Comorbidity adjustment is one of the more overlooked levers in PDGM reimbursement, mostly because it depends on something that feels like a formality: making sure every relevant secondary diagnosis actually makes it onto the chart. Miss that, and the adjustment an agency should be getting simply doesn't happen.

Here's where this consistently goes wrong.

How the adjustment actually works

PDGM applies a comorbidity adjustment based on the presence of certain secondary diagnoses, either individually or in specific interacting combinations, that add real complexity and cost to a patient's care. A patient can qualify for a low comorbidity adjustment with one qualifying secondary diagnosis, or a high adjustment when two specific diagnoses interact in a way CMS has identified as meaningfully more complex than either condition alone.

The entire adjustment depends on those secondary diagnoses actually being coded, not just present in the patient's history somewhere in the chart.

Where agencies consistently miss it

The most common miss is straightforward under-documentation. A patient has a real secondary condition, it's mentioned in a physician's note or a med list, but it never gets pulled through to an actual coded diagnosis on the OASIS or the claim. If nobody's specifically checking for it, a real, billable comorbidity adjustment goes uncaptured, not because anyone did anything wrong, just because nobody connected the dot from "mentioned in the chart" to "coded as a diagnosis."

The second miss is more subtle: interaction adjustments. A coder might correctly code two secondary diagnoses individually without realizing that specific pair, together, qualifies for a higher adjustment tier than either one alone. That's not a documentation gap, it's a knowledge gap, and it's an easy one to have without a system flagging the interaction.

The third is specificity, which compounds the first two. An unspecified version of a qualifying diagnosis may not trigger the same adjustment a more specific code would. A close-enough code can look complete while quietly forfeiting the exact adjustment that specificity was designed to capture.

Why this adds up more than it seems

A missed comorbidity adjustment on one chart is a modest dollar difference. But this isn't usually a one-chart problem. If an agency's documentation habits or coding review process consistently miss secondary diagnosis capture, it's happening across every chart with that same pattern, quietly, all year.

What actually catches this

This is exactly where a clinical AI model earns its place in the process. A referral packet can run hundreds of pages, and a human reading it alone, under time pressure, is going to miss things simply because of volume, not skill. Olli's clinical AI is built to read through that entire packet and surface every relevant secondary diagnosis it finds, so a certified coder isn't starting from a blank page or relying on memory to catch a condition buried on page 140. The coder still makes the final call on every code, but they're making it with a complete picture in front of them instead of whatever they had time to skim.

That combination, AI doing the exhaustive first pass and a certified coder applying judgment to what it finds, is consistently more accurate than either one working alone. A human coder working solo, however skilled, can't read every page of every referral with equal attention across a full chart volume. An AI working without a coder can surface a diagnosis and still get the clinical context wrong. Together, they catch more than either does by itself.

See what this looks like on your own charts

We'd like to walk through how our coding process is built to catch comorbidity adjustments agencies typically miss.