24. November 2020

SCCH: Combating Insurance Fraud Using Data Analysis

Intelligent algorithms can detect patterns – and consequently anomalies – in large amounts of data quickly and reliably. This fact can be exploited to combat insurance fraud, as shown by a project being carried out by SCCH in association with the ÖGK, the Austrian health insurance provider.

Insurance fraud is costly for an economy. In 2015, a law on combating benefit fraud was passed in Austria whose aims included protecting medical insurance companies against fraud. Together with the RAD competence centre (risk and anomaly analysis in the employer sector) and the Österreichische Gesundheitskasse (Austrian health insurance provider) SCCH developed an automated suggestion aid for suspected cases on the SmartDD COMET project. This led to the development of an inspection system that was successfully implemented all over Austria.


Automated detection of anomalies

When trying to detect possible cases of fraud, the experts at RAD had previously been faced with a large number of data sources, local disparities and a wide range of different enterprises and insured parties. Moreover, there is an enormous variety of fraud patterns. They therefore used analyses that concentrated on particular scenarios, for example certain branches of business, past pecuniary inconsistencies or specific patterns of fraud, to carry out random checks. The researchers at SCCH used these scenarios and time-series data to develop a model that compares individual branches and companies quickly and automatically.

Our task was to automatically detect suspected cases resulting from employers’ actions,” explains project manager Johannes Himmelbauer, data science expert at SCCH. “With an intelligent combination of machine learning methods and statistical anomaly analysis, irregularities in the available data can be identified on the basis of the findings reached by RAD. The result is a monthly scorecard of those companies that have atypical figures. The power of decision remains with the experts.”

“The scorecard helps our experts to pick out relevant suspected cases,” says Dr Gerhard Mayr, head of the insurance services department of the ÖGK with satisfaction. “The advantages are precisely directed and prompt measures for tracking down the small number of black sheep. The higher the score of an employer on one of the lists, the greater the probability of finding an instance of fraud.”

Central system for the whole of Austria

The project was continually expanded and in 2018 rolled out from Upper Austria for the whole of Austria. To this end, the existing model was expanded and trained with the specific data from the other provinces to show the differences between the regions. The result is a comprehensive experts’ system that can be applied to every enterprise in Austria in precisely the same way. New patterns of fraud and new legislation mean that the battle against insurance fraud is not static. To account for this, the models were designed dynamically and can be recalibrated if need be.

(c) RAD (screenshot of the RAD dashboard tool: detail view of a company)