OHDSI Home | Forums | Wiki | Github

CMS algorithm for planned/unplanned admissions

(Evan Sholle) #1

TL;DR: Has anyone here made any efforts at replicating the CMS algorithm for classifying admissions as planned/unplanned in the CDM?

For those outside the US or unfamiliar, the Center for Medicare/Medicaid Services (CMS) is the biggest payor in US healthcare - and they attempt to incentivize/disincentivize providers and hospitals using many carrots and many sticks.

One of the biggest sticks in their arsenal is refusing to pay for 30-day readmissions they consider to have been avoidable. There’s a pretty detailed algorithm based on a suite of diagnostic codes and procedural codes that classifies certain admissions - like an inpatient visit for a course of chemotherapy, or to have a baby - as unavoidable, so hospitals don’t get dinged for them. Any other 30-day readmission is considered preventable.

This ends up feeding a lot of predictive analytic techniques for readmission, as it becomes a convenient way of filtering out noise by getting rid of admissions that aren’t really what you’re trying to predict. It’s been implemented pretty well in SAS, but I wanted to try to take a crack at building out something that would go against the CDM and classify each admission as planned/unplanned based on the criteria laid out. I have some code that I’m pretty sure does what I want it to do, but I’m afraid I may have reinvented the wheel, or built this out in a way that, hindsight being 20/20, may not end up being easy to share with other consortium sites (it’s literally just a SQL query).

Does anyone have any experience trying to replicate this algorithm against the CDM? Is this something people are interested in? If so, can someone with more experience than me give some suggestions or thoughts on how to most effectively share this resource with the consortium?

(Mark Danese) #2

Do you have a link to the algorithm or SAS code? Just curious to look at this algorithm.

(Nigam Shah) #3

My understanding is that the algorithm changes every year. The 2013, and 2014 are available online. One of the pdf says:

The new SAS analytic package and documentation are available upon request by emailing cmsreadmissionmeasures@yale.edu