CMS, the Office of Inspector General and other governmental agencies have ramped up their stance on healthcare fraud and abuse, estimating that roughly $4.1 billion in taxpayer money was recovered in fiscal year 2011 in new prevention and enforcement efforts.
Ted Doolittle is the deputy director of CMS' Center for Program Integrity, a new department within CMS that has emerged under President Barack Obama's administration to fight healthcare fraud and weed out the bad providers, so to speak. Here, Mr. Doolittle talks about his department's current priorities regarding Medicare enrollment and claims fraud and how the heavily scrutinized predictive modeling system is shaking out so far.
Question: HHS recently released a report showing anti-fraud efforts recovered more than $4.1 billion in fraudulent healthcare payments last year. What are some of the main things you and CMS are focusing on right now in terms of healthcare fraud?
Ted Doolittle: One thing to remember about that $4.1 billion figure is that the bulk of that comes from qui tam suits — those that are initiated by individual citizens and pursued by others, sort of along the lines of a class action suit. So we have to keep that in mind. There were recoveries in the program, but they all were not necessarily CMS recoveries.
CPI is a pretty new organization within CMS. It was created a little less than two years ago. It took some components that were pre-existing in CMS but created some new groups and functions. Strategically, we're trying to do two main things.
On the provider [Medicare] enrollment side — those who want to be a Medicare provider or supplier — we are expanding our screening efforts. We have an automated provider screening program, which categorizes providers for the first time by risk, and that can be either a category based on a particular type of high risk potential — for example, durable medical equipment supplies are high risk — or more tailored view. We tier them by risk group: high risk, moderate risk and low risk. But this is about prioritizing our tasks. We need to get out the high-risk folks first. We have set up and are still building a system that will allow us to conduct a lot of research-type searches of public records, and [we] are trying to automate a lot of that using a variety of web sources. For example, once you get enrolled, if we're looking at your business affiliations and if something changes, our system will pick up you are no longer owner of a related hospital. We can then assess whether we need to ask for more information. The point is it used to be you got enrolled and then you were good until revalidation, if that ever happened. Now, [the provider] is enrolled, and we are able to get dynamically updated records about it and ascertain whether something needs further evaluation. We are trying to make it harder for bad guys to get in but also to make it easier for the good guys to get in. For example, online enrollment has been really popular with providers. It is customer service friendly to good guys while being tough on the bad guys.
The other side is on [Medicare] claims processing, where we identify bad claims and bad providers. The central game is not about stopping a $500 claim. It's about getting a provider out of the program. If you get the provider out, their operation comes to a standstill. We are trying to improve, have improved and continue to improve our ability to see what's going on in the claims and dynamically stream those claims against algorithms. It's a much cruder way, like how credit card companies operate, to validate who you are. It's the Fraud Prevention System — a computer modeling, predictive analysis system. No one is doing this in healthcare on the scale we are. We started this on June 30, 2011. As a claim comes in, before we pay it, we stream all of the Part A, Part B and durable medical equipment claims. At CMS, we have over 4.5 million claims per day. We've started to stream those claims against a variety of algorithms, and we are continuously developing, arm-in-arm with contractors, algorithms that are more complex and create alerts that can be checked out by us or contractors. Then we can decide if we need to take action. We're ultimately trying to get people out — revoke their billing privileges and kick them out of the program — that are submitting bad bills. Some of the highlights on claims processing have to be integrated with our provider enrollment goals. That's kind of a high-level view of our twin pillar strategies.
Q: The second pillar you mention is the predictive modeling program, or the Fraud Prevention System, which as you said started last summer and aims to catch Medicare fraud before it happens. However, according to a recent Associated Press story, the system has only prevented $7,591 in fraudulent payments. What is the status of predictive modeling? Can you explain how it works and why it has only caught a low number of fraudulent claims?
TD: A lot of people focus on what we're doing on the "payment suspension" side. When people ask about a fraud prevention system, it's natural to gravitate to that single metric. Payment suspension is great, but it's a totally interim solution. Providers can still bill, and claims are still adjudicated. We won't send the check out and will hold it in escrow.
With payment suspension, providers are still in the [Medicare] program, providing services to beneficiaries. I have to get them out of the program. That's where we want to get to. Payment suspension is taking the bus to the airport — you ultimately want to get to the airport. In that article, I described suspending payments as "unsophisticated." You don't just want to put people on payment suspension.
It so happens that we are training contractors on how to use this system, and as it is used and contractors understand how to use it, those payment suspension numbers will go up. But the point I was trying to make is that payment suspension is a new tool, a great tool, a [Patient Protection and Affordable Care Act] tool, but look at where the rubber really hits road. It's like you're talking to a guy at the end of the first day of the Normandy invasion. The general starts criticizing the private because he only used his bayonet once. The infantryman says only one guy got close enough to me, but I used all my other things: grenades, etc. The FPS is not designed to just generate payment suspension. It's designed to give us leads to ultimately kick people out of the program.
When you say CMS has one payment suspension for $7,000, you haven't said how much money we've saved and stopped, and that's why it's an unsophisticated metric to focus on. It should be included along with automated denial edits and revocations and other more desirable solutions. [Payment suspension] is an interim solution while we investigate the real issue.
Q: In what way will this predictive modeling affect hospitals down the road?
TD: We are doing this for Part A and Part B claims as we build the predictive modeling system. We have a constant flow of algorithm developments with CMS and with contractors, and right now, there are a number of algorithms we release every three months. We do an internal development process and figure out what priorities are for the next release, new algorithms for next release. We can target certain algorithms for certain claims types or provider types. When the system is mature in a couple years, I can predict we'll have algorithms with a wide variety of payment provider types. We're still building up that capability.
Nobody has ever done this before on a scale and detail that we're talking about here. Targeting algorithms to certain provider types — we have that capability, and we are starting down that road. We have a lot of algorithms in place to all provider types, but in terms of targeted, specific algorithms — orthopedic surgery, for example — we have to get there. Some of the algorithms are more general, but the plan is to get as specific as fast we can.
Ted Doolittle is the deputy director of CMS' Center for Program Integrity, a new department within CMS that has emerged under President Barack Obama's administration to fight healthcare fraud and weed out the bad providers, so to speak. Here, Mr. Doolittle talks about his department's current priorities regarding Medicare enrollment and claims fraud and how the heavily scrutinized predictive modeling system is shaking out so far.
Question: HHS recently released a report showing anti-fraud efforts recovered more than $4.1 billion in fraudulent healthcare payments last year. What are some of the main things you and CMS are focusing on right now in terms of healthcare fraud?
Ted Doolittle: One thing to remember about that $4.1 billion figure is that the bulk of that comes from qui tam suits — those that are initiated by individual citizens and pursued by others, sort of along the lines of a class action suit. So we have to keep that in mind. There were recoveries in the program, but they all were not necessarily CMS recoveries.
CPI is a pretty new organization within CMS. It was created a little less than two years ago. It took some components that were pre-existing in CMS but created some new groups and functions. Strategically, we're trying to do two main things.
On the provider [Medicare] enrollment side — those who want to be a Medicare provider or supplier — we are expanding our screening efforts. We have an automated provider screening program, which categorizes providers for the first time by risk, and that can be either a category based on a particular type of high risk potential — for example, durable medical equipment supplies are high risk — or more tailored view. We tier them by risk group: high risk, moderate risk and low risk. But this is about prioritizing our tasks. We need to get out the high-risk folks first. We have set up and are still building a system that will allow us to conduct a lot of research-type searches of public records, and [we] are trying to automate a lot of that using a variety of web sources. For example, once you get enrolled, if we're looking at your business affiliations and if something changes, our system will pick up you are no longer owner of a related hospital. We can then assess whether we need to ask for more information. The point is it used to be you got enrolled and then you were good until revalidation, if that ever happened. Now, [the provider] is enrolled, and we are able to get dynamically updated records about it and ascertain whether something needs further evaluation. We are trying to make it harder for bad guys to get in but also to make it easier for the good guys to get in. For example, online enrollment has been really popular with providers. It is customer service friendly to good guys while being tough on the bad guys.
The other side is on [Medicare] claims processing, where we identify bad claims and bad providers. The central game is not about stopping a $500 claim. It's about getting a provider out of the program. If you get the provider out, their operation comes to a standstill. We are trying to improve, have improved and continue to improve our ability to see what's going on in the claims and dynamically stream those claims against algorithms. It's a much cruder way, like how credit card companies operate, to validate who you are. It's the Fraud Prevention System — a computer modeling, predictive analysis system. No one is doing this in healthcare on the scale we are. We started this on June 30, 2011. As a claim comes in, before we pay it, we stream all of the Part A, Part B and durable medical equipment claims. At CMS, we have over 4.5 million claims per day. We've started to stream those claims against a variety of algorithms, and we are continuously developing, arm-in-arm with contractors, algorithms that are more complex and create alerts that can be checked out by us or contractors. Then we can decide if we need to take action. We're ultimately trying to get people out — revoke their billing privileges and kick them out of the program — that are submitting bad bills. Some of the highlights on claims processing have to be integrated with our provider enrollment goals. That's kind of a high-level view of our twin pillar strategies.
Q: The second pillar you mention is the predictive modeling program, or the Fraud Prevention System, which as you said started last summer and aims to catch Medicare fraud before it happens. However, according to a recent Associated Press story, the system has only prevented $7,591 in fraudulent payments. What is the status of predictive modeling? Can you explain how it works and why it has only caught a low number of fraudulent claims?
TD: A lot of people focus on what we're doing on the "payment suspension" side. When people ask about a fraud prevention system, it's natural to gravitate to that single metric. Payment suspension is great, but it's a totally interim solution. Providers can still bill, and claims are still adjudicated. We won't send the check out and will hold it in escrow.
With payment suspension, providers are still in the [Medicare] program, providing services to beneficiaries. I have to get them out of the program. That's where we want to get to. Payment suspension is taking the bus to the airport — you ultimately want to get to the airport. In that article, I described suspending payments as "unsophisticated." You don't just want to put people on payment suspension.
It so happens that we are training contractors on how to use this system, and as it is used and contractors understand how to use it, those payment suspension numbers will go up. But the point I was trying to make is that payment suspension is a new tool, a great tool, a [Patient Protection and Affordable Care Act] tool, but look at where the rubber really hits road. It's like you're talking to a guy at the end of the first day of the Normandy invasion. The general starts criticizing the private because he only used his bayonet once. The infantryman says only one guy got close enough to me, but I used all my other things: grenades, etc. The FPS is not designed to just generate payment suspension. It's designed to give us leads to ultimately kick people out of the program.
When you say CMS has one payment suspension for $7,000, you haven't said how much money we've saved and stopped, and that's why it's an unsophisticated metric to focus on. It should be included along with automated denial edits and revocations and other more desirable solutions. [Payment suspension] is an interim solution while we investigate the real issue.
Q: In what way will this predictive modeling affect hospitals down the road?
TD: We are doing this for Part A and Part B claims as we build the predictive modeling system. We have a constant flow of algorithm developments with CMS and with contractors, and right now, there are a number of algorithms we release every three months. We do an internal development process and figure out what priorities are for the next release, new algorithms for next release. We can target certain algorithms for certain claims types or provider types. When the system is mature in a couple years, I can predict we'll have algorithms with a wide variety of payment provider types. We're still building up that capability.
Nobody has ever done this before on a scale and detail that we're talking about here. Targeting algorithms to certain provider types — we have that capability, and we are starting down that road. We have a lot of algorithms in place to all provider types, but in terms of targeted, specific algorithms — orthopedic surgery, for example — we have to get there. Some of the algorithms are more general, but the plan is to get as specific as fast we can.
Related Articles on CMS Predictive Modeling:
Senators Ask Marilyn Tavenner About Her Fraud-Fighting Plans
Senators Call for Better Metrics to Assess CMS' Predictive Modeling Program for Fraudulent Claims
What Does CMS' Predictive Modeling System Entail?