For many hospitals and health systems, artificial intelligence is another hyped technology that one day may find its way into an information technology budget and be of some use.
But for one 335-bed community hospital located in St. Augustine, Florida, a $75,000 pilot project for artificial intelligence has already saved the hospital nearly $850,000 in unnecessary costs and holds real potential to offer savings of as much as $20 million over the next three years.
“We are also seeing positive results for better quality of care and outcomes,” says Michael Sanders, a physician, computer scientist and chief medical informatics officer for Flagler Hospital, an institution that first opened its doors in 1889. “I’ve been a physician for 30 years and before that I was a computer scientist and I’ve never seen anything like this.”
Artificial intelligence, a branch of computer science dealing with the simulation of intelligent behavior in computers, gets talked about a lot in healthcare these days because of its potential to better manage clinical and outpatient services, speed up and improve a diagnosis, better predict patient outcome and rapidly scan images and medical records, among other uses.
At Flagler, Sanders and the hospital’s information technology staff began a pilot project using artificial intelligence tools from Ayasdi Inc., a machine intelligence development software company based in Menlo Park, California, to improve the treatment of pneumonia, sepsis (a potentially life-threatening complication of an infection) and other high-cost medical conditions that providers call “high morality” or a higher death rate.
Specifically, the hospital wanted to use artificial intelligence to scan hundreds of thousands of electrical medical records from the institution’s system from Allscripts Healthcare Solutions Inc. and other billing and administrative data and analyze patterns for care, length of stay and patient outcomes.
Rather than have a piecemeal approach to how the hospital and an array of clinicians treated cases of pneumonia and sepsis, Sanders wanted to use artificial intelligence to develop specific universal clinic treatments and administrative procedures, or pathways, that doctors, nurses and other personnel could use the same way.
“The idea was to let the data guide us,” Sanders says. “Our ability to rapidly construct clinical pathways based on our own data and measure adherence by our staff to those standards provides us with the opportunity to deliver better care at a lower cost to our patients.”
The artificial intelligence pilot project took nine weeks, and Sanders and his staff used artificial intelligence tools to study volumes of medical and claims records dating back five years. The pilot delivered some surprising, and expensive, results, Sanders says.
“We looked at things such as the frequency with which we were ordering complete blood count (CBC) tests and for some patients with certain conditions we were doing it every day,” he says. “That’s hugely expensive and how frequently this was done was pretty eye opening.”
Using artificial intelligence tools to rapidly machine-read medical and claims data and see patterns, Sanders also looked at other variables such as patient length of stay, the type of medication—such as antibiotics—doctors were prescribing and when and why physicians were ordering CT scans.
When the data was fully analyzed, Sanders next used the artificial intelligence tools to build guidelines and standards into the Allscripts electronic records that doctors could use to more universally treat patients.
When implemented, the first pathways for better and more universal treatment of pneumonia resulted in $1,356 in administrative savings for a typical patient hospital stay and cut two days off the time patients were in the facility.
All in all, the administrative savings reduced costs by $800,000 in treating pneumonia patients. Now Flagler and Sanders are using the same AI tools to develop similar pathways to see improved clinical care for sepsis and for other chronic conditions for diabetes, substance abuse, heart attacks and heart conditions and gastrointestinal among others.
Originally, Sanders had planned to use AI to take on 12 conditions, an objective that has since been expanded to one condition each month with one major goal of saving the hospital as much $20 million over the next three years.
“These aren’t pathways based on some ivory tower approach but based on our own data and the ability to reveal the best pathways with the best outcomes at the lowest costs,” Sanders says.
Keep up with latest coverage on digital healthcare by signing up for Internet Health Management News today.