From just about the time we’re born to just about the time we die, we’re expected to follow the rules. There are rules at home. There are rules at school. There are rules at work. There are rules in society generally.
That’s one of the things that makes the Mad Max series of films so appealing to so many of us. In that post-apocalyptic world there are no rules, per se. You learn what works (and what doesn’t) as you go.
We are now at a similar point in the world of healthcare charge capture, albeit for a different reason. Instead of everything being chaotic a la Mad Max, we have actually been very successful in building a charge capture system that has a high signal-to-noise ratio. Most organizations today are billing correctly roughly 98 percent of the time.
Still, the remaining 2 percent that’s sneaking through the current rules-based system represents millions or even tens of millions of dollars in lost revenue to large hospitals and health systems. That’s where you need to think a bit more like a Mad Max character – throw out the rules and take a different approach. But instead of a souped-up car or truck, the vehicle to charge capture success is predictive analytics and machine learning.
The problem with rules
As we all know, healthcare is dynamic and changing. Organizations grow through mergers and acquisitions, patient demographics change, provider staff changes, new drugs, devices and care patterns are coming out all the time. This is a tough environment for rule-base systems, especially for charge capture.
If you are relying completely on a rules-based system for charge capture, it means you must think of every possible situation that could occur and write a rule against it in order to be successful. You may be able to do that for an individual hospital or even a small health system. But as the organization grows, attempting to write rules to capture inconsistencies when new facilities come online gets too complex.
Here’s a great example. A common rule is if there is a code or claim for the insertion of a device, such as a pacemaker, there should also be a code for the device itself. Makes sense, right? So let’s consider a case where a patient receives a pacemaker. The patient may also receive an inexpensive “lead wire” in addition to the pacemaker.
If the “lead wire” is classified as a device, then the rule is satisfied – there was a code for the insertion of a device (the pacemaker), as well as a charge for a device (even though it was for the “lead wire”). The problem is, the claim is missing the pacemaker charge (typically tens of thousands of dollars)!
It won’t come back as a denial so the organization can make the correction, either. Why would it? The payer isn’t looking for missing charges. Do that a few times and it adds up quickly.
There are so many individual situations like that it’s impossible for any organization, no matter how diligent it is, to think of all of them ahead of time. Worse, if you are relying on rules to highlight missing charges in these situations, the same costly mistakes can keep occurring until someone catches them – either through an audit or by accident. As the organization grows, and more different ways of working are introduced into the mix, the potential for these issues grows exponentially along with it. So does the revenue lost as a result.
Machine learning intercepts missing charges
This is where predictive analytics and machine learning can make a huge difference. Rather than having to think of every possible missing charge scenario up-front, machine learning analyzes what has happened historically across the organization and highlights any trends or patterns that fall outside the norm.
In the case of the pacemaker, as long as the organization has done a reasonable job of coding for that specific device historically, machine learning is able to identify that there is a serious discrepancy between the claims that have the device and those that don’t. It then alerts the proper personnel to look into the aberrant claims to see if there should be a code for the pacemaker itself so the claim can be corrected (and the revenue captured) before it is submitted.
This ability to identify and alert the organization to issues has a longer-term benefit beyond that specific charge. Knowing there is a problem, the organizations commonly work to fix the problem upstream.
Since machine learning approaches to charge capture are based on extracting and leveraging historic charging trends, they perform much better than a completely rules-based system in the face of variation or ambiguity in care patterns. It does this by creating a finer granularity around the organization’s charging practices as well as the physicians who are delivering the care. Here’s how that might apply in an orthopedics setting.
Imagine you have two physicians who do knee replacements. One uses two screws and a bolt to repair the joint, while the other typically uses five screws and two bolts. Both methods are considered acceptable under evidence-based medicine.
If you are completely rules-based, you would have to know how each physician works ahead of time. You would need to develop a hypothesis, test, and write a rule that covers each one. In a large hospital or health system that’s a lot of potential rules to create just around bolts and screws.
Machine learning systems can look at the historical patterns by physician and recognize the number of screws and bolts that are within the norms for each. By adjusting this way, the system ensures charges are captured fully while simultaneously reducing the number of false positives the billing or coding departments must follow up on.
New sense of order
While rules are necessary and even helpful, they can only take you so far – especially when it comes to ensuring proper charge capture.
To realize the millions of dollars of revenue your organization may still be missing, you may need to take more of a Mad Max approach. By taking advantage of predictive analytics and machine learning, you can discover the outliers hidden throughout the organization and become a revenue cycle management hero in the process.
Author Bio: Paul Bradley, PhD, is the Chief Data Scientist at ZirMed, a recognized leader in cloud-based revenue cycle software and predictive analytics. He is an in-demand expert in the field of predictive analytics. In his role at ZirMed, Dr. Bradley oversees the research and development focused on leveraging predictive modeling and machine learning on revenue cycle applications.
The views, opinions and positions expressed within these guest posts are those of the author alone and do not represent those of Becker's Hospital Review/Becker's Healthcare. The accuracy, completeness and validity of any statements made within this article are not guaranteed. We accept no liability for any errors, omissions or representations. The copyright of this content belongs to the author and any liability with regards to infringement of intellectual property rights remains with them.