Drafting examples: PT, OT and SLP

• Example (synthetic patient). Every patient, date and value is invented. The values are ones the product's pick-lists accept; those pick-lists are drafts pending clinical review.
• Source lines show how ArkCloud words each selection before any AI is involved. They are the same lines Build from your selections (no AI) uses.
• Illustrative drafts are written by hand to show what the checks allow and remove. They are not output captured from the model, and a real draft will word things differently.
• Reference cases are synthetic test cases committed in ArkCloud: their inputs, and what a correct draft must and must never state. They are test expectations, not model output, and are pending clinician review.

Example (synthetic patient). Physical therapy, outpatient. Visit note on 2026-10-28.

Source lines for today (no AI):

• Patient reports left knee pain 4/10, better since the last visit.
• Patient reports: stiff after sitting.
• Therapeutic exercise targeting strength and range of motion. Neuromuscular re-education targeting balance and proprioception. Gait training targeting stairs / curbs.
• AROM left knee flexion 105 degrees.
• PROM left knee extension -5 degrees.
• Strength (MMT) left knee extension 4-/5.
• Response to treatment: tolerated without reported difficulty.
• LEFS (patient-reported questionnaire) on 2026-10-28: 46 points (range 0-80). 8 points higher than baseline (38 points), recorded 2026-10-07; higher scores are better on this scale (no validated change threshold applied).

What the checks would remove if the model wrote it: "extension 5 degrees" (minus sign dropped); "right knee" anywhere (side not recorded); "minimal assistance on stairs" (level not recorded); "flexion 95 degrees" in Objective (an earlier value presented as today's). A judgement such as "LEFS improved significantly" is not something the checks test: no change threshold is applied, so that wording is yours to accept or delete.

The readiness panel would still list two gaps: no earlier value for PROM left knee extension or for the knee extension strength grade, so change cannot be stated for either.

What the therapist must still add or check

• Post-operative status and any surgeon's protocol or restrictions: there is no structured field for these, so they reach the draft only if typed or recorded as a precaution.
• Knee swelling today: knee circumference is not in the capture lists, so type it if measured.
• The clinical reasoning in Assessment - why the patient is progressing, barriers, readiness to progress - and any change to the plan.
• The PT discipline sections (Mobility, transfers & gait; Balance; Strength & range of motion; Functional movement), which the draft does not write.
• That the extension deficit kept its minus sign, and every value its side.

Example (synthetic patient). Occupational therapy, outpatient, Hand therapy specialty. Visit note on 2026-10-28.

Source lines for today (no AI):

• Patient reports right wrist pain 3/10, unchanged since the last visit.
• Splinting / orthotic management targeting fabrication and wearing schedule. Self-care / home management targeting dressing. Fine motor / dexterity training targeting pinch and precision.
• AROM right wrist extension 40 degrees.
• AROM right wrist flexion 35 degrees.
• AROM right forearm supination 60 degrees.
• Dynamometry right hand grip 22 lb.
• Dynamometry left hand grip 58 lb.
• Edema right wrist circumference 17.5 cm.
• Orthosis right wrist hand orthosis: fabricated; low temperature thermoplastic; night-time.
• Response to treatment: required modification.

With no plan of care and no plan typed, Plan is always left blank: the draft does not decide treatment. The Assessment can only restate today's findings, and the readiness panel says why it cannot say more: no accepted goals, no active plan of care, and no earlier value for each wrist and forearm measurement. A sentence putting 58 lb on the right hand, or upper body dressing at moderate assistance, would be removed.

What the therapist must still add or check

• The OT sections: Occupational profile and Daily activities (ADL/IADL) & participation, and the Hand therapy sections Edema & limb volume, Sensation and Orthotic / splint management. The draft does not write these.
• What modification was required, and why - "Required modification" alone does not say.
• Fracture management details and any weight-bearing or loading restrictions from the referring surgeon (record them as precautions if they should reach future drafts).
• Skin check, fit and donning education for the orthosis, if done: only what is recorded can appear.
• A left-side edema value if you want a side-to-side comparison; none was recorded, so the draft cannot make one.

Example (synthetic patient). Speech-language pathology, outpatient. Visit note on 2026-10-28.

Source lines for today (no AI). The trial line always carries its cueing:

• Language treatment targeting word finding.
• Confrontation naming of common objects (stimulus: "picture cards"): 9/16 correct (56%) overall. Independent: 4/10 (40%) without a cue. A further 6 cued attempts (up to phonemic cue) count toward the overall figure only, and are not evidence of independent mastery.
• Response to treatment: tolerated without reported difficulty.

The readiness panel would list no accepted goals, no active plan of care and no functional status recorded today. A sentence giving a different percentage would be removed. The checks guard numbers, not the cueing qualifier: a sentence that kept "56%" but dropped "with phonemic cues" would read as mastery. Check that every accuracy figure still says how much support it took.

What the therapist must still add or check

• What the patient or family reported - nothing was recorded for Subjective, so the draft leaves it blank.
• The SLP Communication (speech & language) section, which the draft does not write.
• Goals with a baseline and a plan, so later drafts can relate today's accuracy to them.
• Any standardised language measure: the outcome catalogue has no aphasia or naming instrument, so a score from one cannot be entered as an outcome score today.
• SLP drafting has had the least testing of the three disciplines. Review every sentence.

• Only today's records can fill Subjective and Objective; earlier values appear in Assessment with their dates.
• Sides, units, assistance levels, minus signs, denials and test results stay as recorded, or the sentence is removed.
• Assessment and Plan are left blank, with the reason, when nothing recorded supports them.
• The discipline and specialty sections of the template are always the therapist's to write.
• Signing an AI-assisted note requires the AI review attestation. See signing.