Using AI-assisted information display to drive improved recovery scenarios

In 2015, an Optum team of clinical pharmacists and statistical data scientists recognized the need for a better way to alert payers and claims professionals to high-risk claims early in the claims cycle. When looking at claim histories, they noticed that 5% of claims were still active 3 years later. Their first objective was to find a way to intervene on that subset of claims to prevent them from becoming chronic. This quickly led to the need to assign risk values to claims of all types, no matter how old.

Optum Predictive Analytics use one of the industry’s largest datasets, in combination with clinical guidelines and benchmarks, to support clinical intervention needs. As part of the analysis, our systems trigger a risk-based model to identify future high-cost, long-term claims with potentially inappropriate therapies in a client’s injured worker populations. This AI-driven identification allows our clinicians to triage claims more quickly to address potential needs within an injured worker’s treatment plan that may help create a better therapeutic outcome.

After extensive development, the Optum team introduced another solution: the single “Risk Score,” based on a proprietary set of criteria systematically reviewed by our machine learning algorithms on an ongoing basis. The value is derived from an algorithmic calculation based on an injured worker’s in and out of network pharmacy transactions, compiled and analyzed against 37 different risk factors, including:

• Utilization concerns: Physician dispensing; high-cost claim; use of non-typical medication
• Potential misuse: Multiple pharmacies; multiple prescribers; prescribers flagged for potential fraud, waste, or abuse

These and other factors are used to generate a single Risk Score of 0-10, which accounts for medical, clinical, and economic risk.

Once calculated, the Risk Score and its 12-month trend history is continuously displayed on our adjuster portal at the claim level: green indicates low risk, yellow indicates medium risk, and red indicates high risk.

The Optum Risk Score broke new ground in the industry by employing AI — through an extensive set of risk criteria — to identify claimants at the highest risk for delayed return to function and work, a potentially poorer outcome, and a higher cost to the payer.

By coupling AI models with claims professional education and clinical intervention, we have seen the trajectory of an injured worker’s care dramatically improve. Using that same approach in aggregate, we can positively influence the health of injured worker populations, leading to better therapeutic and economic outcomes overall for individuals and payers.

Through machine learning, Optum continually refines our AI models to improve the identification of risk as our databases grow over time. And as the industry continues to embrace the power of AI, we are well positioned to release ongoing AI developments that deliver even more positive and quantifiable results for payers and injured workers.