Santéclair: Detecting Fraudulent Claims More Effectively

Insurance organizations are all exposed to
 fraud risks, and given the complexity of data sources involved, introducing machine learning and AI is a great opportunity to improve fraud models.

Whether dealing with false claims, 
false billings, unnecessary procedures, staged incidents, withholding of information, and much more, the insurance industry must be on the cutting edge
 of technology to stay ahead of fraudsters and reduce losses. With limited resources on fraud investigation teams, every investigation into a case ultimately identified as low risk is wasted time.

 

Advanced Analytics and AI in Healthcare

Go Deeper

Hiring more staff to conduct these manual audits is an expensive and inefficient option. Instead,
the key is optimizing that team’s work by using
big data to detect fraudulent activity with a higher degree of accuracy. This means using data from multiple sources (from patients and providers) and analyzing them together so that audit teams look only at the highest-risk cases to detect more fraud.

Santéclair is a subsidiary of several supplementary health insurance companies (Allianz, Maaf-MMA, Ipeca Prévoyance, and Mutuelle Générale de la Police). They support the health care of more than 10 million beneficiaries, helping to cover optical, dental, and aural expenses as well as dietetic and orthopedic services. For more than 13 years, Santéclair has proven their expertise in risk management, benefiting more than 50 health insurance companies.

Image result for Santeclair"

Challenge 

Santéclair found fraudulent reimbursements stemmed both from opticians as well as patients, however it didn’t have a system that would analyze the right data and adapt with increasingly sophisticated fraudsters. Instead, it relied on “if-then-else” business rules to identify likely fraud cases, which resulted in the manual audit team spending their time on too many low-risk cases. With the increase of reimbursement volume (more than 1.5M a year), Santéclair needed to improve efficiency and productivity.

In less than a year, Santéclair has developed
 an unprecedented fraud detection system using Dataiku that allows our company to handle a growing volume of invoices and to control costs. By choosing Dataiku, Santéclair was
able to internalize its data skills and pursue additional analytics projects.

Jocelyn Philippe,
Head of Partnerships and Development| Santéclair

 

Santéclair found Dataiku via a POC led by the IMT TeraLab platform. Eulidia produced an algorithm using Dataiku to help the manual audit team identify more fraud by feeding them cases with a high likelihood of actually being fraudulent.

Santéclair identified these high-risk cases using Dataiku by:

  • Outsmarting fraudsters with advanced machine learning algorithms that continually update and automatically learn or retrain using the latest data so that any new fraud patterns are immediately identified and audited. Dataiku handles the entire workflow, from raw data to exposing the predictive model to the operational applications.
  • Automatically combining hundreds of variables from different datasets, including patient/prescriber history, interaction graphs, prescription characteristics, and other contextual data.
  • Allowing teams to develop their data science skills through Dataiku’s collaborative, easy-to-use interface.

Without Dataiku, the marketing team would have to rely on the technical team to continually provide or update data, which would be inefficient and ineffective for both teams.

Due to the comprehensive solution developed with Dataiku, Santéclair and Eulidia have:

  • Enabled fraud detection teams to target actual fraud cases three times more effectively.
  • Reduced time-to-market for similar projects by making a POC in a few weeks and then industrializing the project within a few months with a low impact on the IT team.
  • Saved their customers a lot of money by decreasing fraudulent behaviors in the health network and excluding the fraudsters from the network.
  • Saved time with a model automatically updated and monitored along the way to prevent drifting of performance with little human supervision.

Machine Learning-Based Fraud Detection in Healthcare

Health fraud is estimated to cost $68 billion annually, and it can be effectively tackled with AI.

Read more

Go Further

ADA: Aviva's Algorithmic Decision Agent

Hear from Aviva, as they look to leverage their vast data resources to drive world-class customer data science into action.

watch now

Anomaly Detection at Scale

Because of its wide array of applications, mastering anomaly detection in the age of AI is critical.

get the guidebook

Privacy-Compliant Data Projects

This guidebook is meant to help data teams negotiate tricky waters, assisting everyone in getting value out of data in increasingly regulated times.

get the white paper

Improving Efficiency and Accuracy in Healthcare

Data science, machine learning, and AI applications can be the key to improving current logistical and economic issues that influence patient care.

learn more