Identifying Anomalies in Claims Data Leads to More Savings for Healthcare Payors

How anomaly detection works

Working with our data mining team, we use statistical machine-learning methods to identify anomalies in the data. Our experts then review the anomalies identified and determine which are the most important in terms of the overpayments they cause and whether action can be taken to correct and prevent them from reoccurring. Next, the team helps to design a programmable rule that can prevent the anomaly from occurring again. In the vaccine example above, for instance, the rule could be: a single patient cannot receive more than one shot of the same vaccine on the same day. And finally, the team keeps an eye on that rule after implementing it.

Anomaly detection works hand-in-hand with rules based concept development

Our outlier anomaly detection complements the rules-based concept development we already employ within our Payment Integrity services. With rules-based concepts, we rely on our expertise, experience and information we read or learn to program rules-based concepts into our systems. These are rules based on what we know shouldn’t happen, for example, two codes that shouldn’t be billed together.  In short, we have an idea, we code for it and then see the value.

In contrast, the concepts we program from our anomaly detection findings are based on oddities we identify in the data. In these cases, the data, not our experience, tells us what needs to be programmed. Additionally, unlike with rules-based concepts, we already have an idea of the value.

The two approaches complement each other.

Anomaly detection leads to more savings for Claritev’s Payment Integrity services

To date, every client for whom we have run data through the model has experienced incremental savings.

Learn about Claritev’s Payment and Revenue Integrity services here.