Uma abordagem para indexação de dados no padrão FHIR em banco de dados NoSQL Colunares

Abstract

Apache Cassandra’s scalability makes it an attractive option for healthcare ecosystems; however, its primary key-restricted query model hinders flexibility with FHIR-defined search operations, such as range queries, wildcard text searches, and the combination of multiple attributes. To overcome this fundamental limitation, this work presents and validates CLiF-Arch (Cassandra Lucene-indexed FHIR Architecture), a database architecture that enables advanced queries directly within Cassandra. The architecture implements a custom secondary index that integrates the HAPI FHIR library for semantic analysis of resources with the Apache Lucene search engine for parameter indexing. Experiments with 10 million records demonstrated that CLiF-Arch supports FHIR-required searches directly via CQL. This flexibility, however, comes at a performance cost: the synchronous analysis and indexing process increased the average insertion latency from approximately 1 ms to the 3-4 ms range. The results demonstrate that CLiF-Arch is a functional solution that establishes an explicit compromise between write latency and query flexibility—an indispensable requirement for adopting Cassandra in environments that demand full compliance with the FHIR standard.