What the PDGM comorbidity adjustment actually is
Under the Patient-Driven Groupings Model, every 30-day payment period is classified on five variables: admission source, episode timing, clinical grouping, functional impairment level, and comorbidity adjustment. The first four are driven by the OASIS assessment and the primary diagnosis. The comorbidity adjustment is different — it is decided almost entirely by the secondary diagnoses that reach the claim.
That single fact explains most of the money agencies lose on this variable. A clinician can document three chronic conditions in a visit note, a coder can code two of them, and the claim can carry one, and nobody in the chain is necessarily doing anything they would call wrong. The adjustment is calculated from what was submitted, not from what is in the chart.
The mechanism itself is simple. CMS applies a case-mix weight to every period, and the comorbidity adjustment is one of the modifiers applied to that weight. There are three possible outcomes: no adjustment, a low adjustment, or a high adjustment. When no secondary diagnosis on the claim meets CMS criteria, the period receives nothing — which is the correct outcome for a genuinely uncomplicated patient, and a silent revenue leak for a complex one.
How CMS decides between no, low, and high
CMS publishes comorbidity subgroups — often called interaction lists — and a diagnosis has to appear in one of them to count. A secondary diagnosis that falls in a single subgroup can produce the low adjustment. Secondary diagnoses that span interacting subgroups can produce the high adjustment. Diagnoses outside the lists, however clinically serious, do not move the adjustment at all.
The lists are not static. CMS revisits the case-mix weights, functional impairment levels and comorbidity subgroups through rulemaking, and the CY 2026 final rule (CMS-1828-F) finalized recalibrated versions of all three. That has a practical consequence for anyone auditing a quarter: a list memorised during the PDGM transition in 2020 is not the list in force now, and a code that triggered the high adjustment two years ago may no longer interact the same way.
Before you audit comorbidity performance, pull the current grouping logic and case-mix files from CMS rather than working from memory or from a vendor summary. If you cannot say which subgroups were in effect for the period you are reviewing, you cannot tell a coding failure apart from a rule change.
Two diagnosis decisions, one claim
The primary diagnosis sets the clinical grouping, which is the largest single driver of the period's payment. The secondary diagnoses then decide the comorbidity shell around it. Both decisions have to be defensible on their own, and the order matters: the primary diagnosis establishes what the episode was for, and the secondary diagnoses describe what else the patient carries that the plan of care is managing.
This is also where coding integrity gets tested. The primary diagnosis should be the condition that most contributed to the need for skilled services in that period — not the highest-paying code that the record can be stretched to support. Selecting a primary diagnosis to influence the grouping, or adding secondary diagnoses the record does not support to chase an adjustment, is upcoding regardless of who signs the claim. The revenue objective is complete capture of what is documented, not maximum capture of what might be argued.
A useful discipline at review: for every secondary diagnosis on the claim, ask which note in this period documents the condition as current, and what the plan of care does about it. If there is no answer, the diagnosis does not belong on the claim, and it probably should be a query instead.
Where comorbidity capture breaks in practice
The failures repeat across agencies, and nearly all of them happen before the claim is submitted.
Documented but never coded. The visit note mentions chronic kidney disease stage 4 and heart failure with preserved ejection fraction; the coded diagnosis list stops at the primary. This is the most common loss and it is purely a workflow gap — the coders never saw the note, or the note was closed after the coding was completed.
Unspecified codes when the record supports specificity. Type 2 diabetes coded as E11.9 when the note documents hyperglycaemia, or heart failure coded without its type, collapses detail CMS needs to place the diagnosis in a subgroup. Specificity is not cosmetic — it is what makes the code eligible.
Active conditions coded as history. A condition that is being treated and monitored this period is not a history code. Downgrading an active condition to a personal history code removes it from the adjustment while looking superficially accurate.
Copy-forward diagnoses. Last period's secondary list carried into this period's claim without current support. It inflates the claim, invites denial and audit exposure, and it is the mirror image of under-capture.
Coded on the OASIS, lost on the claim. Diagnoses entered on the assessment do not automatically populate the claim's secondary list. Where the two are built by different systems or different teams, they drift, and the claim is the one that pays.
Query never sent. The coder suspects a diagnosis but the record is ambiguous, so they code conservatively and move on. A query with a 24-hour turnaround would have resolved it; an unasked query is a permanent under-capture.
What documentation has to show
A secondary diagnosis is supportable when the record names the condition, describes its current status with enough specificity to code it, and shows the plan of care responding to it. "History of CHF" supports nothing beyond a history code. "CHF with preserved ejection fraction, daily weights, furosemide increased this week, weight up 3 lb since last visit" supports the diagnosis, the specificity, and the medical necessity of the monitoring in one sentence.
That standard is teachable. Most clinician education on this topic fails because it asks for better documentation without showing what better looks like in a note the clinician already writes. Two or three annotated examples from your own charts, per discipline, do more than a slide deck about ICD-10.
Coders need the same clarity from the other direction: a query that names the condition, states what the documentation already supports, and offers clinically plausible options. "Please clarify" produces rounds of email. A focused query — is this documented as acute, chronic, or a history, and what is the current severity — gets answered.
The workflow that keeps it clean
Comorbidity capture is a sequencing problem more than a knowledge problem. Coding has to happen after the period's documentation closes and before the claim is built, and the claim has to carry the full coded diagnosis list rather than a truncated version chosen by a billing template.
On top of that base, four controls carry most of the value. First, a pre-bill diagnosis check on every period, not a sample: is the secondary list complete against the coded record, and are the codes at the specificity the documentation supports. Second, a second-level review on periods where the adjustment matters most — complex patients, long stays, readmissions. Third, a query log with an owner and a turnaround time, reviewed weekly. Fourth, a quarterly audit that samples periods across coders and compares what the record supported with what the claim carried.
The feedback loop is what makes it durable. Every audit finding should change something upstream — a documentation prompt, a query template, a coding tip sheet, a clinician education example. Agencies that treat the audit as a report rather than an input repeat the same findings next quarter.
If the internal team cannot absorb the review, that is the point at which an external coding and OASIS review function makes sense — but the standard should be the same either way: coded from the record, reviewed before the claim, and measured afterwards.
What to measure, honestly
Track the share of periods landing in each adjustment level against your own trailing baseline, not against a benchmark you cannot source. A month-over-month move you can explain is worth more than a comparison to an industry figure nobody can reproduce.
Four numbers are enough to start: the percentage of claims carrying at least one secondary diagnosis, the percentage of coded diagnoses that survive to the claim, query volume, and the share of queries that changed a code. The first tells you whether capture is happening, the second whether the handoff is working, and the last two whether the coder is resolving ambiguity or working around it.
Keep this in proportion with the rest of the model. Functional impairment, set by OASIS items such as grooming, dressing and bathing, is a larger lever than comorbidity for many agencies — see how Section GG scoring drives PDGM payment — and the primary diagnosis's clinical grouping outweighs both. Comorbidity capture is the variable where a clean process reliably converts documentation you already have into payment you are already owed.
Bottom line
The PDGM comorbidity adjustment is not a coding trick and it is not an OASIS problem. It is the last 200 metres of a long relay: the clinician documents the condition, the coder captures it at full specificity, the claim carries it, and CMS scores it against current interaction lists. Break any leg and the period is paid as if the patient were simpler than they are.
Fix the sequencing first, then the specificity, then the queries. The complete PDGM guide covers how the adjustment sits alongside the other four classification variables, and clinical documentation review is where most of the upstream fixes land.