Get in touch with us. We'd love to hear from you.
Part 2: How Intelligent Automation is Reshaping Revenue Cycle Management
In part one of this series, we focused on the challenges facing providers in the front-end of the revenue cycle and laid out the implications of errors or inefficiencies during the registration, prior authorization and cost estimation processes. Most importantly, we presented intelligent automation (IA) solutions that can improve revenue cycle performance by streamlining tasks and reducing inaccuracies before claims reach the payor.
Now, let’s focus on the responsibilities and challenges facing the provider’s billing department, specifically in denials management and cash posting. These two phases of back-end revenue cycle management (RCM) offer the greatest value and ease of implementation, respectively, for IA solutions. It’s important to note that each back-end segment presents impactful opportunities for IA (as illustrated in the following graphic) but for the purpose of this post, we’ll place a central focus on denied claims and how machine learning (ML) capabilities can drive significant financial improvement.
The billing department is responsible for many tasks including preparing, reviewing and correcting patient claims. Described below, back-end revenue tasks are aimed at recognizing (or not recognizing) revenue. An effective and efficient billing department is crucial to an organization’s financial success.
The vast majority of claims successfully navigate their way through the revenue cycle and are submitted to the payor and payment is received. The volume of remittance received daily poses a challenge for billing departments as they attempt to address each claim in a timely manner. Processing, posting and account adjudication can require a significant number of resources. This crucial, but labor-intensive process, is ripe for automation and can be completed by optical character recognition (OCR) and robotic process automation (RPA).
Cash posting can be completed in two ways – manually (physical checks) or electronically (e-checks). In the manual process, checks are received from the payor and scanned into the system with the explanation of benefits attached. OCR can be utilized to extract the necessary information from the scanned document. RPA will then take the extracted information and enter in the patient accounting system.
When money is transmitted electronically, the amounts are deposited directly with electronic file transfers (EFTs). The information is received by the billing office with the insurance reference number. RPA can be utilized to extract the necessary information and allow the posting to be applied to the correct account.
This relatively easy-to-implement solution can provide a high rate of return by freeing up resources and allowing them to devote their time to more challenging tasks.
Unfortunately for providers, a clean claim submission and prompt payment is far from a reality. These exceptions, while a relatively small portion of total claims, pose significant financial risk. Change Healthcare, working with McKesson’s Health IT business, . This denial rate, on average, puts roughly $5 million in payments per hospital in jeopardy. According to a detailed report conducted by the Advisory Board, . The study estimates that 90% of all denied claims are avoidable.
There are two types of denials – expected and avoidable denials. Expected denials are those that an organization know will occur based on the current contract or the payor’s ability to accept supplemental information along with an initial claim. These claims will always be returned as ‘denied’ and can be resolved by responding to a request for additional information. A lack of preparation to provide this additional information can increase total days to payment. Avoidable denials are claims that were unknowingly submitted with insufficient or incorrect information, posing a risk to realizing expected revenue. Both of these scenarios can be assisted through the utilization of IA.
As discussed, there is a subset of claims that cannot be submitted with ‘complete’ information due to the payor’s inability to accept supplemental information with the initial claim. An organization must wait for the denial to provide the ancillary requested information and finalize the request for reimbursement.
Through IA, a bot can pre-determine which claims will return with a request for supplementation during coding. A bot can be trained based on previous data and current data – as bots can continually evolve with additional information – to determine the probability of a pending request in addition to what information will be needed for re-submission. Based on this preemptive action, the bot can collect all the necessary patient records and save them to a secure location in anticipation of the denial. Once returned, the bot will then attach the pre-collated information automatically. When attached, the bot will analyze the data to understand any other reason the claim could be denied before re-submitting. To finalize the process, a member of the RCM team will complete the submission.
It’s worth restating that approximately 90% of all claims are preventable. According to the analysis performed by Change Healthcare, the leading cause of denied claims stem from registration and eligibility issues, accounting for 23.9% of total rejections. Missing or invalid claim data was identified as the second most prevalent cause, occurring 14.6% of the time. Addressing these issues can result in a significant improvement to the first-time payment rate. Some of the issues that fall into the category of ‘avoidable’ denials will be corrected in the front-end through the utilization of IA, as discussed in the first part of the series. However, there is a significant portion of these claims that can reach the billing department and be resolved prior to submission.
To reduce avoidable denials holistically, a ML model can be trained to review all incoming denials, categorize them based on the root cause and identify the required corrective action to be completed by a person. This model learns from both successful and denied claims and can be trained to avoid a denial by identifying replicated mistakes prior to claim submission. This is completed through a tagging process by the end user.
Once a claim is denied, it is put through the necessary processes for resolution. The end user will train the model on how the claim should have initially been submitted. The ML model is then exposed to what was initially submitted and what should have been submitted for reimbursement, preventing a similar mistake in the future. As more denials are received and analyzed, the model continues to learn new resolution paths, root causes and ultimately improves its accuracy. With more information, the bot gains a greater understanding of why a denial occurred and how to resolve the issue. These processes can then be replicated in the future prior to submission and preemptively avoid a denial.
Providers need a revenue cycle solution that doesn’t just help resolve denials AFTER they occur, for example, but rather a solution that helps mitigate them altogether. IA comes in a variety of forms with varying applications throughout the entire revenue cycle. Back-end functions are vital to the financial performance of an organization and pose great risk if not properly managed. IA integration can help billing departments holistically. By training machines to preemptively identify problems, alert the necessary staff and allow for corrections BEFORE submitting a claim, organizations can usher in a new, more effective approach to revenue cycle management.