To deal with fraud cases in addition to the traditional and important ways to combat them (underwriting & claims management), we have turned our attention to new more advanced technical methods with the help of data science.
Through data science & data mining techniques that include statistical and mathematical methods, artificial intelligence and machine learning techniques, useful information is extracted and located from a large volume of data requirements without the intervention of a human factor.
Our company for years has made several steps in the field of advanced analytics and is one of the leading players in the Private insurance sector on data management and utilization. Through the analytics team that exists in the company and consists of data scientists and data analysts, efforts are made to create mainly supervised machine learning algorithms on data of customers who have committed fraud, which help the company and the health industry to have an indication of how likely each new customer is to commit fraud.
This process can not replace the important traditional part of risk taking. It can, however, offer an extra weapon that will significantly reduce the processing time of new insurance claims, automate the process of anticipating potential fraud claims, reduce the actual cost of company claims, and in the long run provide better and fairer premiums. (fair pricing) to customers.
All the above help our company to achieve the broader goals of its business plan aiming at a profitable organic growth. Of course, as mentioned above, the existence and management of both internal and external data sources is quite important for the development of algorithms as well as for their optimization.
The most important categories of data used to develop and optimize our algorithms, if any, are:
a) Patient data (Age, Gender, Place of residence, Medical history, Contract data)
b) Data of visits (Interval between visits, Total visits, Number of providers visited).
c) Data of the contracting physician (Area, Specialty)
d) Provider Data (Area, Provider Specialization)
e) Cost Data (Total Cost, Drug Cost, Nursing Cost, Medical Visit Cost) f) Drug Data (Drug Cost, Active Substance, Production Company).
However, fraud can occur at different levels and not just at the level of individual compensation.
For this reason, very important variables for the machine learning models that are created when we look at the above information cumulatively at the following levels of focus:
Examination of compensation claim individually,
The overall claims of a patient,
The total claims of a provider,
All claims of any provider-patient combination,
Claims of insured persons under the same contract,
Claims arising from the same doctor.