Industry field guide
Therapy goal, measurement and episode records for AI
A progress entry can use a different condition or measure than the baseline.
- 2022 NAICS
- 621340
- Practice pack
- Example and answer key (JSON)
- Updated
What this guide tests
What do the two observations establish under their own goals, and why is the draft's unchanged-goal comparison unsupported?
See the specimen and fields · Work through the fictional records
Task and source context
A progress entry can use a different condition or measure than the baseline. APTA publishes documentation guidance across the care episode. A therapist can create a fictional or tabletop example that tests whether a model preserves measurement conditions and goal versions. APTA documentation guidance
Use a dummy object-placement task and an original worksheet. Record one count with a fixed time window and another with a longer window. Change the goal wording between versions. Ask which results can be compared, and have a reviewer verify the units, timing and goal independently.
Use inert props or consenting adult staff in a training room. Exclude patients, gait videos, actual speech or hearing assessments and case histories. Keep outcomes and treatment judgments outside the exercise.
Industry research
The U.S. Census Bureau’s industry definitions establish which activities these codes cover. The workflow sources in this guide describe the task or record type. Fiund’s pilot below uses fictional material to test a specific decision within that scope.
621340 · Offices of Physical, Occupational and Speech Therapists, and Audiologists
This industry comprises establishments of independent health practitioners primarily engaged in one of the following: (1) providing physical therapy services to patients who have impairments, functional limitations, disabilities, or changes in physical functions and health status resulting from injury, disease or other causes, or who require prevention, wellness or fitness services; (2) planning and administering educational, recreational, and social activities designed to help patients or individuals with disabilities regain physical…
Build a pilot sample
Set up this pilot
Create fictitious therapy goals with revised observation conditions. Grade each count against its own goal and require a common window before comparing.
Prepare the test packets
- Start with the worked case. Use the fictional record types below as a guide to the source files you need. Give each packet a stable ID and save its source versions with it.
- Change one fact. Make one packet with enough evidence to answer the question and one where the evidence points to a different answer. Keep the other fields fixed so a reviewer can tell what changed.
- Remove the decisive evidence. Give the reviewer a third packet with a plausible near-match in its place. Have them write what can still be established and what evidence would resolve the rest.
Fields the reviewer needs
- Goal ID and approved wording
- Defines the target.
- Observation count, window and condition
- Preserves measurement basis.
- Goal-linked result and missing common window
- Limits interpretation.
Use What health care and social services teams can contribute to AI to turn the specimen into a brief with a defined task, source and review owner.
Fictional exercise
A longer task window and a revised goal need separate comparison fields
An original tabletop object-placement task uses inert props and artificial counts. It contains no patient, gait, speech or hearing assessment, treatment or outcome claim.
| Record ID | Source | What it records |
|---|---|---|
| TG1 | Goal G1 baseline | Original goal: place 5 objects within 30 seconds. Baseline B1: 4 objects in 30 seconds, condition C. |
| TG2 | Goal G2 revision | Approved goal changes to 8 objects within 60 seconds. Follow-up F1: 7 objects in 60 seconds, same stated condition C. |
| TG3 | Summary draft | Draft says count improved by 3 under the same goal and that revised target was met. No common-window follow-up supplied. |
Question for the model. What do the two observations establish under their own goals, and why is the draft's unchanged-goal comparison unsupported?
Show the reference answer
TG1 documents four objects against G1's five-in-30-seconds target. TG2 documents seven against G2's eight-in-60-seconds target, so neither stated count meets its own threshold. TG3 overlooks both the doubled window and changed goal. Report the original measures and versions rather than call the raw three-object difference an unchanged-goal improvement. The fixture does not establish clinical progress or treatment effectiveness.
How to check the result
Ask for the answer, the record IDs that support it and any fact the evidence leaves unresolved. Have a reviewer compare the response against the reference before deciding whether the sample is ready to expand.
Errors to flag in this case
- Comparing raw counts across different time windows as if the task basis were unchanged.
- Scoring a revised goal using an earlier target or claiming success below the stated threshold.
Change the inputs and run it again
Add a 30-second F1 subinterval count of four. Require the common-window comparison to show equal counts while retaining the separate 60-second result.
Score the decision and the evidence separately
Save the reviewer’s expected answer, supporting record IDs and unresolved fields in a separate answer key. Score the model on the decision, its citations and its handling of the deliberate gap. A correct answer citing a different case counts as an evidence error.
Before expanding collection, have a second reviewer run the same checks on the held-out packet. Keep related versions, views and exports of the same case together so they stay on one side of the training and evaluation split.
Check permission and scope
Use original measures and consenting adult staff with inert props. Exclude actual assessments and histories; practitioners retain outcome judgments.
Use no patient data or actual care footage in a practice package. Handle real clinical data through a separate authorized governance path.
Describe an authorized archive through Fiund’s existing-data assessment, or describe a new capture capability through the collection brief. The contributor agreement covers the terms for participation.
Sources and further reading
Resources for this workflow
Includes worksheet
What health care and social services teams can contribute to AI
Separate a staff training simulation from a governed clinical or care-data project.
Includes worksheet
Build an AI evaluation sample for service handoffs
Use fictional records to test whether a handoff reached a documented state.
Company and program reviews
Use these reviews to compare sourcing, capture or evaluation approaches. Confirm the applicable program, accepted sources and terms with the provider.
Compare the approach
Protege review for AI teams and data owners
Compare the company’s healthcare data-provider pathway separately from a fictional staff workflow.
Compare the approach
Sapien review: dataset quality review and expert checks
Compare an independent quality review process for an administrative or nonclinical evaluation.
Related industry examples
Industry guide · NAICS 621310
Chiropractic visit and exam documentation data for AI
A template can carry an observation from an earlier visit into a later draft.
Industry guide · NAICS 621330
Therapy intake preference and correction evidence for AI
A counseling practice can evaluate whether an assistant preserves a person’s stated preferences and corrects a summary when new information appears.
Industry guide · NAICS 621399
Dietitian food log and serving-basis records for AI
A food log can list an amount per package while a label lists a value per serving.