Part IX: Understanding How PDMP Data is Scored
BY JIM HUIZENGA, M.D. AND CHIEF CLINICAL OFFICER OF APPRISS HEALTH
Series Introduction – Appriss Health has long held that chronic pain and addiction (substance use disorder) are separate medical conditions and that each deserve a unique clinical approach. We have spent years developing clinical decision support tools and in doing so have obtained the feedback from countless medical professionals, including many pain providers. Through those providers we have also heard (directly and indirectly) from many chronic pain patients as well.
This series will explore our approach to benefit/risk assessment, data presentation, and clinical support…and how they relate to our goal of creating a usable and balanced, clinical viewpoint that protects access to care while highlighting areas of risk that clinicians should be aware of.
Okay, let’s advance the discussion on scoring. First off, let’s agree that all of the following research based risk factors (metrics) that we’ve been talking about can be considered themselves to be scores:
- 90 MME per day
- 5 providers per year
- 4 pharmacies per 90 days
They’re all represented with numbers (we can’t get away from them in medicine) per some unit of time. If we look at 90 MME per day you might be thinking that “90” is the score; however, it is probably more accurate to look at the score associated with all of these metrics as either a “0” or “1.” In other words, if you are at 90 MME per day or above you get a “1” and if you are less, you get a “0.” Let’s represent all of this in a table as follows:
So, if you are higher than the metric (even just a little bit) you get the maximum score. If you are lower than the metric (even just a little bit) you get the minimum score. Now, there are general challenges with threshold approaches such as this, and they are:
- There is a real difference between seeing five providers in a year and seeing 35 providers in a year, but both situations get the same maximum score.
- Conversely, there is no real difference between using four pharmacies in 90 days vs. using four pharmacies in 91 days but the associated scores for each are at entirely opposite ends of the spectrum.
- We examined additional concerns around this type of approach in our multiple providers episodes blog post.
The Appriss Health Approach.
Our approach to scoring is markedly different from the approach described above. First of all, we DO NOT set any artificial thresholds. For instance, if we were to draw an analogous table to the above using our approach it might look something like this (note this is simplified quite a bit and our actual tables are much larger, sometimes with thousands of divisions).
As I said in previous posts, we have 20 different measurements that go into the calculation of a single Narcotic Score. We get to 20 because we look at:
- The number of providers seen
- The number of pharmacies visited
- The MME dispensed
- The amount of sedatives dispensed
- The number of times different providers overlap medications
And we look at these metrics in four different time periods:
- The last 60 days
- The last 180 days
- The last 365 days
- The last 2 years
If I show that using tables (you are not intended to actually read all of these tables, just get an idea for the complexity) this is what it looks like.
This is how we get to just one of our scores. Because we use different time periods, our scores can be made higher for more recent activity, and lower, for more distant activity. And because we use many different factors for just a single score, the only way to get a very high score is to be very high in all 20 areas. Let’s pause for a minute — and compare this to the single threshold approach, which we do not use, where you get a high score if you’re over just one measurement, even by a little bit.
As we continue to dig deeper, there is another interesting aspect of our scoring methodology I’d like to share with you. There are multiple ways to get to a single score. For instance, if we grouped providers, pharmacies, and overlaps into one category and we grouped MME and sedative amount in another category we can create the following table.
What this table reveals is that there are multiple ways to get to a similar score. What this means is that our scores tell providers in general how much risk is in a PDMP report, but providers must look at the PDMP report to find out where the risk is coming from (remember we created a graph to help providers see the details). A provider cannot look just at our scores and tell exactly what is going on.
We did this on purpose.
All patients are unique, and their histories, and reasons, and risks, and benefits can only be understood by reviewing their actual data. And, of course, talking with them about how they’re doing and what’s been going on.
There are an infinite number of possibilities.
- A patient with a high score may truly be in trouble and heading towards an overdose.
- Another patient with a high score may have chronic pain, lost their insurance and pain doctor and then must use emergency departments and urgent cares to obtain enough medication for pain relief and to avoid withdrawal (remember, withdrawal doesn’t mean addiction).
- A patient with a mid-range score may be entirely fine with a great benefit risk ratio.
- Another patient with the same mid-range score may be getting little benefit and may be trending towards substance use disorder (and a higher score) and needs to have a new plan developed ASAP.
And so on. And so on…the key message being that while our scores represent overall risk at a certain point in time. The specifics around the types of risk and the direction it is going in is what’s important and any risk noted by a provider should be compared with a benefit assessment to make a final treatment decision.
In our next post we’re going to discuss something called machine learning and show how it can be used to really narrow in on those patients who need help (those at risk of SUD and overdose) and help allay concern for those patients with chronic pain who get good benefit from chronic opioid therapy.
And after the next post, we’re going to wrap all of this up with some final thoughts.
Thanks for joining us through this blog post series. I hope it is informative, and for some of you, reassuring.
To read more on this series: