The Weakest Component Picks Your PT Evaluation Code

Since 2017, every physical therapy initial evaluation has billed at one of three complexity levels — 97161, 97162, or 97163 — and most clinics picked a default within a month of the codes arriving and have coded on autopilot ever since. The autopilot habit survives because Medicare pays the three levels identically, so the claim system never pushes back. But the code you report is a representation of the patient in front of you, and each level carries a four-part descriptor the record has to satisfy. This article maps the decision the way a reviewer reads it: component by component — history, examination, clinical presentation, and clinical decision-making — with the rule that actually picks the code, a worked example, and the pre-signature check that keeps the level defensible.

Each descriptor ends with a face-to-face time — 20 minutes for 97161, 30 for 97162, 45 for 97163 — and that line does more damage to accurate coding than any other. The times are labeled typical: they describe what an evaluation at that level usually takes, and they are not a selection criterion. The evaluation codes are untimed, billing as one unit for the date of service however long the visit ran, and their minutes stay out of the timed-code math entirely. A 50-minute evaluation of a healthy runner with one uncomplicated problem is still a low-complexity evaluation, and an experienced clinician who moves a genuinely complex patient through in half an hour has not billed anything down by being fast.

What actually selects the level is a set of four components stated in the descriptor itself, and the structure of the test matters as much as its content: a level is reportable only when the documentation supports every component at that level. There is no averaging, no two-out-of-three, and no credit for the most impressive finding in the note. That single fact — all components, not most — is what the rest of this article is built around.

Each component asks a question a reviewer can answer from the note alone. Before the comparison table means anything, be precise about what is being counted in each one.

1. 01

   History: how many personal factors and comorbidities impact the plan of care?

   The count is not of diagnoses on the intake form — it is of personal factors and comorbidities that impact the plan of care. A well-controlled condition with no bearing on your plan does not move the level, however serious it sounds. The converse is the documentation trap: a comorbidity you genuinely planned around only counts if the note says how it changes the plan. “History of type 2 diabetes” is a list entry; “diabetic peripheral neuropathy limits weight-bearing progression and requires skin checks with new orthotic use” is a countable factor.

2. 02

   Examination: how many elements did you address with standardized tests and measures?

   The elements are drawn from body structures and functions, activity limitations, and participation restrictions — the ICF categories — and the descriptor asks that they be addressed using standardized tests and measures. Range of motion and strength testing, a gait speed measure, a balance instrument, and a participation questionnaire are countable elements; a passing narrative observation is weaker support. Count what the objective section actually shows.

3. 03

   Clinical presentation: is it stable, evolving, or unstable?

   This is the descriptor language most worth quoting in your assessment: stable and/or uncomplicated for low, evolving with changing characteristics for moderate, unstable and unpredictable for high. A chronic condition can be stable; a six-week post-operative knee is evolving almost by definition; a presentation with rapidly shifting symptoms or comorbidity interactions that change visit to visit is unstable. Characterize it in words — the reviewer should not have to infer it from the vital signs.

4. 04

   Clinical decision-making: how complex is the plan you built, and can the note show it?

   Low, moderate, or high — demonstrated, per the descriptors, using a standardized patient assessment instrument and/or measurable assessment of functional outcome. In practice this is the assessment section doing real synthesis: the interaction of findings, the options you weighed, the prognosis reasoning behind frequency and duration. A plan any colleague would have written from the referral line alone is low-complexity decision-making no matter how sick the patient is on paper.

Read the table by row, not by column: score each component honestly against the note you are about to sign, then let the lowest row set the code. The time row is included only because the descriptors include it — it describes, it does not decide.

The rule is easiest to trust once you watch it run. Here is a composite outpatient evaluation — the kind that tempts a clinic toward 97163 on gut feel — scored component by component instead.

A reviewer scoring your evaluation does not interview you; they count what the note shows. The difference between a level that survives review and one that gets recoded is rarely the care delivered — it is whether each component is stated in countable form. Four sentences do most of the work: one naming each plan-affecting factor and how it changes the plan, one listing the elements examined and the measure used for each, one characterizing the presentation in the descriptor’s own vocabulary, and an assessment paragraph that shows the reasoning rather than announcing a conclusion.

When CMS finalized the tiered codes for 2017, it declined to tier the payment: all three evaluation levels were finalized at the same value the old single code 97001 carried, and Medicare has not separated them since. That design cuts both ways. There is no revenue reason to overcode — a 97163 pays what a 97161 pays under Medicare — and no revenue excuse for defaulting to the middle either. What the level does carry is compliance weight: it is a coded clinical representation of the patient, contractors can profile a clinic whose evaluations cluster implausibly at one level, and an audit reads the components against the note exactly the way this article has. Commercial payers set their own fee schedules and their own review behavior, so verify how yours treat the three levels rather than assuming Medicare’s parity.

Payment parity also explains the most common quiet failure: the practice that bills 97162 for everything because “they all pay the same anyway.” Under Medicare the money is the same; the representation is not. A caseload of genuinely uncomplicated post-operative referrals should produce mostly low-complexity evaluations, and a chart that says low while the claim says moderate is an inconsistency you handed the reviewer — the one kind of error the parity design gives you no benefit for making.

The fourth code in the family, 97164, replaced 97002 and has a two-part descriptor of its own: an examination including a review of history and the use of standardized tests and measures, and a revised plan of care informed by a standardized patient assessment instrument or a measurable assessment of functional outcome. The operative word is revised. Routine reassessment — re-running your outcome measures, updating goal progress, adjusting exercises within the existing plan — is part of ongoing treatment and of the progress reporting your payer already expects inside treatment visits, not a separately billable re-evaluation. 97164 is the code for the visit where something changed enough that the plan itself had to: a significant and unanticipated change in status, new clinical findings, or a patient who is not responding and needs the plan rebuilt. Payers scrutinize this code precisely because the line between reassessment and re-evaluation is easy to blur, and coverage conditions vary by payer and contractor — when you bill it, the note should show the change that forced the revision, and the revision itself.

“

“The level is not a reward for a hard day. It is a claim about the patient — and the note either backs the claim or it does not.”