By Ben Wright, Senior Solutions Architect in Fraud and Security Intelligence at SAS
Health care is complex. It’s based on rules – rules about diagnoses, treatments, benefit plans, policies, payment schedules, exclusion and inclusion lists, bank accounts, in-network rates, prior treatments and on and on. For the most part, these rules can be (and usually are) programmed into claims adjudication systems in exquisite detail.
Health care fraud, on the other hand, is governed by three simple, immutable rules:
- Fraudsters will always find a way to misappropriate funds.
- They will attempt different strategies until they succeed.
- If/once their fraud is uncovered, they will change tactics and start again at 1.
In the realm of fighting health care fraud, false positives occur when an alert, edit, allegation or other term signals a claim is potentially fraudulent but, upon investigation, reveals no fraud. The argument goes that false positives are bad, because they waste investigative resources.
However, curtailing errant fraud triggers at all costs assumes the only benefit of fraud detection technology is fraud prevention – but false positives reveal so much more.
When an anti-fraud solution yields few or no false positives, that typically indicate the solution is merely detecting payment errors occurring due to the complexity of health care and benefit plans, or resulting from undetected payment systems errors. Often, the solution provider is analyzing the payer’s own policies and applying clinical rules that were overlooked or too complex to implement in the payment system.
These discoveries can lead to system remediation, general improvements and, yes, even cost savings – but such cost savings is minor compared to the magnitude of fraud losses yet undetected. Again, a lack of false positives signals the solution is finding improper payments and the unintended consequences of complexity, while largely leaving fraud unchecked. This is one reason why the term “payment integrity” has become so popular, even supplanting “fraud detection” in many circumstances.
Analytics-driven false positives tell a story bigger than fraud
While a rules-based approach can get bogged down in complexities, an analytics-driven approach reveals opportunities to improve systems. Analytics discovers anomalies and assigns risk to these events. Analytics can also predict which events are improper and even prescribe actions to take. This is done using robust, valid statistical methods and predictive models. These methods are not new inventions. Rather, computers and software methods have grown powerful enough to apply them to the processing of millions of claims.
Consider these methods the modern answer to yesteryear’s fraud tip line. They find the “unknown unknowns” in health care claims. And they generate false positives in the process…but fraud investigators should appreciate these false positives. The anomalies identified are, in fact, true anomalies – even the ones that aren’t actual fraud – and their discovery is therefore valuable to the organization. They reveal possible problems and inconsistencies that may indicate issues like:
- Unintended consequences of plan design, claim policies, provider or patient populations. An ambiguous definition of eligibility can, for example, result in payments for treatment of ineligible patients. Instances have cost payers millions before being detected.
- Emerging issues for disease/condition management. For instance, providers may be billing traditional modalities or procedures in a non-traditional or unapproved application.
- Adjudication system errors/gaps. Such errors and gaps result from an inability to respond to unanticipated changes, such as expanded telehealth during the pandemic.
- Targeted audit opportunities. Dynamic changes in volume or incidence of any procedure, diagnosis, or drug code may warrant a closer look, for example.
- Overpayment recovery opportunities. While not fraudulent, unintentional overpayments will show up as anomalies, providing an opportunity to recover funds.
- Oh, yes…and potential fraud. A case in point, some unscrupulous providers have taken advantage of loosened COVID rules to upcode or overbill, and to disguise identity theft as telehealth.
- And probably some true false positives. These can be generated by such things as benefit provisions that are managed “by administration” rather than by standard plan definition. If the system doesn’t know about the change, it will generate a fraud alert.
Reducing false positives as part of an enterprise approach
There is a right place and time for quashing false positives. The best opportunities lie in analyzing “known good” and “known bad” historical transactions, which is always a good practice. Advanced analytics, such as machine learning and artificial intelligence, can increase what is known, making models more effective over time. And such methods will still produce a continuum of results, as outlined above.
Applying analytics to what is known and unknown is a voyage of discovery. Enterprise analytics systems break down silos and foster better collaboration and information sharing across the organization. They combine rules-based analyses with a deeper, more sophisticated understanding of a health system’s intricacies and foibles. In doing so, these systems also reveal opportunities for improvement and massive savings while delivering the ability to detect and prevent fraud. They also produce false positives. Thankfully.
Ben Wright, AHFI, is a Senior Solution Architect at SAS, supporting the planning and implementation of solutions for health care fraud detection, investigation management and payment integrity in the state government and commercial health markets. Among his 45 years in commercial health payer IT, Wright has dedicated more than two decades in pre-payment fraud detection, data analytics, and payment integrity process support.