Reactive Business Models No Longer Deliver Optimal Value
The stream of recoverable losses that can be generated from pre-defined detection models is running dry as schemes sophisticate, says Dr. Cheek. “It’s becoming more difficult for rules-based systems to identify fraud, waste, and abuse and show year-over-year return because fraudsters are becoming more adept at creating claim patterns that look more generic.”
This gap leaves health plans more exposed than ever. Rules-based systems can only catch highly aberrant behavior.
The typical process a rules-based system supports has limited reach:
- Run multiple queries
- Combine queries to see the bigger picture
- Look at top of list
- Take action on X%
Interestingly, whereas this approach was once “safe,” it is quickly becoming riskier as the SIU’s efforts are no longer fully quantified by serving as a sentinel, notes Dr. Cheek. Now, payers need to show their stakeholders how they are proactively reducing FWA and preventing leakage. Especially with virtual care models and other disruptions. Investigators are expected to deliver results by showing increased avoidance and recoveries.
Where can the SIU find value today? According to Dr. Cheek, and the health plans he consults, it’s in the unknown.
Adaptive Business Models Hold Promise for the SIU
Adaptive methods use “smart” technologies that surface actionable findings which can deliver greater value to payer business units that are responsible for detecting and reducing FWA. Dr. Cheek advises SIUs to consider how more sophisticated detection methods – like the deep learning models his team developed for Pareo Fraud – are the only way payers can uncover equally sophisticated anomalies.
Adaptive tools and business models offer value by making processes more efficient, reducing costs, and minimizing overlap between the SIU and payment integrity. Benefits include:
- More efficient claims selection and validation processes
- Realizing administrative and operational cost savings
- Reducing cost and provider abrasion by reducing false positives
- Less overlap and greater insights offered through bi-directional SIU and payment integrity integration
For detecting healthcare fraud, deep learning systems must ingest an incredible amount of data from multiple sources, link the information together, and analyze it to identify complex trends and detect potential signs of fraud. A.I. can do this at a much faster speed than humans, and with greater accuracy. But smart technology relies on learning from user interaction to better classify the “problem” in order to get better over time.
“Detection solutions using A.I. cannot replace investigator expertise,” says Isbitts, who notes this is a common concern that may cause SIUs to dig in further on legacy solutions. “What they can do is help the investigator find the deeper-rooted schemes and aberrance much quicker, with greater accuracy and less maintenance”, he adds.
With new directives to show value, payers are realizing that rules-based system are static and unable to provide real-time detection, thus limiting overall analytic capability. Furthermore, it may take 20 or more rules (all manually developed) to capture emerging schemes, whereas a deep learning solution can detect new patterns without any programming, and continue to evolve, change and grow as more data is scored. With the rapid evolution of A.I., systems can begin telling us more about the unknown.