Saturday 07 November 2026About PsychregWrite for Us

Business

Why Your ICD-10 Code Is a Clinical Decision, Not Just Paperwork

by Adam Mulligan · Updated

business desk

A therapist bills a 90837 with F43.20 on it, week after week, for a client who moved past “adjustment disorder, unspecified” months ago. The notes are thoughtful. The work is good. Then the payer stops paying and asks for records, and the only thing wrong in the whole chart is the code that never changed while the client did.

For a solo counsellor or a small group practice that takes insurance, that’s not a filing nuisance. That’s payroll. A denial you catch three weeks later on the ERA is cash you already spent treating someone. A clawback is worse, because you have to give it back. And the thing triggering both is usually the part of the visit clinicians think about least: the diagnosis code.

Here’s the position I’d defend to any therapist who takes insurance. The ICD-10 code isn’t the administrative tail end of a session. It’s the last clinical decision you make about it, and payers read it exactly that way.

Payers read the diagnosis code as your reasoning

The diagnosis code is the billable version of your clinical thinking. It’s the one line a payer uses to decide whether the service was medically necessary and whether the next stretch of sessions gets approved. DSM-5 tells you what the client has. ICD-10-CM turns that into the code a claim can’t move without.

When the code and the service don’t line up, the claim often comes back as CARC 11: the diagnosis is inconsistent with the procedure. That’s a payer telling you your own paperwork contradicts itself. It doesn’t mean the therapy was wrong. It means the code didn’t carry the reasoning the session actually contained.

“Unspecified” is a habit, and payers can see it

Unspecified codes aren’t off-limits. They exist for the genuine case where the picture isn’t clear yet. The problem is reflex: grabbing the unspecified option because it’s faster, when the chart would support something sharper. Run across a year of claims, that reflex reads as thin documentation, not careful care. Before a code goes out, it helps to run it up a short specificity ladder:

  1. Is the diagnosis established, or still provisional?
  2. Does the note support a severity, like mild, moderate, or severe?
  3. Is the course captured, like single episode versus recurrent?

Each “yes” pulls the code one rung off “unspecified.” Depression is the classic trap here. “Depression, unspecified” now indexes to F32.A, which the US added in 2021, while F32.9 stays reserved for major depressive disorder, single episode, unspecified. Those are two different clinical claims, and swapping them misstates what you found. If you want the full category-by-category map to check against, this breakdown of ICD-10 codes for mental health is a solid desk reference.

Specificity is continuity of care, not just revenue

The code outlives the claim. It follows the client to the next provider, anchors the treatment plan, and frames what the clinician after you assumes before the first session. A precise code hands all of that off cleanly. A vague one leaks a little information every time it’s passed along.

Comorbidity order carries the same weight. When someone presents with PTSD and a co-occurring substance use disorder, the code you list first tells the payer what’s driving this episode of care. Lead with the diagnosis that explains why this person needs this level of support right now. List the adjustment disorder ahead of the PTSD and you’ve written two different stories, one in the chart and one on the claim.

The patterns that trigger a records request

Payers don’t pull charts at random. They pull them when the data shows a pattern. F43.20 (adjustment disorder, unspecified) repeating across a dozen sessions when a subtype is clearly documentable. A diagnosis that never changes across a year of weekly work. F99 showing up past the intake. A code on the claim that doesn’t match the note underneath it. Each one reads as documentation that stopped keeping up with the client.

Two habits head off most of it. Code from the note in front of you, not from memory. And check the active code set every fall, since ICD-10-CM updates on October 1 and a retired code is an automatic denial. The official set is free from CMS, so being a year behind is a choice, not an accident. It’s also worth knowing federal parity rules generally bar insurers from applying stricter limits to mental health than to comparable medical care, which is leverage if a diagnosis-based denial looks off.

You don’t have to code like a coder

You trained to treat people, not to memorise a tabular list, and no one got into this work for the thrill of fifth-digit specificity. Coding like a clinician isn’t the same as coding like a biller, though. You need the diagnosis documented well enough that the right code is obvious, and someone checking claims before they go out so a stale code or a mismatch gets caught on the front end instead of in a denial. For a lot of small practices that second set of eyes is an in-house biller. For others it’s an outside team that does mental health billing all day and tracks the annual changes so the clinicians don’t have to. Either way, the clinical call stays yours. The cross-check is what you’re delegating.

Takeaway

Pull your last month of claims and count how many diagnosis codes end in a zero or say “unspecified.” For most practices that number is higher than they’d guess, and it’s a running total of denial risk sitting in plain sight. The specificity is almost always already in your notes. The only job left is making the code say the same thing.