ICD Coding (and AI Code Suggestion)
ICD coding is the process of translating a clinical encounter into standardized codes from the ICD (International Classification of Diseases): diagnosis codes — ICD-10 in most systems today, with ICD-11 adoption underway — and, in Indonesia's casemix workflow, ICD-9-CM procedure codes alongside them. These codes drive health statistics, casemix grouping, and payment: under BPJS Kesehatan (Indonesia's national health insurer), the INA-CBGs (Indonesian Case Based Groups) tariff a hospital receives is computed from them. AI can read the clinical documentation and suggest candidate codes with their supporting evidence, but a trained coder reviews and confirms every code before it enters a claim.
AI code suggestion works at the reading layer: the system parses the note or discharge summary (resume medis), proposes the diagnosis and procedure codes the text supports, points to the sentences that justify each one, and flags gaps — a suspected secondary diagnosis that was never written down, or a procedure with no matching indication. Done this way, suggestion speeds the coder up and catches omissions, without touching the two boundaries that matter: the clinician remains the author of the record, and the coder remains the decision-maker on every code.
This is how Micromeet approaches it as governed healthcare AI: Claim Readiness is built to suggest ICD codes and completeness checks from structured clinical documentation, for the casemix coder to review and confirm before submission. The quality ceiling is honest, too — AI can only suggest what the record supports, so better documentation upstream is what makes coding suggestion genuinely useful. AI writes. Doctors decide.
Micromeet — AI for governed healthcare. See the public benchmark →