Health care fraud: We can't afford to 'pay and chase'

In June, Attorney General Loretta Lynch announced “the largest criminal health care fraud takedown in the history of the Department of Justice.”  More than 240 people were arrested and charged with stealing $712 million from Medicare.  The suspects included 46 doctors, nurses, pharmacy owners and other medical professionals.

While this is a huge blow to fraudsters across the country, the federal government must hasten its move away from this type of “pay and chase” model – where Medicare and Medicaid routinely pay every bill that comes in and investigate cases only when it’s blatantly obvious that something is wrong. To achieve this transformation, the government must closely examine how it is currently using data and analytics.

Much of the government’s data about citizens is siloed in federal and state agencies, limiting the view agencies have of the larger picture. However, third-party data is available, and data integration technologies can be used that are capable of searching, mining, analyzing, linking and detecting anomalies, suspicious behaviors and related or interconnected activities and people. By running a data validation model on multiple sources of data, agencies can see whether a potential beneficiary has provided the government with a false storefront or address, is indeed a citizen or legal permanent resident, has actually been prescribed a medicine or is associated with any known fraudulent activity.

Although checking claims against validated data compiled from multiple sources is the critical first step to fraud prevention, it is far from the be-all, end-all solution. Federal health agencies also need a reliable method for anticipating fraud or risk of fraud even in cases where the fraudsters are not immediately revealed by flags or conflicts in multisource data. This requires a comprehensive fraud framework, with elements capable of detecting suspicious information, activity and anomalies using rules-based algorithms.  It also requires the ability to predict fraudulent trends using scoring engines that can rate, with high degrees of statistical accuracy, behaviors that warrant further inquiry.

Rules-based algorithms allow agencies to detect known types of fraud or abuse based on specific patterns of activity. These algorithms flag inconsistencies or vulnerabilities, such as instances where services provided are inconsistent with medical history, diagnosis or provider specialty. Similarly, claims may be flagged in cases where a services provider was recently established, changed its place of service or address, provided similar services to multiple family members within a narrow timeframe or billed heavily immediately after enrollment.

Again, no single technology is a silver bullet for preventing fraud, and rules-based approaches also have their blind spots. For instance, under a rules-only approach, some fraud may appear to be legitimate, and, likewise, legitimate activity may return false positives. Agencies can mitigate against these possibilities by supplementing their framework with anomaly detection to normalize events and set thresholds to identify outlier behavior. Using distributional analysis, agencies can then analyze variables to identify claims and practice patterns that are extreme outliers relative to the rest of their respective distribution. With this capability, new fraud patterns, identified through outliers, may be brought to agencies’ attention.

The last weapon in an agency’s anti-fraud arsenal should be predictive modeling, which further complements the previously discussed capabilities, especially in attempting to eliminate false positives. Predictive modeling tends to be more accurate and reliable than other analytical methods for fraud scheme discovery. By looking at past behavior, a predictive model will determine what variables have the most impact on the outcome. The strongest variables should then be combined into an algorithm that can be applied to multisource data to highlight otherwise undetectable patterns. This gives agencies a way to quantify their fraud risk, allowing them to calculate a score or likelihood and determine where fraud is most likely to be occurring.

There are hundreds of news stories showing how criminals, organized or opportunistic, are committing fraud to the detriment of our nation’s health care programs. Data and new technology can provide federal agencies with the real-time analytical framework that is needed to curb the rising tide of fraud. These analytic tools allow for fast, reliable decision making, in turn validating claims under the Affordable Care Act.  Employing these tools can save millions of dollars before they are paid out, and help keep taxpayer dollars directed towards their intended recipients: the deserving patients that most need them.