Analyzing Payer Problems
A recent engagement for a client was to advise them on constructing an analysis that would help them solve some of their billing problems, in particular by payer and what is being paid or not paid, and focusing in on what procedures are being paid or not paid.
This leads to a key question for us at Apache Health, which is, “Who is our audience? What are we trying to accomplish in these podcasts? How are we trying to help you? What sort of tools do we assume that you have available for you?” We are assuming that these podcasts’ users for most of what we are encountering will be relatively adept at mid-level analytical tools, whether it’s Excel, pivot tables, Power BI, or some other advanced tools for some organizations. We’re assuming that if you do not have high-level analytical resources within your organization and highly-developed tools, that you are bringing in outside resources to develop data cubes or other types of solutions to meet your needs.
We’re assuming that our audience (that’s you) is not new to to RCM analytics, where people are just getting into an analysis of their financial data for the first time and learning/discovering the basics, nor is our audience likely to be the most advanced users out there that are hardcore full-time data wranglers and data scientists, so we are sort of melding IT and functional needs to the audience somewhere in-between. This could be practice managers, accountants, a CFO or a VP or Director of Finance, or an analyst that uses multiple sources and files and is pulled in many directions in their organization.
Defining the Problem
One of the most important factors in success in analytics is making sure to clearly articulate the problem that is to be solved and then making sure you are actually solving that particular problem. “Okay, what do we need to be able to calculate to do that?” We were generating for that client an analysis by payers by procedure code. And while it may be done in our software rather than Excel, it is essentially a pivot table, and then seeing what’s getting paid and what’s not, looking at denial rates, and trying to identify the patterns to get to what’s reimbursable and what’s not? This is in part because often payer policies are not written, transparent, clear, or even always followed by the payer themselves.
A really simple example might illustrate this since most people understand E&M codes. If you’re looking at five levels of the E&M coding (yes, we know some have other numbers like three…), payers may deny at a disproportionate rate a level five or some payers even categorically deny a level 5, for example. This is fairly widely known that payers tend to deny larger dollar claims and higher coded encounters at alarming rates.
In order to perform this analysis, there needs to not only someone that is adept at analysis and pivot tables, but also someone that has a functional understanding of revenue cycle management, where that person knows to group those five codes into a particular category and identify through analysis, “Okay, this one is being denied systematically. That means that it’s likely that they have a payer policy where they don’t pay or require additional documentation or delays, even if it’s not an explicit payer policy.” Payer policies and whether they are explicit or secret is a separate little tangent into a whole conversation we’ll do on another podcast.
The Pivot Secret
Looking for de facto problems with particular procedural codes, the client had been looking for groups of denial codes and listing them in descending order. That didn’t make sense. The pivot needs to be by payer by procedure code, not procedure code.
Insurance companies tend to send out many different reasons why they won’t pay something that may or may not be misleading. In other words, they might send 10 different types of denials all with the intent of never paying a level 5 and those denial codes may all be in different categories like coding, EDI, etc. Ultimately, it’s really that they’re just never going to pay that code and making things up to put hurdles in place to slow the provider down and waste their time. The insurer may not think it is clinically reasonable to perform that service or they financially don’t want to pay it, so they’re just going to deny things categorically. Unless you can you can figure out that pattern you’ll just be banging your head against a wall.
In order to answer the particular question they were looking for, which was “If there is a group of CPT codes or group of procedures, whether that’s a group of tests in the laboratory or a group of E&Ms or surgeries, for example, is this code reimbursable in combination with this?” You have to know that if you’re doing a chondroplasty or some orthopedic surgical procedure, whether that code is similar to other codes, or whether it’s not going to be paid in combination with this, or whether it might be bundled.
Therefore, unless you have machine learning in your wheelhouse, you have to put together those analyses and ultimately even link up multiple CPT codes, where if you are looking at an encounter level and will want to create keys that link the CPT codes together. It’s probably the easiest way to do that if you don’t have a complex database and software to analyze this, if you’re doing it in something like Excel.
Creating a Data Key Secret
Following is a simple example to illustrate the idea of creating a combined data key (sometimes called a compound key) that would allow you to see if there are problems with a particular code within a group. Take the inpatient E&M codes 99221-99223 and a chondroplasty 29877. Let’s say you billed accidentally a an E&M with the surgical procedure code in the same visit. You would create a compound key that was 99223-29877 and that claim got denied. You would then be able to see a pattern where all compound keys that included an E&M and a surgical code were denied because you shouldn’t do those on the same date of service. This is a simplified version to illustrate the concept, but you can envision where there are combinations that will never be paid and you can flag them by doing a compound key.
So that you actually can see, “Oh, if we have these in combination with these, the payer is going to deny, although they might not deny either code individually.” That’s a little trick that you can use if you’re trying to figure out “does the payer not deny an individual code, but they deny codes in combination with each other?”
The key secret we have imparted today is: Pivot by payer by CPT code and group procedure codes to find what is denied disproportionately. You can then proactively provide documentation to those payers that require it or bill a lower CPT code (not for federal FYI) in order to get reimbursed without getting denied.