Podcast: The Evolution of AI in the Revenue Cycle

Steve Wright:

Welcome to the first of what will be an ongoing Optum360 podcast series on topics related to hospital utilization review. I'm Steve Wright, a marketing director for Optum360, and our first podcast topic is about the evolution of artificial intelligence in the utilization review area of the provider space.

Steve Wright:

With application of artificial intelligence moving swiftly in healthcare and other industries, today we want to examine how AI can be applied to the utilization review area to make it more effective and drive more appropriate reimbursement.

Steve Wright:

I am joined today by my colleague, Dr. Kurt Hopfensperger, who is a subject matter expert in all things related to utilization review. Welcome, Kurt.

Kurt Hopfensperger:

Thanks, Steve. Glad to be here.

Steve Wright:

Okay, Kurt, let's jump right in. For someone that is totally new to artificial intelligence, how would you describe what it is?

Kurt Hopfensperger:

Well, Steve, the phrase artificial intelligence is, as you know, quite a bit in the news nowadays, but actually it goes back for decades in the field of computer science. I like to look at it as, most of what we do commonly with computers, for example we're using a spreadsheet, writing a letter in a word processor, or even playing a game, is linear. The computer's response follows a pathway. That pathway is a result of the human intelligence of the programmer who wrote the spreadsheet or the game.

Kurt Hopfensperger:

Artificial intelligence refers to computer programs that can perform tasks that we ordinarily think of as requiring human intelligence. Example, recognizing pictures of objects in the real world or reading reports and figuring out the actual meaning behind the words in those reports. Artificial intelligence computer programs can change, they can adapt, and most importantly, they can learn from their mistakes.

Steve Wright:

Great. That's a good basic background on AI, Kurt. But how did AI start, and how has it evolved in the healthcare space?

Kurt Hopfensperger:

The first attempts at computer-based artificial intelligence actually go back to the middle of the 20th century. At those times, there were a number of predictions that computers would be able to equal the intelligence of a human within just a few years; but of course, those predictions proved false.

Kurt Hopfensperger:

Initially, artificial intelligence focused on mathematical reasoning, symbolic computation, language understanding, and areas such as machine vision. But in the healthcare space, I think it's useful to break down artificial intelligence into clinical and nonclinical applications. There's been an enormous amount of work in, for example, applying artificial intelligence to help physicians with diagnostic problems, such as reading mammograms or other radiology studies. These kinds of AI programs are steadily becoming more useful and reliable.

Kurt Hopfensperger:

But in the nonclinical healthcare space, AI in the form of natural language processing, which is often referred to as NLP, which is the ability for a computer to read and have an understanding of a medical record, has been used for about 20 years now and is constantly improving as well. Computer-assisted coding, for example, has proven very successful, and now AI has expanded into utilization review.

Steve Wright:

More related to that middle revenue cycle area, can you tell us a little bit more about how AI can be applied directly to utilization review?

Kurt Hopfensperger:

The traditional, and what's still present at most institutions, model of utilization review involves reviewing large quantities of usually lengthy medical records, applying some sort of standard or guidance to those records, and then taking some action in terms of further review or affirming patient status or changing patient status. It would be very useful to have a computer learn how to read those records and also how to apply guidelines to make an assessment of whether a particular patient is more or less likely to be an inpatient versus an observation patient.

Kurt Hopfensperger:

It would also be very useful to have a computer recommend when particular cases need further review by a physician advisor. Lastly, it would be very useful to have the computer help the physician adviser by pointing out the most important clinical factors in the patient's chart.

Kurt Hopfensperger:

To make utilization review better, AI should be used to sort cases in realtime into those which should be reviewed by a physician adviser and those which are highly likely to be inpatient or highly likely to be observation cases. Then when a physician advisor uses the AI application to review specific cases, the record is already parsed, and it's highlighted to make the review more efficient and more consistent. We've been working a lot on these approaches at Optum360.

Steve Wright:

Interesting approaches, Kurt. Can you expand a little bit more on the key problems that AI can solve in this utilization review area?

Kurt Hopfensperger:

When I think about that question, I think there are really five key problems in utilization review that can be helped by artificial intelligence.

Kurt Hopfensperger:

First, it's a highly manual process. It requires quite a bit of time and quite a bit of attention of case managers. Case managers don't know which cases they should really be focusing on until they've already had to do the initial review. So we have sort of a chicken and egg problem where they don't know where to focus until they've actually focused. AI can sort all of the cases in a hospital, for example, and allow the case managers to focus on those really requiring their attention.

Kurt Hopfensperger:

Second, manual reviews are necessarily error prone, and it's difficult to prevent subjectivity from entering into the review. Subjectivity leads to variability in the UR process. AI can be used to apply a database of past medical [inaudible 00:06:05] reviews to perform consistent case sorting.

Kurt Hopfensperger:

Third, staffing and availability of clinical resources are often issues. Computers don't get tired, they don't need breaks. So the case sorting by AI can occur 24 hours a day, seven days a week, and that leads to reduced variability and a timelier outcome.

Kurt Hopfensperger:

The fourth area I see is that AI can allow case managers to focus on the most important cases, but it also therefore frees them up to work in other areas that are more impactful, such as discharge coordination, and have them also work at the top of their license.

Kurt Hopfensperger:

The fifth area I'd like to comment on is, how do case managers and physician advisors keep up with changes in so many areas of medicine and surgery as they relate to patient status? AI can tie a specific case to current evidence-based medical research so that the recommendations coming out of the utilization review process are more defensible against denials.

Steve Wright:

Really interesting points, Kurt. It really appears that AI can fundamentally change how utilization review performs and in really impactful ways. But where do you think AI can be applied in hospital administration functions beyond medical necessity?

Kurt Hopfensperger:

That's a really fast-growing area. We're seeing AI used to help with length of stay management. The AI can show an expected length of stay based on the patient's severity and their diagnosis and can help case managers focus on those outliers.

Kurt Hopfensperger:

Natural language processing can be used in the clinical documentation improvement process as well so that hospitals have a more accurate picture of their patients' true risk and true severity of illness.

Kurt Hopfensperger:

We can also see AI used in a compliance perspective. For example, identifying potential condition code 44 cases or cases to consider rebilling post-discharge.

Kurt Hopfensperger:

I would also expect that, just as artificial intelligence is used in the broader financial world, we will soon see more and more inroads into hospital finance and helping guide decisions at the administrative level.

Steve Wright:

It looks like the horizon's pretty broad for AI in healthcare, Kurt.

Kurt Hopfensperger:

Very much so.

Steve Wright:

Well, that's all we have for today. Thanks, Kurt, for sharing your views and expertise. I hope this has been helpful for our listeners. Please be on the lookout for additional episodes of our podcast and thanks to our listeners for their commitment to make healthcare better for everyone.