A false assumption is made that hospitals of the future are all online and fully automated. This assumption pervades the big data quants. Big data indexing is not as easy as plugging into an EMR and indexing everything in a cloud. First of all, even with a link from the search hits to the original EMR source, all of the data is not present, it may reside in an ECM system used for scanning and workflow. It may still be 20% on paper. OCR fails at 99% accuracy, Big Data fails at 99.99% accuracy. It has to be flawless to work, that's the issue.
Back to the quality issue: the information motivation is split between data integrity within tables and integration, and with applications, many of which serve the requirements of ICD-9 not the patient's continuity of care. The way access services creates a medical record number and metadata is fine. The problem is that with multiple visits new records are added, some are incorrect. If the procedure is correct and that account number is correct, send it to revenue, let's get paid. Hospitals have deadlines for submitting bills. This forces them to be constantly one step behind the curve for automating their processes and cleaning up their data. They don't have the time.