This article first appeared in the AAHAM Summer Journal and was authored by Jason Petrasich, Sr. VP of AI for Meduit.
When a large healthcare system in the upper Midwest struggled with staffing and resource challenges due to the impact of COVID-19, they engaged an expert artificial intelligence (AI) revenue cycle management (RCM) provider to drive support for staffing shortages, solve specific RCM problems and accelerate account resolution. The partnership leveraged AI across the following areas to drive efficiencies, reduce cost and accelerate revenues collected.Editing Claims
The healthcare system has built numerous safety measures on claims to ensure that everything on a claim is correct before it gets sent to a payer. One example focuses on claims related to physicians seeing patients in the hospital.
These claims are reviewed to ensure that any authorization number for the hospitalization is included with the claim. The former process involved simply having a biller check the authorizations screen, copy any new information onto the claim and then release it. Though a relatively simple process, it is highly repetitive.
AI was trained to complete this function and quickly got up to speed, working 2,000 claims a month. With AI managing the work of editing claims, the healthcare system’s RCM staff was able to focus on other claim edits that better utilize their skill set and accelerate the release of claims to keep discharged not final bill (DNFB) statistics low.
Like every health system, checking the status of pending authorizations prior to scheduled services is extremely time-consuming. The healthcare system may have more than 1,000 scheduled services at any point in time. Services requiring prior authorization need to be monitored to see whether the request has been approved, denied or pending further information.
Claims falling into the pending further information category are very time-sensitive and need to be acted on as soon as a problem with the scheduled service is identified. In order to identify these claims, most organizations employ an army of staff to check and manage the process, because it may require five to ten minutes to check each pending authorization.
AI learned to navigate the payer authorization portals, locate the current status and determine the action required. Once AI learned the process for one payer, it was able to quickly learn additional payers. Now AI is statusing more than 250 authorizations a day and freeing up valuable resources in the patient access team.
Modern patient accounting systems include some sophisticated logic required to manage claims and charges, enabling charges to be corrected, transferred, voided, etc. The byproduct of this complex task is that many transactions cannot be automatically attached to the correct claim/service.
The healthcare system had accumulated a backlog of over 17,000 “undistributed” transactions. Every single item needed to be examined and assessed to determine whether it should be applied somewhere, transferred to another account or refunded. These tasks would have taken an employee several months to resolve.
AI learned this complex but repetitive process; it started working thousands of transactions a week and was able to clear the backlog in a little over a month. Now AI keeps that list clean every day, which helps keep the A/R clean and keeps employees focused on accounts where they can drive collections and resolution.
Through AI, the healthcare system was able to automate:
- Clearing backlog of 17,000+ undistributed transactions
- Working 2,000+ claims per month
- Statusing 250+ authorizations a day