Implementation of Correlation and Regression Models for Health Insurance Fraud in Covid-19 Environment using Actuarial and Data Science Techniques

Fraud acts as a major deterrent to a company’s growth if uncontrolled. It challenges the fundamental value of “Trust” in the Insurance business. COVID-19 brought additional challenges of increased potential fraud to health insurance business. This work describes implementation of existing and enhanced fraud detection methods in the pre-COVID-19 and COVID-19 environments. For this purpose, we have developed an innovative enhanced fraud detection framework using actuarial and data science techniques. Triggers specific to COVID-19 are identified in addition to the existing triggers. We have also explored the relationship between insurance fraud and COVID-19. To determine this we calculated Pearson correlation coefficient and fitted logarithmic regression model between fraud in health insurance and COVID-19 cases. This work uses two datasets: health insurance dataset and Kaggle dataset on COVID-19 cases for the same select geographical location in India. Our experimental results shows Pearson correlation coefficient of 0.86, which implies that the month on month rate of fraudulent cases is highly correlated with month on month rate of COVID-19 cases. The logarithmic regression performed on the data gave the r-squared value of 0.91 which indicates that the model is a good fit. This work aims to provide much needed tools and techniques for health insurance business to counter the fraud.