participated, and why the plan continued, changed, paused, or ended.

A numerical matrix can organize thinking. Keep the underlying evidence and uncertainty visible; a single total can conceal a rare severe harm, unequal burden, or client objection.