Machine Learning is touching almost all kinds of industries including the healthcare industry. Techniques in machine learning and artificial intelligence are covering the healthcare industry in an enormous way, including the Medicare vertical. With the graph of medical data growing exponentially, it is easier now to achieve great insights by machine learning methods. But on the flip side, it can have serious issues such as the susceptibility to commit Medicare frauds. In one instance in Uttar Pradesh, 21 people were infected with HIV from contaminated syringes by a fraudulent physician in the name of cheaper treatment.

Some Of The Common Types Of Medicare Frauds Are:

• Medical Identity Theft: It is being done to obtain false insurance payments. Here, a person’s identity or personal information is stolen that will be used without the knowledge of that person for obtaining medical benefits such as goods, services, etc.
• Upcoding: One of the major processes is known as upcoding where the code of an affected person’s diagnosis is inflated and is coupled with a false code
• Diagnostic-testing: Medicare frauds in diagnostic-testing schemes are also one of the common cases. It is done by performing unnecessary medical procedures for claiming huge insurance payments
• Fraudulent Schemes: Private insurers, as well as the government-based programs on healthcare, have to deal with fraudsters who roll out fraud schemes, for instance, provider-billing data and genuine patient data are used to fabricate false claims from imaginary clinics
• False Representation Of Non-covered Medical Treatments: It is one of the most common kinds of frauds, where unnecessary treatments are shown to be of importance to gain higher insurance payments

How Is Machine Learning Solving These Frauds

Machine learning portraits a crucial role in detecting these fraudsters because of its quick prediction results based on the available data. Both the techniques i.e. supervised learning technique and the unsupervised learning techniques can be used to detect such fraudsters at a high accuracy rate

The supervised machine learning technique follows its traditional process to detect Medicare fraud. For instance, this method can be carried out by feeding both the data of fraudulent cases and billing data created by the physicians or the medical centre where the patient was checked. As per the learning technique, a model between the two datasets based on these data connections will be created such that if a new instance is provided into the model, it will quickly predict whether it is a fraudulent case or not.

On the other hand, in the unsupervised machine learning technique, the model usually predicts through cluster analysis. This analysis uses the statistical method to split the data into different groups in some specific dimensions.  In this method, the Medicare claimed data is fed into the model and then it tries to detect meaningful patterns to predict future frauds.

In this research paper, the researchers from Florida Atlantic University, USA compared several machine learning methods to detect Medicare fraudulent. They performed a comparative study with supervised, unsupervised and hybrid machine learning algorithms using four performance metrics and class imbalance reduction via oversampling and 80-20 under-sampling method. The research includes 10 techniques that are categorized into three groups; supervised, unsupervised and hybrid.

The supervised learning techniques include gradient boost machine, random forest, deep neural networks, and naive Bayes. The unsupervised learning category includes autoencoder, Mahalanobis distance, k-nearest neighbours, and local outlier factor and the third group includes a neural network model that is pre-trained using the unsupervised autoencoder and also a combination of multivariate regression and Bayesian probability. As a result, the supervised learning technique gives more accurate result than the rest of the two techniques.  

How It Is Being Implemented So Far?

India has been quite ahead in implementing machine learning and analytics to overcome fraud in Medicare. SAS for instance, which is a leader in analytics has developed a robust fraud analytics engine that helps to spot payment integrity breaches either in real-time or in batch. It uses advanced analytics with embedded artificial intelligence and machine learning techniques to uncover suspicious activities with greater accuracy.

Another initiative called Ayushman Bharat PMJAY (Pradhan Mantri Jan Arogya Yojana), launched in September 2018, which is the biggest government health insurance scheme in the world, used tech to fight fraud. It has developed an anti-fraud framework, guidelines, and policies, responsible for providing broad oversight of the various processes and if there are any fraudulent cases. They have designed infrastructure and protocol for advanced analytics, developed data standards, guidelines for data consolidation, mining, and more. They use predictive modelling, machine learning models, regression techniques and social network analysis to detect frauds, for which they have hired five analytics firms including SAS to carry the work. The other companies that they have roped in are MFX, Optum, Lexis Nexis and GreenOjo.