In 2003, we presented an open source, web-based prostate cancer database developed to integrate clinical practice and research workflows at the IEEE Symposium on Computer- Based Medical Systems [1]. Caisis was designed around structured chronological patient histories which could be displayed to clinicians and processed by computer algorithms. We extended the system to handle 9 diseases. It has been adopted by 15 sites in 4 countries with over 400 users and is currently used to manage data for over 165,000 patients. Lessons learned include the following: 1) the original structure accommodates new diseases well, but users are quick to return to spreadsheets or denormalized data models; 2) design and support should be extended to address a variety of system-adoption scenarios; 3) perceived usability is affected by the entire set of applications an individual must use to perform their job; and 4) user training and support must be enhanced as the system evolves and becomes more complex.