This is part two of the discussion around average reimbursement analysis, where we received a spreadsheet from a billing company for one of our clients. Diving in more deeply, we referenced earlier in the last session, you should go back and listen to that one if you haven’t done it yet, that the summary table adds up. It’s a sum of averages by CPT code, and that’s completely wrong.
I am just diving in more deeply to explain why that doesn’t work. It falls into the general category of averages, and the paradox, and things like that. Bad! Don’t do it! You have to be careful about adding averages, averaging averages, and doing other goofy things when it comes to averages for a whole series of reasons.
A simple explanation of that is, let’s say, 5 encounters have CPT code A and then 5 encounters that have CPT code B. In this case, what they’re doing is they’re saying that makes 10 encounters. And in theory, that’s possible if there’s no overlap, where there’s not a single encounter that has both of those CPT codes in. However, every single encounter could bill both of those. You could think of situations where you bill two CPT codes in combination: a surgical procedure, something like that with injection or anesthesia, or the injection code with the units for the drug, or whatever. They’re always billing a combination. So if they’re billing a combination, then it’s not 5 plus 5 equals 10. It’s 5 plus 5 equals 5. There are only 5 encounters, each one of which includes those 5 CPT codes.
The reality is, we don’t know what that is. It could be 5, it could be 10, or it could be anywhere in between. As the numbers get larger, this gets much more complex, so you can’t just assume that those are additive.
Diving in more deeply than into reverse engineering of what they’re doing, we’re trying to figure out why they’re doing some of this. For example, we see on one of the line items, call it, CPT code 10001. For Humana, for example, we’ll see that they show encounter count is 5. When we go to the raw data and try to validate and check because an excellent way to QA things is to just spot-check, find, pick one individual, line item, code, calculation, or something like that, where it’s a relatively small number. You can go into the raw data, throw some filters on it, and see what pops up. When we filter for Humana’s payer for that CPT code, we don’t get 5. We get 55 encounters with that CPT code in the raw data.
We’re trying to figure out what they’re doing. We thought, “Okay! Well, when you’re calculating average reimbursement, typically, you want to look at only adjudicated claims because if you billed something yesterday, you don’t want to put that 0 in the average because that’ll bring down your average unnecessarily if you’re likely to get paid on that in a month.” We then filtered and only looked at adjudicated claims. Nope! Not that either. It turns out there are 6 claims that are adjudicated for that payer for that CPT code. What the heck are they doing?
It turns out they’re only counting claims that have gotten paid a non $0 amounts. Let’s get that, quote, average straight. You got paid, for example, $10 on one claim, and on the other, you got paid 0. You had to write it off. So you had two claims. They’re saying you averaged $10 because you got paid 10 on the one you got paid on. They do not include the zeros in the averages, which is wrong, really wrong.
It’s a little convenient for a billing company. It makes them look better because you’re only counting the ones you got paid on. Some of these payers are denying at a very high rate.
We’re guessing that they’re not doing this intentionally because we don’t think they intended to make themselves look better. Still, it might be perceived that way by a client because it just so happens. To make the billing company look a lot better.
Why, in this case, for that particular payer, for that CPT code, have only 5 out of 55 been paid? Some of these claims go back 6 months for data service and submission. Again, when we look at the second tab, referencing back to the last podcast, that’s the adjudication rate tab. An adjudication rate shows that there’s a too high rate of denials. In fact, for this particular payer, it’s saying the denial rate is 82%. Only 18% are successfully making it through. We calculate a slightly different number with the same raw data. We get 88% that are unpaid.
If 82% or something north of 80% are getting denied, and 6 months later, they’re not paid, should it be that those are written off? They should either be appealed and paid or written off by that time. Otherwise, you end up with garbage data.
The bigger issue in our mind is, you don’t know what’s going on. Is it that this is a bad payer, and you effectively should not take it because you’re only getting paid 10% of the time? Or is it that somebody just isn’t following up and has not gotten these appealed and paid for yet? It shouldn’t be out there forever. You should clean this up and figure out whether or not you’re going to get paid so that you know what to do because the data should be informing you have something to do, some action to take.
Problems With This Analysis
Going back to the main average table, we then have the subtotal line below each payer summarize encounters, payments, averages. In theory, it then shows a dollar number for the average, which should be dollars per encounter per payer. You got paid $100, on average, every time you saw a patient for this payer, or $400, or whatever it might be. Though we’ve discovered that there are problems with this analysis, it’s not dollars per encounter. It’s dollars per line item per payer, excluding anything that wasn’t paid, which means you’re only averaging things that got paid. You have $1 per line item per payer just when things got paid. I don’t know what to do with that information, like what question are you answering? When you got paid, this is what you got, but who cares about that?
In most cases, we only use adjudicated claims. And in this case, since the billing company said that they’re denied at an 82% rate for that particular payer and our data shows about 88% on unadjudicated because, again, we don’t have full transparency into this data given the fact that we only have this information provided to us. We can safely assume, given that information, that the true average is closer to including all of the claims, including the unpaid ones, not only the adjudicated ones.
That is a good billing company because we’ve worked with them before. We’re not trying to throw rocks at this billing company. They’re good, they do a good job, and our perception overall is that they’re good at reporting, in general, also.
One of the things that there’s a nuanced analysis that we might be picking up on is they’ve been very good at getting information out of systems and getting it to us to do analysis. Therefore, we like billing companies like this. However, it’s possible that while they may be good at that, which call it reporting, they may not be good at analysis, which is taking that raw data, that information, and crunching that data. In this case, a billing manager provided this report back or this analysis back to the client.
The net is, remember, we talked about line of business A and line of business B. Instead of $100 on average for a line of business A, it’s more like $92. Instead of just under $400 average per patient encounter for a line of business B, it’s more like $165. That’s a staggering difference. We’re talking about more than double or approximately one-third of the reimbursement that they showed. They’re getting on average once you include all of the denials. That misinformation would likely misguide a business in their decisions, so it’s vital to correct the data to have the analysis right.
It’s clear that this client doesn’t understand how Medicare Advantage works since they asked the billing company to bill Medicare if the Medicare Advantage plan doesn’t pay. They still need to receive good information so that the client can make data-driven decisions about their business.