Diagnostic coding in clinical records: ICD-10, ICD-11 and DSM-5-TR

The order of operations gets reversed constantly. Clinicians talk as though the code were the diagnosis, and as though picking the code were the clinical act. The World Health Organization’s own ICD-11 Reference Guide puts it the other way round: “The health care practitioner responsible for the patient’s treatment is also responsible for documenting the patient’s health conditions… A properly completed medical record is essential for good patient management. It is also an essential prerequisite to the creation of a valid coded record of patient diagnoses, derived through a coding process from written information describing a patient’s medical condition.”

The written formulation comes first. The code is derived from it — sometimes by you, sometimes by a trained coder. This is a documentation guide, not a diagnostic one: no criteria, no decision trees, no code lists. Record rules vary by country and by state, so verify what binds you where you practise.

There are no DSM codes

The American Psychiatric Association says this in as many words on its DSM FAQ: “There are no DSM codes.” The numbers printed beside each DSM-5-TR disorder are ICD codes. The APA frames the two works as “companion publications” — DSM-5-TR carries the criteria and the descriptive text, while “the ICD contains the code numbers used in DSM-5-TR and all of medicine, needed for insurance reimbursement and for monitoring of morbidity and mortality statistics by national and international health agencies.”

Which ICD, specifically, matters more than most clinicians realise. The APA’s coding fact sheet is precise: since 1 October 2015 the official coding system in the United States has been ICD-10-CM, a clinical modification of WHO’s ICD-10 maintained by the National Center for Health Statistics at the CDC, and it “provides the only permissible diagnostic codes for mental disorders for clinical use in the United States.” DSM-5-TR prints ICD-10-CM codes for exactly that reason.

So ICD-10-CM is a national modification, not the international classification, and it is not what other countries report on. If you practise outside the US, or you see clients across borders, do not assume the code in your DSM is the code your health system collects. In Mexico, for example, the federal record standard requires diagnoses to be recorded but names no classification at all — the coding duty comes from separate norms, as the guide to Mexico’s clinical-record standard sets out.

ICD-11 exists, and it is not the same map

WHO reports that ICD-11 “was adopted by the 72nd World Health Assembly in 2019 and came into effect on 1st January 2022,” and that as of May 2024 — two years after it took effect — 132 Member States and areas were at various phases of implementation: 72 had begun the process including translation, 50 were running or expanding pilots, and 14 had begun to collect or report data using ICD-11 coding.

That is a global count, not a timetable for any one country. We are not going to tell you when your jurisdiction switches; ask your own regulator or payer, because the answer is administrative, not clinical.

What did change is the mental-health chapter itself, and it changed enough that a mapping is not a rename. Summarised at high altitude from Reed and colleagues in World Psychiatry (2019), written by the WHO team that led the revision: personality disorders are now rated by dimensional severity plus trait domains rather than by type — one of the dimensional approaches the authors say “eliminate artificial comorbidity”; complex post-traumatic stress disorder and gaming disorder appear as distinct categories; bodily distress disorder replaces the ICD-10 somatoform disorders; and gender incongruence moved out of the mental disorders chapter into a new chapter on sexual health, “meaning that a transgender identity is no longer to be considered a mental disorder.”

Where to get each classification legally

Search results for free copies of either manual are dominated by unauthorised uploads. You do not need them, and in one case you do not need to pay either.

ICD-11 is free, online and openly licensed. WHO hosts the browser and coding tool at icd.who.int, in ten languages including Spanish. The licence is Creative Commons Attribution-NoDerivs 3.0 IGO: you may consult it and cite codes and titles with attribution, but “you are NOT permitted to make ‘adaptations’ of the Classifications,” and the licence bars using the material to develop a competing standard. Consult and cite freely; do not publish your own reordered code list.

The clinical manual is a separate publication. WHO’s Clinical descriptions and diagnostic requirements (CDDR) for ICD-11 mental, behavioural and neurodevelopmental disorders was published on 8 March 2024, runs 852 pages, and is described as the clinical version of ICD-11, complementary to the statistical linearization. Alongside English, the WHO publication page lists, as of September 2026, exactly one language version: Italian. That is what the page offers, not proof that no other translation exists or is coming.

DSM-5-TR is a commercial copyrighted work. The APA’s own materials carry the notice “© 2022 American Psychiatric Association” and its DSM page routes you to buy the book. Never paste criteria sets into your note templates and never load a pirated PDF into a practice system; the reasoning generalises to every instrument you use, which is the subject of the guide on what psychological tests you can legally use.

What actually goes in the record

A dated, signed, attributed formulation. Every clinical-record rule worth reading asks for the same four things on an entry: what happened, when, who wrote it, and their signature. A diagnosis without a date is not auditable, because nobody can tell when you knew what. The first formulation usually takes shape at the intake interview.

What supports it. APA Ethical Standard 9.01(b) is the floor: psychologists provide opinions about the psychological characteristics of individuals “only after they have conducted an examination of the individuals adequate to support their statements or conclusions,” and where that is not practical they must document the efforts made, clarify the impact of the limited information, and limit their conclusions accordingly. In the chart that is one line: what you examined, with what, and what you found. A score is not that line — a PHQ-9 result is a test result you record, not a diagnosis you assert.

Main condition and other conditions. The ICD-11 Reference Guide defines the main condition for hospital episodes: the one determined, at the end of the episode, to be the reason for admission. The principle carries to an outpatient chart — name the condition that drove the episode, and list separately all other conditions dealt with that affected the patient’s management. Conditions from an earlier episode with no bearing on this one do not belong in that list.

When nothing is confirmed — the rule almost nobody quotes. WHO is explicit, and it cuts against common practice: “Rather than qualifying a diagnosis as ‘possible’, or ‘suspected’, when a diagnosis has been considered but not established, when applicable, record the symptom, abnormal finding or problem.” Write the presenting problem, not a hedged disorder label. Ruled-out conditions run the other way. Where examination settles the question and no further care is needed, the Reference Guide asks for the ruled-out condition itself as the main condition; only where you treated a symptom, or follow-up is still needed to find its cause, do you record the presenting sign or symptom instead.

“Clinical problem: three months of sleep-onset insomnia and anhedonia with occupational impairment. Formulation in progress; reassess in four sessions.”

When the diagnosis changes. You add, you do not overwrite. A new dated entry states the new formulation, what information prompted it, and from when it applies — and the earlier entry stays exactly as written. Silently editing the old note destroys the very thing the chart is for: evidence that you reasoned, and when. Many record regimes make this explicit by banning alterations or requiring an audit trail that can reconstruct earlier states; check yours. Overwriting is high on the list of clinical documentation mistakes that are impossible to explain afterwards. Session to session, the working formulation lives in the A of a SOAP note; the closing one is recorded when you document a therapy termination.

Who may record a diagnosis

This is where honest guidance stops short of a rule. Whether a given clinician may record a diagnosis, and under which classification, is set by the licensing law of the place they practise — in the United States by state licensure, in Mexico by federal health law that names psychology as a licensed profession without defining its scope of practice. There is no portable answer, and a blog that gives you one is guessing.

What is portable is the ethical floor, and it binds regardless. APA Standard 2.01(a) limits practice to the boundaries of your competence based on education, training and supervised experience. Standard 9.06 requires you to take into account the purpose of the assessment and the test-taking, situational, linguistic and cultural factors in play, and to “indicate any significant limitations” of your interpretations. Both are written into the chart, not merely believed.

Checklist

The diagnosis or clinical-problem field is filled in, not left blank.
The formulation carries a date, a name and a signature.
What supports it is visible: examination, instruments, findings.
Nothing is established? The symptom or problem is recorded, not “possible X.”
Diagnosis changed? A new dated entry says so, and the old one is untouched.
If you code, the catalogue is the one your jurisdiction actually collects.

None of that is hard once and all of it is hard across a hundred charts. gesell.ai keeps the chart structured — intake, clinical history, treatment plan and session notes — and drafts each note in SOAP, DAP, BIRP or GIRP from your session record, so the diagnostic field exists in every note with its date and its supporting detail. When the formulation changes, you write a new dated entry, the way the record rules ask. The draft is decision support: you review it, correct it and sign it, and the diagnosis stays yours.