From ANVISA leaflets to extended interoperability with global health databases: some pitfalls and success stories

Data interoperability in Health Information Systems (HIS) has long been recognized as a challenge, sometimes even within a single institution, given the number of different databases, systems, standards and requirements adopted. In Brazil, this is aggravated by the lack of standardized, consensual data sources in Portuguese. This paper describes a hands-on approach to overcome these hurdles, thereby helping researchers and practitioners interested in these issues. It shows how, starting from official sites of the Brazilian Health Ministry, we built HealDB – an open Portuguese language database with hundreds of thousands of instances. HealDB supports interoperability across multiple international data sources, providing a core for the construction of federated HIS in Brazil. Within this context, the paper contains two main contributions: (1) a discussion of successive approaches to derive disease ICD-10 codes from text in drug leaflets, identifying their pros and cons; and (2) a case study of linkage to RxNorm, a normalized naming system for clinical drugs maintained by the U.S. National Library of Medicine, thereby illustrating potential extensions.