Inspector General’s Department of Healthcare and Family Services
Halting payments to health care fraudsters
State of Illinois identifies and tackles claims fraud with SAS® Analytics
In recent years, the Inspector General's Department of Healthcare and Family Services transformed its Medicaid program using SAS® Analytics to identify overpayments and prevent further improper payments to health care providers.
Weishin Wang, Assistant Bureau Chief and Project Manager, says the agency needed a solution that would do more than just rely on exception-processing runs to compare providers to their peers. It wanted to use insights from existing integrity review audits to generate future fraud case referrals – discovering fraud at the transaction or patient level, as well as uncovering fraud perpetrated by members of criminal networks.
The underlying fraud platform, based on SAS, uses historical data on previous fraud and abuse cases to develop well-honed fraud predictors. By utilizing the insights from known fraud cases, the system can spot provider collusion and identify undiscovered fraudulent providers and criminal networks – avoiding significant fraud-related financial losses each year.
With this modeling approach, our accuracy is much, much higher ... We've also decreased the time to create a model from weeks to just a few hours.
"From the past, when we suspect an individual or provider of fraud, we go through an assessment study to understand the case," explains Wang. "The process has always been lengthy and requires significant human resources to identify the initial referral issue. Our predictive models were based on supervised and unsupervised Medicaid claims information by using SAS analytical tools to orchestrate the provider risk score index table. We are still finalizing our predictive model, but test results have so far proved that it can successfully direct us to targeted providers. Furthermore, we have customized comprehensive routines, with interrelated patient information, to identify suspicious networking activities among providers.
"We can identify fraudulent activity, such as time-dependable billing issues, non-corresponding medical claims and double billing, and decide whether to refer the cases to law enforcement agencies, terminate their status and potentially recoup the claim funds. Our approach looks at identified providers and variables to create a pre-modeling base of patterns. We then apply a model to the entire list of member providers to score and identify previously unknown cases of fraud. Even with cautionary patterns identified, we don't approach the provider until we model the social network aspects of an individual. By combining the score from each model, we have a stronger indication of fraudulent behavior to investigate.
Looking at data at a very detailed level and finding interrelations through social networks is what Wang calls a DNA (dynamic networks association) routine, which provides the proof that a provider is acting in a fraudulent manner.
"With this modeling approach, our accuracy is much, much higher – we've achieved a very prominent accuracy level. We've also decreased the time to create a model from weeks to just a few hours," he says.
Wang says the agency chose SAS because of its ability to access, integrate and manage data from multiple sources, allowing the agency to run analytics against the data and then share that information instantly throughout the organization.
"Sometimes people can't believe what we find," he concludes. "SAS has a tremendous manipulation capacity, and it is unbelievably more efficient in generating and distributing reports than any tool I have ever used."
Needed a solution that used historical and social network data to uncover fraudulent provider Medicaid claims, claims overpayment at the transaction or patient level, while uncovering fraud perpetrated by members of criminal networks.
- SAS® Analytics
- SAS® Enterprise Miner™
- Identified overpayments and prevented further improper payments.
- Uncovered criminal network fraud.
- Avoided significant fraud-related financial losses.
- Decreased time to create models from weeks to a few hours.
- Reporting efficiency.