A healthcare client had five separate IT systems across two clinic locations, including two PACS (Picture Archiving and Communication System) databases and a central hospital information system. In total, there was 14 years' worth of data, with 183,000 patient records and more than 1.5 million NEN radiological studies.
Over time, data quality errors had accumulated in the system. These errors were resulting in significant care issues such as incorrectly assigned test results and duplicate patient records, opening the door for treatment errors and a lower quality of patient care.
Our client decided to migrate all of their data into a new central PACS system. Their IT team would export the five databases, merge them, identify a potential error source, then search for affected records using custom programming.
After several attempts using this strategy, the client's internal team had only been able to migrate 5% of the data. The IT budget had anticipated a much faster rate of progress, and the go-live date for the new PACS appeared to be in jeopardy.
The data quality errors themselves originated from a variety of sources. Some, like typos or dummy names, were introduced by users. Others originated with the client, such as when prefixes were added to account numbers during a migration that split one system into two. Finally, the systems themselves also produced issues: for example, erroneous HL7 messages resulted in subsystems having different readings of the patient data.
We used our model-based approach including our tools and expertise in rule-based matching, to build a solution that automated a significant amount of the error detection and resolution process, then delivered the resulting data to the new PACS database.
Special cases that could not be handled by automation were diverted for a manual assignment of records, ensuring that the client's personnel only had to act when their specific expertise was needed.
We were able to migrate 85% of the client's data in less than a day. The data quality operations, patient matching, and migration into the new PACS database was completed in only a few days more.
