Status: Active development — v5.1 (deterministic pipeline released 2026-05-06).
Owner: Mingyu Kang ([email protected])
Source data: Korean MyHealthway (
나의건강기록) app FHIR JSON export —routineCheckupDataResponse.
Output target: 5-sheet SNOMED CT–coded Excel workbook (v5 schema).
Original (Korean) workspace: parser — 나의건강기록 작업 페이지
This project converts raw FHIR JSON from Korea's national health-checkup data feed (delivered through the MOHW/KHIDI MyHealthway service and the citizen-facing 나의건강기록 mobile app) into a standards-based clinical dataset that downstream CDSS, research, and EHR-integration workloads can consume directly. The output format — known internally as v5 — is a 5-sheet Excel workbook in which every checkup measurement is bound to a verified SNOMED CT concept (with LOINC observables on the side for laboratory values), augmented by an 18-column long-format schema that captures presence_status, verification_status, and presence_qualifier_code for each finding.
The pipeline has been packaged as a Cowork plugin (myhealthway-routine-checkup-to-snomed, v0.1.0, MIT-licensed) exposing four slash commands (/convert-checkup, /batch-convert-checkup, /compare-exam-dates, /verify-v5-workbook) and five MCP tools (convert_subject, batch_convert, verify_workbook_v5, get_codes, normalize_subject). The conversion itself is fully deterministic: a Python core (v5_core.py + v5_writer.py + openpyxl) performs all parsing, unit normalization, presence-status derivation, and sheet serialization. The LLM is restricted to orchestration, validation, and user-facing reporting — it never authors workbook rows directly.
As of the date of this report, the pipeline has been applied to 78 patient workbooks stored in 대상자별_v5/ plus 84 patient files in the older 대상자별_분석/ directory. A canonical golden reference (신정순_v5_canonical.xlsx) anchors regression testing against row-count divergence.
Korean adults receive a biennial national health screening (국가건강검진) covering anthropometry (height/weight/waist/BMI), vital signs, vision, hearing, a blood panel (Hb, fasting glucose, total/HDL/LDL cholesterol, triglycerides, AST/ALT/γ-GTP, serum creatinine, eGFR), urinalysis (protein), chest X-ray, and (depending on age/sex) hepatitis-B, bone density, depression, cognitive function, geriatric ADL/fall/incontinence assessments, and the four big cancer screenings (stomach, colorectal, breast, cervical). Results are delivered to citizens through the 나의건강기록 app, which exposes them as FHIR-style JSON under the routineCheckupDataResponse and observationLaboratoryResponse payloads.
The raw JSON is patient-friendly but not analytics-ready:
"0873" for 87.3 kg, "비해당" for not-applicable, "■ 금연 필요" for lifestyle flags).suspectedDisease, existingDisease, comprehensiveOpinion) are free Korean prose.