The Goal Bank Is the Draft, Not the Goal

Goal banks have a reputation problem they only half deserve. Nobody writes forty goals a week from a blank page, and nobody should: a good library supplies tested sentence structures, developmentally sequenced targets, and measurable criteria formats that took someone years to accumulate. The trouble starts when the sentence that came out of the library is the sentence that goes into the plan of care.

An unedited bank goal is generic by construction. It was written to fit any patient, which means it describes none of them. The fix is not to abandon the bank — it is to treat every bank goal as a draft that has not yet earned a signature, and to run a short, repeatable edit pass before it does. This article shows that pass, then walks one goal through it end to end.

On paper, the bar for goals looks low. Medicare requires the outpatient therapy plan of care to contain, at minimum, the diagnoses, the long-term treatment goals, and the type, amount, duration, and frequency of therapy services — and CMS guidance says those goals should be measurable and tied to the patient’s impairments. A bank goal usually clears that bar on its face: it has a verb, a percentage, and a timeframe. What it fails is the question the requirement exists to answer — does this document skilled, medically necessary care for this specific person?

Reviewers read goals across a chart, and often across many charts from the same clinic. The pattern that gives an unedited bank away is uniformity: the same 80% criterion on every goal, the same “minimal cues” with no definition, the same “4 of 5 sessions” cadence regardless of visit frequency, and no detail that could only belong to this patient. Each goal is defensible alone. Together they read as a library, not a caseload.

A bank is a structure library, not a content library. It is genuinely good at the parts of goal writing that are the same for everyone: observable verbs, criterion formats that can be scored, sensible target hierarchies within a domain, and phrasing that has survived other clinicians’ review. Keeping that consistency across a practice is worth real money and real time, and it is why banning the bank is the wrong lesson.

What the bank cannot know is everything that makes a goal defensible: this patient’s baseline, the environments where the skill has to work, the supports actually available at home or school, and what the patient and family named as worth working toward. ASHA’s documentation guidance points goals at measurable, functional change grounded in the patient’s own participation, and AOTA builds its documentation around the occupational profile for the same reason — the clinical content of a goal comes from the person, not the library. The bank contributes the sentence; the chart has to contribute everything inside it.

Run the edits in this order. The first two are where generic goals hide; the last three make the goal scoreable and honest. With the chart open, the whole pass takes a minute or two per goal — much less than writing from scratch, which is the point of keeping the bank.

1. 01

   Swap the activity for one from their life

   Replace the library’s generic task with an activity that appears somewhere in this patient’s chart — the evaluation, the intake, the family interview. If you cannot name one, the gap is in the evaluation, not the goal, and that is worth fixing first.

2. 02

   Rewrite the conditions from their environments

   State where the skill has to work and with what real supports: which communication partners, which materials, what noise or time pressure, which device. “In the clinic room with the clinician” is a default, not a condition — keep it only when it is genuinely the target context.

3. 03

   Derive the criterion from the baseline

   The bank’s 80% is a placeholder. Replace it with a target justified by documented current performance and your prognosis for this episode of care. If you do not have a baseline for the measure you picked, measure before you set the number.

4. 04

   Define the opportunity so two clinicians score alike

   Say what counts as a trial: elicited or spontaneous, which cue levels are allowed before it no longer counts, how many opportunities make a session scoreable. If “minimal cues” could mean different things to two colleagues, define it or drop it.

5. 05

   Set the horizon from the plan of care

   Tie the timeframe to the certification period, the expected duration of this episode, and the visit frequency actually planned — not the bank’s default. A goal that needs three sessions a week to be plausible does not belong in a once-a-week plan.

The pass is discipline-agnostic because the failure is. A bank dressing goal — “Patient will don upper-body garment with minimal assistance in 4 of 5 trials” — has the same problems as the /r/ goal: a garment that belongs to no one, an undefined assist level, a criterion with no baseline behind it. The occupational profile supplies the fixes: the garment becomes the buttoned work uniform the patient must manage before a 7 a.m. shift, “minimal assistance” becomes “setup and verbal cueing for buttons only, no physical assist,” the criterion moves from a documented moderate-assist baseline toward what the discharge environment actually requires, and the horizon comes from the plan of care. Same five edits, different domain.

If the goal survives the pass unchanged — same activity, same criterion, same conditions as the library version — that is not automatically wrong. Occasionally the generic target really is the right one. The pass exists so that when the sentence stays generic, it stayed that way because you checked, and the chart around it shows the check.

The pass gets cheaper if the bank is built for it. Store frames rather than finished sentences: “Within [horizon], [name] will [functional action from the chart] during [their context] with [defined support] at [baseline-justified criterion], measured by [method].” A frame cannot be signed by accident, which is the property you want. Alongside each frame, keep a criterion menu — the measures that work for that goal family and when each applies — instead of one default number.