Clinical
Clinical Data Collection
Coralia's Clinical Data Collection module is the behavior-target and session-data system inside Coralia, an AI-first practice management platform for applied behavior analysis (ABA) therapy agencies. Board Certified Behavior Analysts (BCBAs) program goals with one of ten measurement systems and five-part mastery criteria; Registered Behavior Technicians (RBTs) record data through entry widgets matched to each measurement type, and every datum is validated against the program, versioned on edit, and tied to the billable session note.
The problem in real agencies
Payers reimburse a 97153 session — the Current Procedural Terminology (CPT) code for technician-delivered applied behavior analysis (ABA) treatment — on the strength of its documentation. When data lives on clipboards or in a separate graphing app, the note and the data drift apart: claims go out for sessions whose notes show blank targets, and a payer audit two years later asks why. Transcribing data after the fact bets the agency's revenue on nobody ever comparing the two records.
The clinical record drifts the same way. A hand-typed 'current level' on a treatment plan disagrees with the graph within weeks. Mastery gets declared from memory — a Board Certified Behavior Analyst (BCBA) remembers three strong sessions but not whether they crossed two settings or two observers. And when someone fixes a mistyped datum, nothing distinguishes the honest correction from a back-edit, which is precisely the distinction an auditor is trained to probe.
Measurement consistency breaks last. If a Registered Behavior Technician (RBT) can pick frequency today and percentage tomorrow, or log trials under a goal that has no operational definition, two technicians end up recording different behaviors under the same target name — and the trend line stops meaning anything. The measurement decision belongs to the BCBA who programmed the goal, not to whoever happens to run the Tuesday session.
How it works in Coralia
- 1
Program the goal
On a client's Treatment tab or in the assessment prep workspace, a BCBA defines each behavior target: one of three types (maladaptive, replacement, or skill), one of ten measurement systems, behavior functions, a clinical domain, baseline and goal values, priority, and start and target dates. Mastery criteria are a five-part composite: performance threshold, consecutive qualifying data points, across-observer and across-setting generalization, and a maintenance schedule. Saving an active goal without an operational definition or a mastery criterion triggers an explicit 'save as incomplete' warning — because two RBTs could otherwise record different behaviors under the same goal.
- 2
Sequence objectives under the goal
Each target carries long-term and short-term objectives (LTOs and STOs) with a description, a criterion value and unit, a consecutive-sessions requirement, and a sequence. Objectives hold one of three statuses: not started, in progress, or mastered. During data entry, the session screen shows the currently active objective directly above the entry widget, so the RBT always knows which step of the program today's data serves.
- 3
Collect with the widget the program demands
On a draft session note, each target renders the entry widget its measurement type requires: a tap counter for frequency, rate, and permanent product; per-trial correct/incorrect entry with a six-level prompt hierarchy (independent through full physical) for percentage targets; plus/minus seconds entry for duration, latency, and inter-response time (IRT); and an interval grid for the three interval methods. The RBT never picks the modality at session time. The server rejects any datum whose structure does not match the target's declared measurement system, and rejects duplicate entries for the same target within one session.
- 4
Gate the note on real data
A technician's 97153 direct-care note cannot be submitted unless every selected target carries at least one real data point — a positive count, recorded seconds, or at least one trial — or a written no-data reason, such as the client leaving early. Agencies can additionally require a minimum number of data-bearing targets per note. Data entry is only possible while the note is in draft or needs-revision status; submitted and approved sessions are locked, and only their data feeds trends and current levels.
- 5
Version every correction
Editing or deleting a recorded measurement first snapshots the prior values — including note text, kept encrypted verbatim — into an append-only version history inside the same database transaction. Deletes are soft archives that remain visible to auditors; nothing is hard-deleted. Behavior-target and target-data changes emit audit-log events, and every target status change writes an attributed history row: who changed it, when, from and to which status, with an optional data-basis rationale.
- 6
Derive progress, confirm mastery
A target's displayed current level is derived from the most recent submitted or approved session datum — never hand-typed — so the treatment card, progress chart, and sidebar always agree, and zero is treated as a real measurement. The Progress tab charts each target against baseline and goal lines over a 4-, 8-, 12-, or 24-week window with mean, standard deviation, a least-squares trend slope (direction-inverted for maladaptive targets), and a 3-point moving average. Clicking 'Master' runs the full mastery check live and shows the verdict; the clinician confirms manually, and status moves through a guarded lifecycle.
The specifics
Ten measurement systems per target: frequency, rate, permanent product, percentage, duration, latency, inter-response time, partial interval, whole interval, and momentary time sampling.
Six prompt levels on trial entry: independent, verbal, gestural, model, partial physical, and full physical.
Per-trial outcomes are stored as an ordered array; totals are server-derived from trial positions, so counts and positions can never disagree.
Five-part mastery criteria: performance threshold, consecutive qualifying data points, across at least two observers, across at least two settings, and a maintenance schedule.
A guarded five-state target lifecycle — active, on hold, mastered, maintenance, discontinued — rejects illegal jumps like mastered-to-active before any write.
Trend statistics per target: mean, standard deviation, least-squares slope classified improving/declining/stable, and a 3-point moving average, over 4-24 week windows.
A batch trends request returns history for up to 40 targets, powering per-row sparklines, previous-session deltas, and mastery chips on the note.
ABC (antecedent-behavior-consequence) narratives are encrypted at rest; lists return up to 200 newest records, and deletion archives rather than erases. Built for HIPAA compliance.
The Data Center workbench charts maladaptive, replacement, and skill targets together — up to six on one comparison chart.
Three target types — maladaptive, replacement, skill — across six clinical domains, with five pickable behavior functions per target.
Integrations
Session notes and the billing gate — collected target data prints on the session note and gates 97153 submission; only submitted or approved data feeds trends and current levels · Assessments and treatment plans — report sections (maladaptive summary, behavior detail, replacement detail, ABC data, intervention matrix) auto-populate from live targets and embed per-target trend charts · The Coralia Brain, the per-agency AI copilot — records target data and behavior incidents through approval-gated tools that run the same validated server checks as manual entry · Audit log — every behavior-target and target-data create, update, and archive emits an audit event
Access control
Four permission codes gate the module: separate view and edit permissions for client programs (targets, objectives, ABC records) and for session data (trends, mastery checks, in-session entry), so program design and data entry can be granted to different roles. A per-client access guard applies on top of the permission codes for trend and mastery views.
Frequently asked questions
What measurement systems does Coralia support for ABA data collection?
Ten: frequency, rate, permanent product, percentage (per-trial correct/incorrect), duration, latency, inter-response time (IRT), partial interval, whole interval, and momentary time sampling. The BCBA picks the system when programming the target; during the session the RBT sees the matching entry widget — tap counter, trial entry with prompt levels, seconds entry, or interval grid — and the server rejects any datum whose shape does not match the declared system.
Can an RBT submit a session note without collecting data?
Not for direct care. A 97153 note cannot be submitted unless every selected target shows real recorded data — a count, timed seconds, or a completed trial — or carries a written reason no data was collected, for example the client left early. Agencies can also require a minimum number of targets with data per note. After submission the data is locked; edits reopen only if the note returns for revision.
How does Coralia decide when a goal is mastered?
It does not decide — it checks. The mastery evaluator tests the criterion — performance threshold, consecutive qualifying data points, across at least two observers, across at least two settings — against recent submitted or approved session data, direction-aware so maladaptive targets qualify at or below the bar. When a clinician clicks Master, a dialog shows the live verdict and the clinician confirms manually. Status changes then follow a guarded lifecycle with attributed history.
What happens when recorded session data is edited or deleted?
Every edit or delete first snapshots the prior values into an append-only version history inside the same transaction, so an honest correction is distinguishable from a silent back-edit. Deletes are soft archives that auditors can still see — never hard deletes. All target-data and behavior-target changes also emit audit-log events, and data can only be changed while the note is in draft or needs-revision status.
Can I try Coralia's data collection before signing up?
Yes. Coralia runs a live demo at coralia.app/demo — a complete synthetic agency, no sign-up required. You can open a client's Treatment tab to read target definitions and mastery criteria, walk through the entry widgets on a session note, and view the Progress tab charts with baseline and goal lines. Nothing in the demo is real client data.