Experts reveal new ways to fight healthcare fraud
How predictive analytics could reduce health insurance fraud
Quite a bit has been written recently about healthcare fraud. And just recently, the FBI seized 15 million assets in a healthcare fraud investigation. The main reason for such attention in this area, however, is that even without fraudulent practices in healthcare, it is one of the most expensive industries in the US. When considering these two factors, many experts predict the bankruptcy of healthcare funds in the near future.
The crux of the problem
Fraud in the healthcare industry can be narrowly defined, but that definition misses the common view among experts that at a minimum fraud amounts to 10 percent of the total bill. The National Health Care Anti-Fraud Association estimates that 3 percent of the health care industry’s expenditures in the United States are due to fraudulent activities, amounting to a cost of about $51 billion. Other estimates attribute as much as 10 percent of the total healthcare spending in the United States to fraud—about $115 billion annually. For example, all the fraud detection methods applied to overbilling processes for medical insurance claims were not able to tackle this significant loss. The main reason is because of the vast amount of big data; the lack of analytical, predictive technologies; and the absence of cost-effective computing power to be able to detect fraud patterns efficiently.
The overbilling challenge is probably going to be even more formidable when the new International Classification of Diseases, 10th revision (ICD-10), code is applied. Because of the increased ICD-10 complexity and the number of opportunistic situations with both the disease diagnoses codes and the procedure codes, the therapy codes have grown by more than 10 times.
If an effective technology for fraud detection—or an opportunistic code variability anomaly—exists, it has to have the ability to handle every single claim in a precise manner, with the highest possible probability of determining the fraudulent or opportunistic billing. This requirement is based on four considerations.
First, this condition means that the detection technology has to handle all types of data. Even more important than handling all types of data, detecting fraud processes requires discovering all possible fraud patterns; currently, statistical methodologies are exclusively concentrated on finding the best patterns that translate into only one pattern. Big data, in comparison to small data, has many patterns between dependent and independent variables, and all are more or less statistically significant by different prediction units.
Second, the detection technology needs elastic computing and networking resources to process every single claim containing a huge amount of data. Presently, having an on-premises, bare-metal networking environment to analyze big data is costly and inefficient.
Third, the more patterns that are captured by an analytical technology, the more accurate will be the predictions. In other words, the more data that analytical technology digests, the better the result that the predictions will be accurate. This consideration reflects the major challenge for all statistical methods that are available on the market today.
And fourth, effective, real-time fraud detection for medical insurance claims needs to be easy to apply, easy to understand as to why particular overbilling happened in the first place and easy to act on the detected fraud. After more than nine months of testing on real medical claims data, Dr. Mo multi-model, self-learning statistical software for automatic prediction responds effectively to all four of these challenges:
- Dr. Mo does not require any human intervention in modeling. It develops multi-model processes for each single claim, in real time, and provides the probability and amount of fraud. And its insightful, predictive scenarios also provide the factors that characterize the fraud.
- Dr. Mo is powered by IBM Cloud, which uses the IBM SoftLayer infrastructure as a service (IaaS). This cluster computing technology enables large amounts of data to be analyzed with great resources, and additional resources can be allotted on the fly based on the workload. This automation also eliminates major workloads that can impact the IT department—for example, having to manually provision computing resources.
Initial results for a health insurance company for a selected region showed that Dr. Mo can accurately detect significant amounts of overbilling in real time. In the long run, getting this problem under control can save patients significant amounts in their healthcare payments. Now, a way to reduce, if not completely eliminate, fraudulent medical insurance claims or opportunistic overbilling practices exists to get the healthcare industry back on track serving people.