For a large private hospital group, we developed an AI-based health data hub that allowed data from various internal and external systems to be brought together into a single queryable interface for analytics and BI reporting.

Using a model trained on a wide variety of health data formats, value dictionaries and sources, we were able to integrate patient records from multiple systems with automated identity discovery (which patient, which visit), automated alignment of visit history (for example, which test was ordered when, referred to in which clinical notes, alongside time-series results and consequent prescribing) and automated matching of source to destination fields for reporting purposes.

Data arrived in a variety of formats, including HL7, JSON, XML, CSV and TXT and via various protocols, including HL7/MLLP, SOAP, REST and flat-file transfer via SSH/SFTP.

Using this approach, rather than mapping the data formats, fields and protocols by hand, resulted in a drastically reduced implementation timeframe. We integrated six different systems and achieved a 96% identity-mapping accuracy with 93% field-matching accuracy within an 8-week timeframe.