I finally wrote a tool that generates static HTML pages with all the combinations of diagnoses and procedures that have enough representation. After optimization and setting the minimum support to 800 patients, the tool created 1200 HTML pages in less than 10 minutes.
800 patients is too high support demand, database users may wish to browse smaller groups. I've set the minimum support to 200 patients and started the generation.
It went crazy. The number of groups to generate apparently rises dramatically when lowering the minimum required group support. After about 15 hours of work the tool generated about 100 thousands pages and has more than 200 thousands to go.
Generating static HTML doesn't make any sense. This tool needs to generate pages on the fly. But I won't rewrite it again. Later, maybe. Now I need to focus on the main scope of the analysis — defining the major patient groups and creating statistical models for them.