By Jim Stinson, Business Observer Florida, May 8, 2004
Bottom line
Key takeaway: Apprio, led by CEO Kali Durgampudi of Tampa, eliminated 75 million annual computer tasks of 10 Tampa Bay hospitals and another 42 locations, allowing automation software to work for the revenue departments.
Core challenge: Convincing people the intensely efficient software won’t lead to layoffs.
What’s next: Apprio plans rules-based artificial intelligence as the next step, a second phase in hospital “revenue cycle management.”
To fight a computer task, one must think like a computer task.
Kali Durgampudi thinks computers, ever since he earned his master’s degree at the University of Central Oklahoma. His company, Apprio, was recently tasked with killing time-wasting computer jobs at one of the nation’s largest hospital chains, with 10 Tampa Bay locations.
Durgampudi declines to disclose the name of the hospital, but a third-party report identifies it as having 52 hospitals nationwide. These Tampa Bay hospitals needed a terminator, a hired gun, to help its billing-and-claims employees use computers faster, or to remove the PC-related input altogether so the employees could move on to higher-level work requiring human intuition and a human set of eyes.
So the employees weren’t the problem, hospital officials concluded. It was the machines. Officials also concluded there must be something else to bridge the gap between human input and artificial intelligence and machine learning — something many organizations are searching for.
There was. There are “bots,” or automated programs that perform what were once human computer tasks. (A task is a step within an overall billing or revenue transaction.)
The hospital system found Durgampudi’s company, Apprio, a tech company Darryl Britt of Sarasota and Washington, D.C. founded in 1998. The software company focuses only on health care revenue departments. (The company is based in D.C. and Orlando; Durgampudi and the chief revenue officer live and work in Tampa.)
The Apprio team headed into Tampa Bay and the other locations for a showdown with the software of the hospital company’s billing departments. The team scouted out what the hospital employees had to go through to perform billing tasks every week, every month, every year.
They took their time, listening to people throughout the department, a listening tour to enhance the intuitive nature of their non-AI automation software. It took about a year of work done over 2.5 years, Durgampudi says. That time was, Durgampudi adds, a must to make sure the software had a deep understanding of its mission — expertise of standard claims, VA claims, authorizations, denial appeals, etc.
Over time, a number was discovered: The Apprio team thought 75 million billing-and-receiving tasks, or process steps, could be automated. With some astonishment, Durgampudi says that was only 10% of what might be automated throughout the hospital system. But it was revenue cycle management tasks they were assigned to dispatch.
A problem identified by Apprio and the billing departments was one assignment took 28 input steps for a human, from logging in to keystrokes, etc. The 28 steps took 33 minutes to input and process. And humans are not computers — mistakes or typos can show up in 10% to 20% of the work, Durgampudi says, meaning they had to be corrected, taking up more time.
Today, the hospital chain uses an automated program to execute the 28-step assignment in 1.5 minutes at its 10 Tampa Bay centers and its other 42 locations. And if the new program has problems, it kicks its query back to a human user. That’s why it’s not A.I., Durgampudi says, because it is “rules-based.”
In total, Apprio believes it saved the hospital system 476,000 hours annually — or the work of 150 people a year. The 75 millions tasks, or steps, amounted to 2.1 million transactions, says Durgampudi. A third-party audit in February confirmed the results.
The design, development and maintenance and associated tech can cost about $1 million for a revenue system that size, Durgampudi says. But they are bots, and they pay their way through savings.
The savings are a combination of things, Durgampudi says, noting the hospital system did not tell him what they saved on using the bots. But he estimates a similar system’s revenue is gained due to automations going after some cash humans did not have time to chase. That’s at least $2 million annually for a system the size of the client.
Then there are cash improvements, as automations fast-track reimbursements that would have been slower with people. That’s another several million dollars, he says. And there are the productivity improvements, easier to estimate when you have a headcount. Durgampudi figures the 150-person system earned $12 million in savings annually.
And “no one got laid off,” says Durgampudi, with some pride. And the employees could hardly believe they were freed from the tyranny of the old programs. They could now focus on higher-level work — tasks that required human eyes and intuition.
Despite the zero layoffs, Durgampudi says the hospital was able to slow or stop billing-related hiring. That was seen as a win, with the tight labor market.
Durgampudi says such tedious revenue-related computer work is commonplace in hospitals’ billing departments, and he estimates hospitals and their chains can cut $150 billion out of their U.S. costs by automating software. That’s 50% of the U.S. total of $300 billion, he says.
Durgampudi makes no apology for wiping that work from the job descriptions as it is so mundane, taking away from deeper, human-led examinations of revenue cycle management. It’s an inefficiency within the departments but it was the old way of doing things, and it replaced a lengthier process. “Nobody should be doing this work,” says Durgampudi, adding Apprio installed 100 automations for this project.
Bots and scripts, of course, are not new tools in health care and tech. These types of automations are noticed by the average U.S. consumer in obviously computer-generated emails and texts from doctors, pharmacies, hospitals and dentists. For example, instead of using humans to call people to remind patients of appointments, automatic texts or emails go out.
But don’t worry. The computer behind that text is so far not AI. Yet.
“We are definitely headed to that second phase,” says Durgampudi.
