How AI can bring crucial help to the claims resolution process
Advanced technologies can assist with revenue cycle management by resolving routine tasks such as managing payer masters.
Artificial intelligence is expected to bring many benefits to healthcare organizations, but perhaps one of the least appreciated is the benefits the technology, in combination with machine learning, can bring throughout the claims resolution process.
It provides a wide range of possibilities, but what provides the biggest bang for the buck? Over time, it’s becoming apparent that the maximum return on innovation occurs when users work independently of software that is designed to accelerate and optimize processes.
Here, and in a subsequent article, we’ll review the cash acceleration component of revenue cycle management (RCM) after claims submission. That’s a crucial time to have support, because it’s estimated that one out of every three claims require some type of review and additional work for resolution.
As these new technologies are applied to the revenue cycle management process, it’s important for those overseeing these processes to understand how AI and ML can improve business processes.
Sorting out payers
One aspect of RCM that AI and ML can address is in managing payer masters and assigning payer IDs to payer names and addresses.
Sounds simple, doesn’t it? It’s a key component because payer masters are provided to RCM and billing firms in a separate file or access point from patient information. The payer master is critical to the revenue process because electronic transactions must be used to accelerate revenue, and accurate payer IDs are required.
A serious disconnect occurs when data is inaccurate, because electronic transactions require the correct matching of payer names and addresses to payer IDs. Doing so is fraught with danger, because the current approach requires an intensive manual process that is prone to errors that can jeopardize a seamless claims resolution process.
Payer masters may contain hundreds of iterations of the same payer name in the same state. For example, a Blue Cross plan in New Jersey may be labeled as Horizon, BCBS, Blue New Jersey, Horizon Blue Cross or incorrectly as Anthem.
Other payer data may be inaccurate or dated, which can further gum up the process. For example, payer addresses can change frequently, and some of those that may be on file may be more than a decade old or may no longer exist because of mergers or discontinuations of corporate addresses.
This is where advanced technology can make an impact. For example, a revenue cycle management firm with more than 93,000 records of payer names and addresses used automation to assign payer IDs to payer names and addresses, doing so in three to four hours with a 99 percent accuracy rate that was confirmed by a validating step at the end of the process.
The firm said the same process was simulated as a manual process, and it determined that it would take more than a week of manual labor, with an estimated time commitment of about five hours per day. Accuracy via the manual process was estimated at only about 75 percent.
Using AI to facilitate efforts
This all works because of the contribution of machine learning, which is a subset of artificial intelligence.
To understand the synergy, think of AI as the brain – it creates systems with human-like intelligence. Machine learning is the actual process of learning – it’s the mechanism by which the brain gets smarter through experience and data, rather than being told every single rule.
Email is a simple example of machine learning. Applications use a spam filter, a capability that uses machine learning to learn what spam looks like from thousands of received emails instead of having a developer write rules for all messages that might be spam.
In applying machine learning to payer IDs, the technology is able to match addresses to organization names, even when there are duplicates operating in the same state. It does so by first parsing addresses into components and then using fuzzy matching algorithms that incorporate address and name data to calculate a similarity score. Think of fuzzy matching as a technique that identifies and links records that are not an exact match but are similar enough to be considered the same entity.
A more advanced approach, beyond the scope of this article to describe, uses hybrid techniques that combine phonetic and semantic similarity, or a multi-pass system, to first narrow down a large pool of candidates and then use more precise methods for final matching.
The use of AI and machine learning to sort out payer details to compile an accurate payer master is just one use for these advanced technologies to make a significant, cost-effective impact on the revenue cycle and healthcare organizations’ bottom lines.
A subsequent article will look at other ways that advanced computing technologies can bring more efficiency to the claims adjudication process. It’s clear that any savings that potentially can be achieved are critical benefits for healthcare organizations.
Ken Poray is CEO of Integrex Health with 20 years of experience in working with claims status, including EDI and web portal transactions.