Try using OpenEMR demo data as a source

[bokov]

If the goal is having a shareable, standard corpus representative of EHR data then perhaps it's better to use one that has already been simulated and/or de-identified by someone else. That way development can focus on actually doing things with the data, and we don't have to take on responsibility for preserving valid relationships in the first case or insuring complete scrubbing of identifiers in the second case. Tony McCormick of OEMR foundation a while ago told me that OpenEMR is widely used in medical IT training programs as a reference EMR system (because of its transparency and open-sourceness presumably). For this reason I assume there are also training datasets floating around. A quick search led me to the following SO post:
http://stackoverflow.com/questions/21400565/openemr-sample-database

OpenEMR itself is here:
http://www.open-emr.org/

On a side-note, Tony McCormick was an enthusiastic supporter of the FreeTL grant proposal, and would have been a subcontractor if it had gone through. There may be some synergy between his organization and ours. Especially because of his experience in engaging developers and adopters.