Public Healthcare Insurance Fraud Mitigation: Big Data Analytics

In the 2010 World Health report, the world health organization listed fraud as one of the top ten leading causes of inefficiency in healthcare with recent studies having calculated that nearly 6.9 percent of all the healthcare expenditure is wasted in fraud.

Historical fraud detection methods only uncover about 10 percent of losses, and because of the post-payment nature of such methods and the resulting pay-and-chase recovery process, less than 5 percent of losses detected are ever recovered.

Fueled by technology advancements that have made crimes such as identity theft and multiparty fraud schemes both easy to commit and hard to detect, healthcare fraud continues to grow. It holds particular appeal for organized crime syndicates, which account for a growing proportion of healthcare fraud, waste and abuse.

Back home the effects of this proclivity are more profound largely driven by the slow adoption of data driven decision management. For instance, the Association of Kenya Insurers commissioned a survey on the health sector Fraud in the country. According to the report, Medical insurance had the highest loss ratio of 81.5 percent for 2010 with an average loss ratio of 78 percent over a five-year period from 2006 to 2010.

The report noted that collusion between beneficiaries and health service providers, lack of efficient interrogation and detection software, poor internal controls and poorly trained claims processing staff were isolated as the drivers of fraud in Kenya. The report further indicated that business leaders were aware of the need to address fraud and implement fraud prevention initiatives.

Data-driven decision management (DDDM) is an approach to business and corporate governance that values decisions that can be backed up with verifiable data. The success of the data-driven approach is reliant upon the quality of the data gathered and the effectiveness of its analysis and interpretation. It also requires a combination of culture change, skill change and technology.

According a previous NHIF Annual Report, the Fund had registered a growth in Benefit Payout Ratio which was attributed to the roll out of enhanced and expanded benefit package for the national scheme accessible to NHIF beneficiaries countrywide. Diversification in the payment mechanism led to beneficiaries having a broad choice of facilities to access the needed care.

With this rapid expansion also came increased instances of fraud. The healthcare domain has been an easy target for people who seek easy money by using fraud methods. Healthcare fraud is expected to continue to rise as people live longer. This increase will produce a greater demand for Medicare benefits. As a result, it is expected that the utilization of long and short term care facilities such as skilled nursing, assisted living, and hospice services will expand substantially in the future. Additionally, fraudulent billings and medically unnecessary services billed to health care insurers are prevalent throughout the world.

These schemes are becoming increasingly complex and can be perpetrated by corporate-driven schemes and systematic abuse by certain provider types. The net effect of this being a need for higher premiums for schemes sustainability in that every shilling spent on a fraudulent or abusive claim reduces the amount of money available to improve quality of care for beneficiaries and this is poised to be a barrier to the Agenda Four drive unless addressed.

Current form of Fraud Management is based on enterprise data and heuristics

Fraud detection & prevention is mainly executed in two methods – fraud audit rules and fraud prediction score card. Fraud audit is the most widely used method compared to prediction scorecard.

Audit rules method involves tedious manual work; one needs to audit all the claims one by one to detect the fraud. The final judgment is taken by the auditor which leads to judgmental error and inconsistency between two auditor judgments. The audit rules or general check points are designed purely based on previous experience and intuition. Audit rules are used to check the genuineness fields like Total amount billed, Total number of patients, Total number of patient visits among other parameters.

While audit rules do not rely on process automation, prediction scorecard is based on computer based statistical analysis; which unfortunately, suffers from lack of real time data. The models built on historical data, which tend to lose their prediction power beyond certain extent. In this method the probability of fraud is predicted by using a mathematical predictive model.

The model is built based on historical data which involves fraud or non-fraud indicator along with other explanatory elements like billed amount, number of patients, Reporting Lags, Treatment Healthcare claim system leverages the power of Big Data platform by delegating its analytical needs.

A Big Data platform has ability to sift through a huge amount of historical data in relatively shorter amount of time, so that the business transactions can use fraud detection on real time. In order to exploit both of these, one needs a more powerful computing platform. Big Data platform is not only capable of processing terabytes or petabytes of data but also supports massively parallel processing.

It’s high time that the health insurance sector starts exploiting big data technology to secure the future sustainability of their organisations.

More about this is covered under the topic Big Data In Healthcare in my upcoming book- Big Data and Predictive Analytics: Raise your Data Quotient