A popular observation these days is that big data democratizes decision-making and analytics – in other words, big data mining technologies can bring large enterprise capabilities to many more organizations at a reasonable cost, in part thanks to the Hadoop platform. For the Insurance industry this means the ability to proactively implement greatly improved methods and processes to speed up and expand fraud detection. The usual modus operandi for the Insurance industry when adopting new technologies has been that of a clear laggard. However the high price of fraud and the increasing availability of cost-effective technology solutions for working with big data should push many Insurers into taking advantage of big data insights.
The Insurance Information Institute estimates that property and casualty insurance fraud accounts for another $30 billion in annual losses.
Fraud and the abuse of healthcare services in the U.S. cost an estimated $125-175 billion annually.
This represents the second largest component of the $600-850 billion surplus in healthcare spending.
Fraud for healthcare services can be initiated by all kinds of people: patients, healthcare providers, organized fraud rings, workers in claims services. A common activity is for an individual to enroll for Medicaid in several states. Another common case: providers bill for services that weren’t performed or inflate the cost of services actually performed. The worst provider abuse is to perform unnecessary services solely for the purpose of being paid for them, which can also endanger the health of the patient.
To date, only 3 to 5 percent of fraud is actually detected—and usually late in the payment cycle. Additionally, only a fraction of money that could have been used to provide care is recovered. Some cases are the result of billing and coding errors, which with prevention practices in place, could have been entirely avoided.
Often individuals and organized rings avoid detection by circumventing the traditional methods of identifying fraud that are used by insurance adjusters and claims processors. Compounding the problem is that traditional methods are frequently manual. Today, with big data mining and analytics, insurance companies can put together the fraud picture from many pieces of data coming from many disparate sources. Entity resolution is another vital aspect of fraud detection and should be included in big data analytics processes. The stealth fraud methods of those “flying under the radar” can be identified and stopped – especially relevant for complex and well-organized fraud networks.
Through predictive analytics, sources such as relational data, machine-generated data and information from social venues can be put to work. Predictive models can be constructed to analyze enormous amounts of data to find patterns that indicate claims fraud and hard-to-uncover networks related to fraud schemes, as well as new kinds of fraud that are just getting off the ground. Analytics can bring to light suspicious transactions, activities and anomalies, connecting unlikely pieces that can only come together through big data analytical processes. Such processes should be real-time or near real-time to stop fraud as early as possible.
The continuous improvement of fraud detection procedures benefits from the intelligence gained from big data analytics. This intelligence can also be used to improve overall claims procedures and other related business processes, which undoubtedly will lead to further cost reductions and efficiency increases.
Insurance providers now can have greater control over all claims, provide better services for legitimate claims, and potentially pass on to customers savings from preventing more and more fraudulent claims. The Insurance industry may well be the poster child for gaining clear monetary value from big data and analytics to accomplish critical functions.