Even more unfortunate still is that virtually all these installed systems do not have the performance capacity or advanced search-ability to adequately mine this growing horde of data. Administrators, under capex constraints, do not allocate resources for secondary systems which would duplicate data for mining purposes while alleviating strain on the principle production system. The no money budget problem will lead in house programmers to build these research systems on top of open source NoSQL solutions. There, the technology will prove itself.
Additionally, "NoSQL" comes in different flavors. Generally, all of them forgo the Consistency in CAP for Availability and Partition tolerance, which is fine for many use cases - just not primary medical data acquisition use cases. As the field matures programmers and system designers will learn how to make this work better to the point where one day NoSQL systems may be used as the primary data repository for medical data. However, that day has not come. For instance, Riak allows you to tweak knobs in order to favor of certain aspects of CAP theorem at different times while in production (specifically the w and dw parameters, http://blog.basho.com/2010/03/19/schema-design-in-riak---int...). But having just started working with Riak in the last month or two I would still only use it as an analytics tool exposing my medical record data to m/r jobs at this point. And before jbellis smacks me, I think Cassandra is awesome and I'm looking forward to spending some time with it but I'm still not putting my med app data in Casssandra just yet as a primary data store.
/Disclaimer. I work for a major University Medical Center and write business web applications in this area./