Our analytics and detection methods
Fraud attempts are ever-changing, hence it is crucial for us to adopt a multi-layered detection approach that uses the latest data and technology. We use advanced analytics and machine learning to complement the traditional rule-based analytics so that we can identify potential fraud in a timely manner. Data-driven detection of potential fraud is a proactive, complex and analytical approach that leverages data to identify abnormal and potentially risky behaviour patterns that could indicate fraud.
Analytics
One mechanism we use to detect benefits fraud is analytics—studying the data. We have a dedicated team of data scientists and data analysts who look for any sign of abnormal behaviour.
Analytics methods
To detect fraudulent behaviour, we take a data driven approach to reviewing claims. We continuously monitor claims behaviours, providers, groups and members for activity that may be fraudulent.
Our claim analytics team uses several key methods to identify potential fraud, including the following:
Rule-based analysis
With rule-based analytics, we identify risky behaviours based on previously known fraud schemes. We constantly get feedback from business experts, review literature, perform environmental scans and incorporate relevant findings into specific analyses to identify new instances of provider and member claims that follow the same patterns or have the same characteristics as past claims that were proven to be fraudulent.
Network analysis
Network analysis is the process of discovering and analyzing associative links between individuals and groups in our data.
This analysis allows us to identify clusters of risky activity where members are colluding with a provider to abuse the plan, from members who are seeking suspicious behaviours by providers and from providers that are targeting specific group members.
Anomaly detection and pattern recognition
These analytics identify abnormal patterns of behaviour. It helps us identify individuals that are outside the “norm”, based on specific risks and behaviours related to their peer groups’ claiming activity.
Clustering techniques
Clustering methods are used to identify groups of similar entities (providers or members) together using multivariate data.
Market basket analysis
This analysis shows the combination of products that most frequently occur together in claims submissions. It helps us to identify abnormal claim bundles.
Machine learning:
Using machine learning technology to determine the likelihood of a claim being fraudulent.
Audits and investigations
After a suspicious member or provider has been flagged by our analytics team, we escalate and conduct an audit or investigation. We deploy multiple layers of protection that include:
- verifying and reviewing a claim before a payment is made,
- continuing to audit a plan member or health care provider after a payment has been made.
You can also play a role in helping us stop fraud. Learn how to detect benefit fraud and how you can report benefits fraud to us.
The information on this page is based on data provided by our private book of business.