AI Solutions for Healthcare

Psychotherapy Analytics and Savings: How Highmark Uses AI to Proactively Combat Payment-related Fraud, Waste and Abuse

At the NHCAA 2020 Annual Training Conference, Codoxo Healthcare Fraud Analyst Derik Ciccarelli and Highmark Director of Financial Investigations & Provider Review Latrisha Oswald detailed how an AI-driven solution from Codoxo successfully helped uncover and prevent provider fraud to keep costs under control.


AI-assisted analytics platforms are valuable tools to assist healthcare companies’ special investigative teams in identifying fraud and wasteful practices. Codoxo’s solution, for example, uses a Suspicious Trends analysis that compares utilization of specific codes across providers to identify outliers with high concentrations of claims for those codes, as well as providers utilizing codes that are not within their range of contracted services. It also offers a time behavior analysis that looks at trends and spikes to identify an increase or decrease in utilization of a particular code set.

These comparisons are critical to identifying outliers and changes in provider claims activity over time to help health plans uncover systemic and costly issues that might otherwise go undetected. Here are two examples of how healthcare leader Highmark has successfully leveraged Codoxo’s AI-driven technology to audit their BCBS client’s professional providers, identify wasteful practices, and realize significant cost savings.

Case example 1: Uncovering claims and processing errors

Codoxo’s Suspicious Trends analysis uncovered a psychiatric substance abuse provider with a significantly higher number of charges for certain hospitalization and psychiatric services procedure codes than its’ peers. This provider’s average charge per code was also much higher. Further data analysis found the service was one that needed to be specifically contracted, which explained the provider’s higher usage of those codes, but not the higher charges.

Highmark conducted an audit that revealed a claims processing error was responsible for the provider billing significantly higher charges for that code. Highmark successfully negotiated with the provider as well as other providers contracted to use those codes and to obtain recovery from those providers. Based on this individual provider issue, Highmark also implemented system-wide contracting and claims processing improvements that enabled them to realize a significant amount of recovered payments and projected cost savings across other similar providers.

Case example 2: Detecting suspicious trends

In a separate case, Codoxo’s analysis flagged a provider whose incorrect coding of psychiatric sessions led to sizable  dollars in incorrect payments. An aggregate query identified this provider as having the highest paid risk exposure and high services for procedure designation for psychotherapy sessions – a single procedure code accounted for 98% of the provider’s billings and 90% of their overall spend.

After identifying this provider, Highmark conducted a standard investigation. A quick internet search revealed the provider was advertising 50-minute sessions, which fell short of the 53-minute time requirement to receive the higher reimbursement amount for that specific code. In addition, it was revealed the provider was conducting group therapy while billing for individual sessions. A review of sample medical records obtained by Highmark, plus patient interviews, supported their findings.

Highmark used this evidence as the basis for an audit and was able to recover overpayments. Without the initial identification of this provider through data analysis, the fraudulent payments may have gone undetected indefinitely.

Identifying fraud, waste and abuse in healthcare payments

Partnering with third-party technology platforms gives health plans access to valuable tools to identify providers, facilities, pharmacies and even members that are engaged in fraud schemes. It can even uncover emerging schemes earlier in the cycle, before claims are paid, giving health systems an opportunity to investigate and address fraud before it impacts their bottom line. In addition, health plans can use their learnings to focus resources on events that have the highest potential for cost savings and recovery.

Technology solutions like Codoxo’s Forensic AI platform use artificial intelligence and machine learning, combined with predictive analytics, to analyze large volumes of data including claims, lab requests, provider history, member eligibility, and other types of data, in real time, to reveal trends and anomalies. Through several types of analyses these solutions can quickly detect fraudulent or wasteful practices that increase costs and build connections across the data to provide actionable insights.

Codoxo’s patented fraud, waste and abuse detection technology has the added distinction of being the industry’s only self-learning AI solution that detects repetitive unusual behaviors automatically to uncover new fraud schemes as they emerge, while also gathering supporting evidence to provide the rationale behind each identified pattern. This information gives health plans evidence to audit specific providers and build a roadmap to expand their investigations to other providers.

Health plans can maximize savings through quick, proactive detection of emerging fraud schemes and wasteful practices. As illustrated by Highmark’s results, enhanced AI technology platforms give health plans a window into “blind spots” that aren’t obvious through routine data analysis. Revealing outliers in their provider data gives health plans a solid starting point for audits and investigations that can uncover additional issues. It can also be the catalyst for process improvements that can aid in recovery of incorrectly paid funds, and result in future cost savings.

You can watch Derick Ciccarelli and Latrisha Oswald’s presentation from the NHCAA Annual Conference on demand.

Watch webinar now >

If you are an existing customer, please contact your Customer Success representative or for assistance with this topic.  If you would like more information or to setup a demonstration of the Codoxo Forensic Platform, please contact