LO
PilotCognitive automation2023-01-01
A Taiwan university hospital deployed and tested an AI system that suggests ICD-10-CM diagnosis codes to its certified coding specialists, who review the suggestions and choose the final codes
Medical coderoccupation page →
Event date / reported
2023-01-01 · reported 2024-09-20
Evidence stage
PilotSmall-scale trial in a real setting. Tells us the deployment conditions are being tested, not that they hold — so one pilot is never enough on its own; two independent ones are.
Tasks this bears on
Coding complex inpatient stays
Reading a long hospital record, choosing the principal diagnosis and the secondary ones, and the codes that set the case's payment group.
Being augmented≈ Platform inference
Reviewing codes that software suggested or assigned
Accepting, correcting or rejecting machine-suggested codes, and working the queue of charts the software could not code with confidence.
New task≈ Platform inference
Where this applies
A peer-reviewed study by researchers working with Kaohsiung Medical University Chung-Ho Memorial Hospital, published in the Journal of Medical Internet Research on 20 September 2024. It states that the user interface and coding system were deployed and tested in January 2023 at the hospital, inside the workflow of its certified coding specialists, who review the system's recommendations and select the final ICD-10-CM codes for Taiwan's diagnosis-related groups; the real-world evaluation used cases from February to April 2023. The month is given, not a day. The authors report that the system helped detect coding errors in 1.9% of cases (50 of 2,632) and say it has the potential to reduce manual workload; they do not measure time saved, and the agreement analysis was done by one senior coder. One hospital, a three-month evaluation, one payment system.
What this means
In at least one hospital, AI code suggestions have run inside certified coders' everyday workflow: the machine reads the record and proposes codes, and the coder decides. For a coder the job shifts from finding every code to checking the machine's — and the check caught errors the coders themselves had made.
What it does not yet show
It is one hospital's three-month evaluation in Taiwan's payment system; it does not show time saved, coding without a coder, or any change in how many coders the hospital employs.
What you can check
Open the study on PubMed Central (PMC11452756), and read the methods section on deployment ('deployed and tested in January 2023 at KMUCHH'), the description of how coders select the final codes, and the result on the 1.9% error rate.
Does it change the assessment?
No. The impact index is never moved by a single event, and this stage does not move one on its own: a "Pilot" record counts toward a judgement but needs a second, independent record before the judgement rests on evidence. This one is counted; on its own it changed nothing.
Source
Journal of Medical Internet Research (2024) — Kaohsiung Medical University Chung-Ho Memorial Hospital study of an AI-assisted ICD-10-CM coding system (PMC full text) · verified 2026-09-26 · Claude (VOLO agent) · interpreted 2026-09-26 · Claude (VOLO agent)
Primary source — published by the party that did this, or the authority of record. No co-signature needed.