Spending hours on hold and transmitting information by fax are business practices from a bygone era. Yet the prior authorization process depends on these costly, inefficient manual processes.

Shawn Tienken, MHA, director of revenue cycle operations at Stanford Children’s Health in Palo Alto, CA, sat with Hospital Access Management to discuss his department’s progress toward automation. (Editor’s Note: This conversation has been lightly edited for length and clarity.)

HAM: How are prior authorizations affecting your department?

Tienken: We are seeing an increasing number of requirements, particularly for services considered to be high cost. The development of targeted, bioengineered pharmaceuticals has been accompanied by stringent authorization requirements, since these therapies are so expensive. As a pediatric specialty institution, we perform a lot of high-cost services. It might be hitting us harder than your average community hospital. There has been more control implemented over access to specialty services. That also has increased the authorization burden. All of those things have made it more time-consuming.

HAM: What progress have you made toward automation of auths?

Tienken: Our organization has looked very closely at automating authorizations for the past three years. Our views on the best ways to do this have evolved over time.

A lot of payers have instituted portal-based authorization and online options, as opposed to [making] phone calls or [sending] faxes. Strangely, we still have to submit a decent number of faxes. Probably 10% to 20% of authorizations are done via fax. There have been some opportunities, either through using a portal or e-faxing, to generate an automated submission. But where we see the opportunity for automation to really make an impact is less on the initial submission than on follow-up. We submit the request, but then we have to follow up. Usually, we can just log into a portal to do it, but it still requires somebody to do that. We were particularly interested in automation for status updates, and there were vendors in the marketplace saying that they could do that.

What we found was that a lot of those vendors could really only do that for a certain subset of payers, or for a certain subset of services. Or, they can do it for the hospital side, but not the professional side.

Some didn’t integrate well with our EMR. All of a sudden, the value we were realistically going to get was shrinking pretty rapidly.

HAM: What about artificial intelligence (AI) or machine learning solutions?

Tienken: We’ve talked to firms that offer these solutions. These seem to offer much greater flexibility in what services they can perform, because the systems learn and adapt over time. We haven’t implemented an AI solution yet. That’s because we are still trying to figure out what’s real and what’s marketing. We are trying to do that by talking to existing customers. The problem is that firms don’t have real track records of actually doing this yet. They say they’ve done it on a small scale, and that it should work theoretically. But we are trying to evaluate what they can really do, and then what the best approach is for actually doing it.

Our decision has been much more about how do we implement something that has the ability to scale and grow over time. Instead of just putting one piece of duct tape on the problem, and then needing more duct tape for other problems, we want to cover the whole issue. Every one of those implementations requires an investment of IT, operational, and project management resources. If you do it piecemeal, implementation costs can become prohibitive. What we’ve learned is there are certain areas where hospitals can get value. If you want to just focus on imaging, and your organization does a lot of that, that’s pretty straightforward. What you have to submit is pretty consistent from payer to payer. That would be a good opportunity to automate. But if that’s a small slice of what you do, the question is: Is it really worth it to invest and get enough of a return?

HAM: How have you made the prior authorization process more efficient?

Tienken: We’ve worked to maximize the efficiency of auth rules within our existing EMR. There’s a good chunk of things we do that don’t require authorization. We want to make sure we only route those things to people ... who actually require it. We have made rules in our system that automatically bypass the whole authorization workflow if payers don’t require it.

But one of our biggest challenges is making sure we align those front end rules with the back end rules we’ve set up for billing and claims. The worst-case scenario is that we bypass the workflow for something that we say doesn’t need an auth, but our claims people on the back end say it [needs an auth]. They stop it when we are trying to send out a bill. We’ve missed the chance to get an auth beforehand, and we’re chasing it after the fact. We’ve come across areas where the rules we initially set up on the front end weren’t matching up with our back end rules. Our front end rules were too broad. They basically said: All visits of this type, for this payer, don’t require authorization. While that may have generally been true, for a certain subset of those usually higher-dollar services, or services with other specific components within that general category, the payer did require authorization. Our front end rules weren’t that nuanced.

We also were paying highly skilled people to follow up on the status of an already submitted authorization. We outsourced that component so we are able to do the follow-up at a much lower cost. We are now focusing our highly trained front end people on initial submissions rather than follow-up. It takes fewer resources to get the submission accurate and complete in the first place than it does to fix things on the back end. The authorization gets approved faster, so payment is faster — with lower cost to collect.