Data collection

Start with a meaningful question

Data collection should answer a named question. Examples include whether communication access is available, how often a chosen request occurs, how long an activity lasts, whether a partner responds, or how the person rates comfort and usefulness.

The person’s priorities and direct communication should shape the goal. Family and clinician observations add context. A convenient number can still be a poor outcome measure.

Match the measure to the response

Common options include:

• count or rate for discrete responses
• duration for total time or time per occurrence
• latency for time from a defined event to response onset
• opportunity percentage for a response possible only in defined situations
• interval or time sampling when continuous observation is impractical
• permanent products when a lasting result validly represents behavior
• self-report, family report, and contextual measures for experience and fit

Write the operational definition, observation boundaries, denominator, supports, prompts, and exclusion rules before collection.

A family-readable record shows the pieces

A useful entry can answer:

• Who or what was observed?
• What was the setting, activity, and observation window?
• Which definition and procedure version applied?
• What were the numerator, denominator, count, or duration?
• Which supports and prompts were present?
• What interrupted observation?
• What did the person and family report?
• Who recorded and reviewed the entry?

A graph point should trace back to this evidence. Corrections should preserve original content, author, date, time, and reason under the applicable documentation policy.

A fictional progress example

Priya is a fictional eight-year-old who wants an easier way to ask siblings to pause a noisy game. With Priya, the team defines an eligible opportunity as noise above the agreed level while her AAC device or backup card is available and a sibling can respond.

Across ten eligible opportunities in one week, Priya communicates “quieter” or “break” in 7 of 10, or 70%. Siblings respond within 15 seconds to 5 of 7 messages, or 71.4%. Two planned observations are invalid because the backup card was missing.

The client percentage, partner-response percentage, and system-access failures answer different questions. The team also records Priya’s report that the pause option felt useful. The week provides descriptive evidence and no causal conclusion.

Missing data are part of data quality

Canceled sessions, blocked views, unavailable communication, device failures, and absent opportunities are different states. None should silently become zero.

Report valid observations divided by planned observations and exclusions by reason. A high result across a small completed subset can hide weak coverage. Age unresolved records so late entries remain visible.

Progress uses patterns rather than one point

One observation can identify an issue worth checking. Repeated comparable observations help show level, variability, direction, and context. Changes in people, settings, definitions, prompts, access, health, or observation time can alter the pattern.

A before-and-after difference does not prove that treatment caused change. Qualified clinicians integrate data with assessment, client and family input, risk, scientific evidence, and contextual fit.

Protect rights, communication, and privacy

Keep AAC, interpreters, mobility and sensory supports, breaks, food, water, bathroom access, prescribed care, relationships, and emergency help available. Record system barriers separately from client performance.

Collect only data needed for the stated purpose. Follow applicable consent, assent, confidentiality, access, security, retention, disclosure, and recording rules. Explain who can see the record and how the person or family can ask for correction.

Ask practical questions at review

Families can ask:

• Which goal does this measure represent?
• Can I see the raw count or denominator?
• What changed in the setting or support?
• How much data are missing?
• What does my child say or communicate about the outcome?
• Which decision would change because of this pattern?

Families can also ask to see how the team handles prompted responses, canceled visits, device failures, and late entries. A plain-language review should show at least one raw example beside the graph. If a measure changes, the clinician should explain the old rule, new rule, effective date, and reason. This keeps apparent progress from being created by a quiet definition or denominator change.

The CASP ABA Practice Guidelines Version 3.0 public summary places assessment, treatment planning, implementation, and evaluation within ABA behavioral health treatment for people diagnosed with autism. It supplies high-level scope and does not prescribe this data-collection workflow.