Standards & Interoperability
ICD-10 Coding
ICD-10 — the tenth revision of the International Classification of Diseases, maintained by the World Health Organization — is the global standard for classifying diagnoses and health conditions. Every disease, injury, symptom, and cause of death is assigned an alphanumeric code, creating a common language for morbidity statistics, public-health reporting, and (especially in the US, via ICD-10-CM) medical billing. For healthcare software and RCM automation, accurate ICD-10 coding is where clinical reality meets reimbursement — and where errors quietly cost organisations real money.
Written and reviewed by the Speed AI Labs engineering team — healthcare AI & EHR integration specialists
Last reviewed July 2026
How ICD-10 is structured
ICD-10 codes are organised into chapters by body system and condition type. A code such as E11.9 breaks down as a category (E11, type 2 diabetes mellitus) plus further characters that add specificity (here, 'without complications'). The US clinical modification, ICD-10-CM, extends this to tens of thousands of highly specific codes, capturing laterality, severity, and encounter type. This granularity supports precise reporting and billing but also makes manual coding labour-intensive and error-prone — which is exactly why it is a prime target for AI assistance.
ICD-10 in the revenue cycle
In revenue cycle management, ICD-10 diagnosis codes pair with procedure codes (CPT/HCPCS in the US, OPCS in the UK) to justify medical necessity and determine reimbursement. A claim with an unspecified or unsupported diagnosis code is a prime candidate for denial. Coders translate the clinician's documentation into the right codes, and the accuracy of that translation directly affects cash flow, audit risk, and compliance. Under-coding leaves revenue on the table; over-coding invites clawbacks and penalties.
How AI assists ICD-10 coding
Computer-assisted coding (CAC) uses clinical NLP and large language models to read documentation and suggest ICD-10 codes, with a human coder reviewing and confirming. Modern systems can flag missing specificity, surface documentation gaps before a claim goes out, and reconcile codes against payer rules. The goal is not to remove the coder but to make them dramatically faster and more accurate — turning hours of chart review into minutes of confirmation while keeping a human accountable for the final code.
ICD-10 vs SNOMED CT, and the road to ICD-11
ICD-10 classifies for reporting and billing; SNOMED CT captures granular clinical meaning at the point of care. The two are mapped, and well-designed systems store rich SNOMED CT data and derive ICD-10 from it. The WHO has released ICD-11, which is digital-native and more detailed, and adoption is gradually beginning worldwide — so any long-lived coding system should be designed to accommodate a future migration rather than hard-coding ICD-10 assumptions everywhere.
Frequently asked questions
What is the difference between ICD-10 and ICD-10-CM?
ICD-10 is the WHO's international version used mainly for mortality and morbidity statistics. ICD-10-CM is the United States' clinical modification, with far more codes and greater specificity, used for diagnosis coding in billing and clinical documentation.
Can AI fully automate ICD-10 coding?
Not safely on its own. AI excels at suggesting codes and catching gaps, but accuracy, compliance, and accountability require a human coder to review and confirm — particularly for complex cases. The proven model is AI-assisted coding with human sign-off.
Is ICD-11 replacing ICD-10?
Eventually. ICD-11 is officially released and adoption has started in some countries, but ICD-10 (and ICD-10-CM) remains dominant in clinical billing and reporting today. Systems should be built to migrate when the time comes.
How does ICD-10 use differ between the NHS and US providers?
The purpose differs, and that changes what your software optimises for. US providers use ICD-10-CM to justify medical necessity on claims, so specificity drives reimbursement directly. NHS organisations use ICD-10 alongside OPCS-4 procedure codes primarily for Secondary Uses Service reporting and national tariff payment, where clinical coders work from the completed case notes after discharge rather than at the point of care. A product serving both markets cannot assume the US claim-centric workflow.
Losing revenue to coding errors and denials? We build AI-assisted coding and RCM automation with human-in-the-loop accuracy. Book a discovery call.