Have you ever wondered how your facility compares to others in the state? Are you among the best? Middle of the pack? Bringing up the rear? How can you know?
There is a convenient technique that we can use to group similar facilities together based on attributes we pick. The official name is "unsupervised clustering", but it's actually pretty simple. For example say we want to classify all SNFs based on the percent of Medicaid caseload on therapy and average ADL score. The data looks like this: (Illustration purposes only, there is no relationship.)
We want to divide these points into two logical groups. There isn't any obvious way to do that but unsupervised clustering will do the hard work for us. In a nutshell, when we cluster these observations we will:
- Randomly pick two points (or more if you want more groups)
- Assign each facility to the closest cluster point from step 1
- Calculate the center point for each cluster point from step 2
- Move the cluster points to the center point from step 3
- Do steps 2 through 5 until the center points stop moving.
(Okay - It's *slightly* more complex than that since we normalize the data first. Normalization is fun so that doesn't count as work.)
You can see that post-clustering we have this:
The grouping may not be what we expected, but the most similar items are now grouped together. (Most similar based on the variables we chose.)
Now we can scale this up using more variables. I chose to use:
- Average facility ADL score
- Percentage of Medicaid patients in rehab
- Percentage of Medicaid patients in reduced physical function
- Medicaid PPD
- Medicaid case-mix
Since I've chosen more than three variables, this is difficult to visualize but we use the same process as before. I chose 5 clusters for this experiment. I don't get to pick ANY attributes for the clusters. They self-group based on the data.
After the grouping I named the groups based on the averages I calculated for each:
- High Flyers: These facilities have high ADL scores, a high percentage of people in rehab, a very low percentage of people in reduced physical function and the highest average Medicaid PPD of $184.71. There are 61 facilities in this group. Successful chains like Peak Resources, Autumn Care and Brian Center appear in this group.
- Rehab: This group is doing a LOT of rehab, more than 51% of Medicaid patients are on rehab on average. ADL scores average second lowest in this group. There are more reduced physical function patients than the High Flyers, but not by much. These 83 facilities are the second highest reimbursed. There are lots of Brian Center and Autumn Care homes on this list as well. Some larger facilities you might recognize are Healthtique in Winston-Salem and Triad Care.
- ADL - This group has the second highest ADL score average of our 5 groups. They do slightly less than average rehab and have about average reduced physical function. There are 110 facilities in this group. They get paid slightly above average by virtue of good ADL scores. This group is an example of what you can do if you focus hard on ADL documentation but aren't as aggressive with therapy. Notable names: Some Autumn Care, some Golden Living, Liberty Commons, Pruitthealth and White Oak.
- Auto-Pilot - This group is characterized by the lowest ADL scores by far. They do have some therapy, but less than average. Reduced physical function is high and pay is less than average. There are 85 facilities in this group. Notable names: Golden Living, Brunswick Cove, Carver Living Center, Randolph Health and Rehab Center and Alamance Health Care Center.
- Reduced Physical Function - This group has very high reduced physical function and almost no rehab. The ADL scores are somewhat low. As you would expect, payment is the lowest for these 50 facilities. Fortunately, many of these facilities have very low Medicaid so they probably don't focus on it. There are, however a handful of large facilities on this list sadly.
If you are curious which group your facility landed in, contact me. (or if you're curious about any facility in the state for that matter.)
As I mentioned before. this is difficult to visualize, but I did give it a go below. I plotted average ADL score versus percentage in rehab versus Medicaid PPD. Hold the Ctrl key and use the mouse wheel to zoom. Hover over a point for detail. Drag the image to rotate. Size of dot indicates the number of Medicaid residents.
You can see that the rehab group (black) can get into the highest pay area, but it's not always a sure thing. The group with high average ADLs and a lot of therapy typically gets the highest PPD. (blue) The reduced physical function group (green) is paid almost $27 per Medicaid patient per day less than the high flyers.