Since 2019, we have supported a national imaging network’s in-house team with prior authorizations. Our prior authorization product combines technology and human expertise to reliably obtain prior authorizations.
The product performs three tasks:
- determination of authorization requirement
- initiation of cases with the appropriate benefit manager
- follow-up on initiated cases
To see if we could improve outcomes and speed while reducing costs, we proposed testing the efficacy of our authorization determination agent (ADA) with the client. ADA is a machine-learning-based capability that has learned authorization requirement outcomes by analyzing millions of historical prior auth cases. Data from hundreds of our customers has been leveraged, spanning a comprehensive mix of payers, procedures, and geographical locations. Because we’ve been doing prior authorizations all across the country for over a decade, we have a very rich data set. Clients benefit from this proprietary data set when they use our solutions.
Manual VS Automated Determinations
Our prior authorization solution uses ADA to determine authorization requirements for most of our customers. Within the regular workflow of the product, a sample of machine-learning-determined “No Auth Required” cases are manually audited for performance tracking and feedback learning. Claims data outcomes for these determinations are also used. For this imaging network, this capability is not used today, instead authorization determinations are performed manually.
ADA is biased towards accuracy over coverage. The focus is on making sure that ADA does not make incorrect determinations. Thus, ADA determinations are restricted to cases where ADA is very certain on the “No Auth Required” outcome.
An In-Depth Trial To Compare Accuracies
To demonstrate the benefits of our ADA system and assess its accuracy, we ran a trial on the cases. As our teams conducted the actual prior authorization determination manually, we would run ADA in the background, collecting and documenting its output for the same cases. Our solution would only log ADA responses instead of using them to update the case. Prior authorization specialists made the actual determinations.
Our shared goal was to establish our machine-learning solution’s accuracy for the client’s data. We measured both against the specialists’ decisions and claims outcomes for the same orders. Claims data is the most reliable source of validation, hence the accuracy metrics outlined below were calculated using claims outcomes.
ADA Accurately Determines Prior Authorization Necessity With 98.5% In Real Time
After collecting our ADA’s answers for three months, we compared them to both the specialists’ determinations and the claims paid out.

The Advantages Of Faster, Automated Authorization Determination
While both staff and a machine-learning-enabled ADA make rare errors, most providers find using an automated ADA has many advantages.
Our ADA’s real-time responses:
- Enable rapid scheduling, improve turnaround time, and increase patient volume.
- Optimize resources – staff is freed to handle higher-value tasks like patient care. Staff no longer have to chase down responses.
- Reduce the need for new hires during a healthcare staffing shortage when wages are rising
If you are looking to achieve similar results at your organization, contact us at www.infinx.com/request-a-demo.
Missing prior authorizations are a top cause of claim denials. How accurate is your team in determining whether a prior authorization is required? Read how one national imaging network with dozens of locations improved its denials rate by activating the authorization determination agent feature of our comprehensive prior authorization solution.