A reproducible analytic data warehouse for Indonesian National Health Insurance (JKN) claims data.
This repository contains the SQL transformation layer that converts raw administrative claims into standardized analytic datasets suitable for health services research, audit, and policy monitoring.
Administrative claims data are not directly analyzable. They are event-based billing records, not clinical episodes.
The JKN Claims Warehouse restructures claims into research-ready constructs:
- patient care episodes
- revisit intervals
- referral patterns
- provider fragmentation
- abnormal utilization signals
This enables population-level analysis without accessing medical records.
The system is part of a larger research platform:
R pipeline (jkn-data-platform) → BigQuery warehouse → Analytic marts → Research studies
This repository implements the warehouse layer.
External data ingested from administrative claim extracts.
No transformation applied.
Data cleaning and normalization:
- de-duplication of visits
- standardization of patient identifiers
- provider normalization
- episode ordering
Construction of analytic units:
- revisit within 7 days
- inter-provider switching
- referral reset
- continuity of care
Research-ready tables:
- episode-based utilization
- provider-level indicators
- audit detection outputs
The warehouse supports multiple studies:
- obstetric claim anomaly detection
- referral fragmentation analysis
- catastrophic cost prediction
- strategic purchasing monitoring
All transformations are SQL-based and deterministic.
Given the same raw claim extract, the warehouse rebuilds identical analytic datasets.
No patient-level data is included in this repository.
The warehouse is executed via the companion repository:
jkn-data-platform
Run:
source("pipeline/run_data_platform.R")This project proposes a framework for using national insurance administrative data as a popul