Substance use disorder ICD-10 codes: a clinician's guide

In this article▾

1. What are the substance use disorder ICD-10 codes?
2. How is a code in this block actually built?
3. Which code do you use when the note says everything?
4. Why do the unspecified codes come back to haunt you?
5. Common pitfalls
6. When can you code someone as in remission?
7. How do you code the client from the opening paragraph?
8. What to write so the code holds up
9. Where the documentation time actually goes
10. FAQ

TL;DR: The F10 to F19 codes ask you three questions: what the person used, how they used it, and what it did to them. Most of the trouble in this block comes from the middle question, because DSM-5 stopped asking it the way ICD-10 still does. Here's how to bridge that gap without guessing.

What this Blog will cover

• What are the substance use disorder ICD-10 codes?
• How is a code in this block actually built?
• Which code do you use when the note says everything?
• Why do the unspecified codes come back to haunt you?
• How do you code the client from the opening paragraph?
• Common Pitfalls

You're forty minutes into an intake. The person across from you drinks most nights, used heroin for two years and stopped eighteen months ago, smokes, and has been using cocaine on weekends since a breakup in the spring. You've a good clinical picture. Now you've to turn it into codes, and the drop-down menu doesn't care about nuance.

This is the part of the job nobody trains you for properly, and it sits right where clinical work meets the billing side of a treatment center. You learned to assess. Somewhere along the way the assessment also became a billing document, and the two use different vocabularies.

What are the substance use disorder ICD-10 codes?

They're the F10 to F19 block, one category per substance family. F10 is alcohol, F11 opioids, F12 cannabis, F13 sedatives and hypnotics, F14 cocaine, F15 other stimulants, F16 hallucinogens, F17 nicotine, F18 inhalants, and F19 other or unknown substances.

Each family is built identically, which is genuinely good news. Learn how F10 works and you can read F14 without looking anything up.

There's one exception worth knowing now so it doesn't surprise you later. F17, nicotine, has no abuse tier at all. Everything else in the block runs use, abuse, and dependence. Nicotine goes to dependence or it leaves the F block entirely.

How is a code in this block actually built?

Substance, then pattern, then complication, reading left to right. The category tells anyone reading the chart what the person used. The digit after the decimal says how. Everything after that describes what happened as a result.

The pattern digit is the one doing the real work. A .1 means abuse, a .2 means dependence, and a .9 means the pattern was never characterized (American Psychological Association Services).

Here's alcohol as the worked example:

Swap F10 for F11 and the same logic gives you the opioid codes. F11.20 is opioid dependence uncomplicated. F12.20 is cannabis dependence uncomplicated. The structure holds all the way across.

Which code do you use when the note says everything?

The most severe one, and only that one. If your assessment mentions use, abuse, and dependence for the same substance, the coding hierarchy says a single dependence code goes on the claim (American Medical Coding).

The four cases work out simply. Use plus abuse means code abuse. Abuse plus dependence means code dependence. All three mean dependence. Use plus dependence means dependence.

Here's where good clinicians create the problem without realizing it. Your history section says the client "abused alcohol through his twenties." Your assessment says "alcohol dependence, current." Both sentences are true, they describe different time periods, and a coder reading quickly may put both codes on the claim.

The history is context. The assessment is the diagnosis. Only the assessment gets coded, and one line in your note can make that unmistakable: "History of alcohol abuse in early adulthood; current presentation meets criteria for alcohol dependence."

Why do the unspecified codes come back to haunt you?

Because they pass every automated check and fail the human one. A .9 code clears the clearinghouse without a murmur, sits quietly on the claim, and then gets read by someone deciding whether the level of care you recommended was justified.

Think about what F10.90 communicates. It says the person drinks and nobody characterized how. There's nothing in that code to support a residential bed or an intensive outpatient slot, which is how a medical necessity denial starts.

Now compare F10.231, which says the person was dependent and withdrawing with delirium. That code makes the argument by itself, and you don't have to be in the room to defend it.

The practical consequence lands on you personally, weeks later, as a records request or a prior authorization call about a client you've partly forgotten. Specificity at the time of the session is the cheapest version of that conversation.

Common pitfalls

Five patterns account for most of the trouble, and none of them are knowledge gaps.

Coding at intake and never revisiting. The diagnosis you assign in session one describes someone you had known for forty minutes. By session six you know considerably more. If the code hasn't moved, the chart says your understanding didn't either.

Inheriting a code you didn't evaluate. A client transfers to your caseload with F10.20 already on the chart, and it carries forward silently through every subsequent note. Read the original assessment before you adopt someone else's diagnosis.

Coding what the drug screen found. A positive result is a data point. A disorder requires impairment, and coding from toxicology alone puts something on a person's permanent record that their behavior may not support.

Documenting DSM-5 severity and stopping there. Writing "moderate alcohol use disorder" is clinically precise and gives a coder no pattern to work with.

Being vague to be kind. Softening a diagnosis to protect a client is an understandable instinct with real consequences. A vague code can cost them the level of care you were trying to get them.

When can you code someone as in remission?

When you write it down. The ICD-10-CM guidelines are explicit that remission codes cannot be assigned without clear documentation in the record by the provider, and the record has to specify that the disorder is in remission (ACDIS).

That rule protects you more than it constrains you. Remission is a clinical judgment about a whole person, and the code set says nobody downstream is allowed to make it on your behalf from a run of clean screens.

Remission also differs from a history-of code. History-of describes something no longer treated or monitored, which does not fit anyone currently sitting in your office (ACDIS). If you are treating it, it is current.

One update worth having: a set of unspecified remission codes took effect October 1, 2022, giving you F10.91, F11.91, F12.91, and their siblings for cases where remission is documented but the original pattern was not (ACDIS).

How do you code the client from the opening paragraph?

Four codes, one per substance, each at the severity you actually assessed.

Every code earns its place. The alcohol code supports the level of care. The opioid remission code tells the next clinician something important. The cocaine code adds clinical complexity. The nicotine code opens the door to cessation counseling you can bill for later.

What you should not do is reach for F19. That code is for substances nobody can identify, and using it as shorthand for "several" throws away everything you just spent forty minutes learning.

What to write so the code holds up

You don't need to write more. You need one sentence naming the pattern in the vocabulary the code set uses.

Weak: "Client reports heavy drinking and past opioid use."

Better: "Alcohol dependence, active, meets six DSM-5 criteria. Opioid use disorder, sustained remission for eighteen months, confirmed by history and collateral. Cocaine abuse, weekend pattern, no dependence criteria met."

The second version takes thirty extra seconds and answers every question anyone will ask about that chart for the next two years.

Where the documentation time actually goes

None of this is difficult. It's one more thing on a list that already runs long, at the end of a day where you saw seven people.

Documentation burden is now well enough established as a driver of clinician burnout that the U.S. Surgeon General issued an advisory on it in May 2022, and reduction efforts have their own national initiative (Applied Clinical Informatics, 2024). The problem is not that clinicians do not know what to write. It is that writing it costs time nobody has.

Supa runs quietly inside the systems you already use. It notices when your note describes a pattern the diagnosis code doesn't match, when a code has carried forward unchanged across months of sessions, and when withdrawal shows up in a flowsheet but never reaches an assessment. It doesn't diagnose. It catches the gap between what you concluded and what the record says you concluded, while the chart is still open and fixing it takes seconds.

If your team spends its evenings on notes and its Fridays on records requests, that is the loop worth closing. Book a demo and we will show you what it catches in your own charts.

FAQ

Q: What is the ICD-10 code for substance use disorder?

A: There's no single code. You code by substance across F10 to F19, then by pattern and complication. Alcohol dependence is F10.20, opioid dependence is F11.20, and so on through the block.

Q: My client uses five substances. Can I just use F19?

A: Not if you can name them. F19 is for substances that cannot be identified. When you know what someone is using, coding each one separately preserves the clinical picture and supports the care you're recommending.

Q: What is the difference between abuse and dependence in ICD-10?

A: Abuse takes the .1 pattern digit and dependence takes .2. Dependence is the more severe designation and outranks abuse, so when your note documents both for one substance, only the dependence code is coded.

Q: Can I code for remission if my client has been sober a year?

A: Only if you document remission explicitly. Time and clean screens support that conclusion clinically, and the code set requires the provider to state it. Without your words in the chart, the active code stands.

Q: Should I code a substance that only showed up on a drug screen?

A: Not on its own. A positive screen is evidence you can use, and a diagnosis needs impairment or a documented relationship to a condition you're treating. Coding from toxicology alone overstates what you assessed.

Q: How often should I update the diagnosis?

A: Whenever your clinical understanding changes, and at minimum before any review or authorization request. A diagnosis that never moves across months of treatment tends to read as a field nobody revisited.