FHIR Data Management Evolution: A Guide to Choosing the Optimal Exchange Format by Use Case

Details

This guide details the optimal FHIR format selection for various data exchange patterns originating from SDC-based patient intake and documentation. The core message is that choosing the correct format based on the use case is critical for maintaining data integrity. Three primary formats are discussed: 1) **Bundle transactions** are recommended for single form submission with extraction, ensuring atomicity (one HTTP POST guarantees all resources succeed or fail). This is best for clinical workflows where partial success compromises data integrity. 2) **NDJSON** is ideal for nightly batch exports used in reporting or ML training. Its streamable nature makes it easy to parse in tools like Python/Spark, though cross-resource references must be resolved downstream. 3) **Individual REST POST/PUT** is suited for progressive save-and-resume workflows, allowing partial updates. However, clients must implement version tracking (ETag/If-Match) to prevent overwriting concurrent changes. The article provides a decision matrix covering use cases like cross-clinic synchronization and ML data pipelines. It also warns against common mistakes, such as using individual POSTs for form submission (leading to race conditions) or using Bundles for bulk export (causing timeouts). This technical guide serves as a comprehensive reference for SDC integration patterns.