Field Notes: Medical Center Uses AI for Prior Authorization Process

Watertown Regional Medical Center (WRMC) is a full-service general hospital and Level 3 Trauma Center located in Watertown, WI, just east of Madison. WRMC has served the region for more than 100 years and, for the past six years, has been in a partnership with LifePoint Health, which is a comprehensive health network with over 90 partner facilities and practices spanning across the United States.

The Challenge

Imagine a world where health information management (HIM) teams knew in advance the documentation a payer needed to authorize an admission or conduct a medical necessity review. Through the use of artificial intelligence (AI), information technology systems would populate the needed data for verification, and the documentation would be sent electronically to the right place within the payer organization for a timely decision. This would be amazing, would it not?

There are several patient scenarios that require providers to share clinical data with payers before certain medical care can continue. Currently, these processes are very manual and are a far cry from the scenario described above. Prior authorization is the process by which providers obtain payer approval before ordering tests, drugs, and particularly expensive procedures like MRIs and other radiology procedures. Payers generally require clinical data to support the need for these expensive medications, procedures, or post-discharge care. Payer approval requirements generally vary—both by payer and by plan—and those requirements could change without notice.

Prior authorizations are not the only trigger events for sharing clinical data with payers. Concurrent reviews, which include inpatient utilization review and case management, and post-discharge processes, which include claims submissions and medical necessity reviews, come with their own administrative burdens and inefficiencies.

At any one of these trigger points, each prior authorization must often be customized through a time-consuming and often manual process. Just 12 percent of the 182 million prior authorization transactions completed in 2019 were fully digital, making prior authorization one of the most manual-intensive federally mandated transactions, according to the 2019 CAQH Index.

These bottlenecks cause administrative burden, care delays, and, in some cases, adverse events. According to the American Medical Association (AMA), 85 percent of practicing physicians consider the burden associated with prior authorizations to be “high or extremely high.”

Prior authorizations have been even further complicated by medical attachment requests. While the typical prior authorization includes data fields for clinical information about the patient, payers are now more often requiring more detailed information that must be fulfilled by sending attachments such as operative or therapy reports. And while attachments give health plans crucial information to resolve some claims, the workflow is time-consuming, labor-intensive, and costly. In addition, many attachment vendors support just a limited number of workflows—typically medical claims and workers compensation—but attachments (both solicited and non-solicited) can be required across multiple health plan-provider transactions, including dental claims, prior authorizations, risk and quality measurements, appeals and overpayments, and more. As mentioned, the content of the payer request varies from payer to payer, and it is often not clear exactly what information is needed—which results in multiple submissions to resolve one prior authorization. In addition to the content of the request, the means in which the information is returned to the payer varies as well: Some payers require a fax, some request it be sent through their secure portal, and some even want direct access to the records. One private firm’s report found that faxes account for almost 75 percent of all medical communication, and the 2017 CAQH report predicted administrative costs would reach $315 billion by the end of the year. These costs are driven largely by the continued reliance on phone calls, fax, and mail to manage claims transactions.

WRMC’s prior authorization team had a very manual, complex matrix of payers and requirements that was challenging to keep up to date. Then, once the information was sent, getting answers was not always immediate. Staff spent many hours following up with the payers, with multiple phone numbers, for the approval or prior authorization number. A further complication was that bulk requests to support payment had been increasing, which requires a manual lift to look up and attach that documentation.

WRMC had a main centralized prior authorization department staffed by registered nurses (RNs). A prior authorization nurse can promote the quality and cost effectiveness of medical care by applying clinical acumen and the appropriate application of policies and guidelines to prior authorization requests. This is a fairly common practice; however, it can be perceived as an expensive option. Some prior authorizations were being done in ancillary departments but, for the most part, they were being done in one main office by RNs. In 2017, a series of staffing shortages in the prior authorization office left a major prior authorization backlog. The staff were struggling to keep the clinics filled because of the large backlog. Many of these clinics required prior authorization, as the care services provided in them were procedural. These slow turnaround times also were affecting patient throughput, leaving idle radiology scanners, empty operating rooms, and frustrated patients. Turnaround times were nearing 15 business days to complete some prior authorizations, which left patients “in limbo” for weeks, and any last-minute cancellations were difficult to fill, leaving provider schedules empty and revenue on the table. Limited administrative time and a backlog of prior authorizations had clinic nurses putting in overtime to work authorizations themselves in an effort to get patients the care they needed. Depending on the complexity of the prior authorization request, the level of manual work involved, and the requirements stipulated by the payer, a prior authorization typically took anywhere from one day to a month to process. The 2018 AMA Prior Authorization Physician Survey revealed that 26 percent of providers report waiting three days or more for a prior authorization decision from health plans.

The Stakeholders

There are two general types of data generated with a patient encounter with a healthcare provider: 1) the clinical data, which is the documentation that is generated when a patient encounters a provider for care; and 2) the administrative data, which is generated once the documentation is coded and the claim is generated to send to the payer for reimbursement. This process has been going on for decades, so what is the issue?

Payers need more and more clinical information these days to sometimes “justify” the bill after the fact and, in many cases, to “prior approve” or prior authorize a test before it can even be done or an expensive prescription drug ordered. Historically, that information request, and the submission of documentation, has been done manually and generally through mail or fax. There are some promising technologies on the horizon, such as the Fast Healthcare Interoperability Resources (FHIR). The Office of the National Coordinator for Health IT (ONC) finalized this technical standard in the recent information blocking final rule adopting the FHIR 4.0.1 standards and will require that certified electronic health records (EHRs) incorporate the standard over the next two years. FHIR takes an internet-based approach to connecting discrete healthcare data and is designed to help developers to build standardized browser applications that enable access to data regardless of a provider’s EHR. A growing number of providers and payers are using application programming interfaces (APIs) in their EHRs and technology infrastructure to use FHIR, which can enable real-time data interoperability.

While many think FHIR has “changed the game,” there remains a HIPAA regulation that originated early on that still complicates matters. The regulation stipulates those prior authorizations must be performed using the X12 278 standard, which most EHRs do not have the capability to use. Even if both parties use FHIR, the data still will have to be converted to a 278 standard in order to meet the HIPAA requirements today.

Just when I thought there was nowhere to turn, I received a phone call from a friend whose friend had just developed technology using AI to solve this problem we were having with prior authorization turnaround time. We did not spend any time searching for a comparable, as this company was new and was looking for a beta site—to which we obliged. The technology was a company called Verata Health, which is now part of Olive.

The Action Plan

The AI engine integrated into our Meditech EHR workflow. With over 40,000 payer forms in their database, this solution presents users with a fully completed prior authorization form and curated clinical documents for a complete prior authorization submission. Users can then use the technology to submit prior authorization to payers and track the submission to resolution. Managers have access to track every authorization as well as the ability to monitor productivity and throughput across all departments.

We pulled together a team made up of individuals from HIM, IT, registration, and the business office. We were excited by the initial demo. The first step was to “teach” the AI engine by “feeding” it six months of claims data. While that was happening, the rest of the team worked on developing the future workflows. The technology was able to also “learn” the business rules for our payers, which helped immensely.

Results and Next Steps

Watertown Regional Medical Center saw significant improvements with Olive’s AI platform. As soon as prior authorizations were automatically prepared, prior authorization turnaround times were decreased by 60 percent. The faster turnaround on prior authorizations meant that procedural clinics were now able to accommodate patients more quickly.

In radiology, the department began to schedule CT scans before submitting a prior authorization, a change that not only made patients happier but also reduced patient abandonment and leakage.

With streamlined prior authorizations, the nurses who were working overtime could focus on patient care. While they previously spent over eight hours per nurse per week working on prior authorizations, this tool helped completely eliminate nurse overtime for prior authorizations, a savings of nearly $21,000 per year per RN.

Automating prior authorizations also reduced missed prior authorizations or errors in prior authorization coding. The result was a 32 percent year-over-year reduction in write-offs across every department. It’s worth acknowledging that, many times, when an organization talks about “automation,” there are immediate fears about the potential elimination of staff. However, in this case, this technology enabled the nurses and HIM staff to devote their time and energy to their core responsibilities, thus enabling them to work more efficiently and effectively.

The bottom line is that, through the use of technology, the prior authorization backlog disappeared and patient throughput dramatically improved.

Provider schedules were consistently filled, and one clinic saw a 45 percent year-over-year growth in scheduled procedures. So, greater patient throughput and less patient leakage translated to greater net revenue between 13 percent and 25 percent increase in monthly revenue following implementation, which translated into more than a $3 million increase in annual revenue.

Lesson Learned: Get Involved as Early as Possible

The health information professional can play a very important role in the prior authorization process. Knowing the topic, weighing in on and participating with advocacy, being a leader, and engaging in crafting solutions are some important ways. Early involvement will help avoid unnecessary delays or missteps. We understand healthcare operations, so it makes sense for us to be involved; after all, the requests for data likely come to and are filled by us. HIM is at the intersection of clinical and administrative data, so I can’t think of anyone better than the HIM professional to be involved or, better yet, lead these efforts.

Jennifer Mueller (jmueller@wha.org) is the vice president and privacy officer at the Wisconsin Hospital Association Information Center.