Saturday, November 20, 2010

Designing a Data Model to Analyze Hospital Readmissions

So I have been tasked now for about a year and half now to support my employers initiative to reduce the number of inpatients who return to our facility for a related reason that could have been treated on a previous visit. My role in this whole process is to provide analysis support to determine 1) How many patients do return within a set period 2) are their second admissions related to their first and 3) what ancillary circumstances can we identify that might have contributed to this.

I pulled together a set of data last year in a real down and dirty fashion that allowed the staff to have an idea of who was coming back and why. However my small amount of BI experience left me with a touchy and barely sustainable SQL Analysis Cube.

In my quest to improve upon this design I've begun working on a new method of doing this, but I was shocked to find that after a whole afternoon of Google searching I found a total of ZERO examples of what other places are doing. I know this is a hot topic here in the US as several Payers are considering not paying for these readmissions in the near future and with healthcare being so much more open to sharing business practices than other businesses here I am just amazed that someone hasn't posted something yet about how they managed to solve this same problem.

So I thought after all that I'd make a post about this. I haven't yet come up with something I consider sustainable, but I am actively working on it. I figure if fortune shines upon me (and maybe you the reader) though something like this I can meet up with other DBA, DWA or BI types and maybe collaborate on something that meets multiple needs. So if you are the type interested in that respond to this post and I hope we can work together sometime to realize this project.


  1. Using CMS definition of a readmission (a patient is either an index or a readmission not both) Maine Medical Cener has created several SAS macros to study non-risk adjusted readmissions. We have created these macros to study any number of days between index discharge and readmission, counting all cases of readmissions(multiple readmissions), only the 1st readmission and can do this as an all cause readmission or same cause readmission. This information is distrubuted to dashboards and scorecards via SAS BI Platform.

    We are just begining to explore olap cubes and will be building one for readmission using SAS.

  2. Just added a new post to this discussion over her with my current E-R Diagram.

  3. Would anyone be willing to share their SAS code for readmissions? Thanks.