By Megan Headley
As COVID-19 infections spread across Minnesota in March 2020, M Health Fairview—a healthcare company made up of the University of Minnesota, University of Minnesota Physicians, and Fairview Health Services—made the decision to dedicate its M Health Fairview Bethesda Hospital in Saint Paul, Minnesota, to serving COVID-19 patients. However, providing the rapid care needed for these patients became complicated when individuals were transferred to Bethesda from one of the 11 other hospitals in the health system.
Like many health systems developed through mergers and acquisitions, M Health Fairview sported consistent branding across what had previously been the Fairview Health and HealthEast Care systems, but it had a disparate array of back-end medical record systems behind the scenes.
“In the normal day-to-day kind of interaction with the patient, that’s really invisible to the patient. Most patients get their treatment within one hospital or you continue to see your same specialists, or if you move across systems over a period of time, we can make that really invisible to you behind the scenes,” says Eric Murray, director of IT data solutions and technology at M Health Fairview. “Then along came COVID.”
Once the pandemic hit, patients might be admitted in one hospital and have their vital information recorded, only to be transported across town once they were diagnosed with COVID-19. The health system struggled with slow diagnoses and was wasting time on tracking down and reentering vital information.
“We had a problem around how we could best treat these patients from a system perspective,” Murray explains. “There really was this need for this ability to have a longitudinal view of a patient record regardless of where that patient came from.”
An effective patient matching solution
One of the challenges with patient matching, Murray explains, is creating effective rules for matching records. “The fact that the name and address and birthday are the same doesn’t necessarily mean that they’re the same patients,” Murray says. “We found out we were either undermatching or overmatching based on the information that we had.”
The IT team worked with technology provider Verato to improve its patient data algorithms. Verato relies upon referential matching, a technology that matches patient records to a comprehensive reference database rather than simply comparing the data within two patient records. This approach, according to Verato, better overcomes the challenges of working with records that are outdated or contain errors.
“It really allowed us to get much better and much more accurate how we matched and merged the patients,” Murray says.
This optimized patient matching technology addressed two challenges that tend to come up with patient matching. First, it removed the need for time-consuming human interventions to match data. Second, it removed the risk of undermatching that comes from applying rudimentary matching rules. As Murray puts it, “If you’re undermatching patients, you’re not providing them the best possible treatment because they’re showing up at another hospital and their vitals aren’t there.”
How patient matching drives faster diagnoses
Murray’s team was already working toward streamlining medical record access across the health system when the pandemic accelerated the need for improved connectivity. Within two weeks of converting Bethesda Hospital for COVID-19 response, M Health Fairview had accordingly shifted the objectives for its patient matching technology.
Speedy diagnosis and treatment were essential. As the IT team worked to support faster record sharing, University of Minnesota researchers were focusing on making more rapid COVID-19 diagnoses. “Back when the COVID pandemic was just starting, waiting for bloodwork and things took a long time—often more than a week,” Murray says. “That made it really difficult for us to treat patients.”
However, university researchers were beginning to develop an artificial intelligence algorithm that could identify patterns in the chest x-rays taken from patients presenting with COVID-19 symptoms. The algorithm could thereby rapidly predict the patient’s likelihood of having COVID-19. Within seconds, the health system’s Cognitive Computing platform could pull an x-ray, run the algorithm, and enter a predictive diagnosis into the EHR software that was, by this point, integrated across all its hospital locations.
“The reason that we were able to do that was because we could create this longitudinal view of a patient regardless of where their data came from,” Murray explains. “It could come from one of our EHR systems, imaging system, our lab system, or the telehealth systems we were setting up.”
Building on a data-based foundation
The game-changing success of the COVID-19 diagnosis tool helped demonstrate the usefulness of the patient matching system and has driven the health system to explore additional use cases.
For example, the technology now supports Fairview’s goal of improving its Net Promoter Score, which measures patient satisfaction. When a patient calls to book an appointment, receptionists can access patient records across the health system’s 32 locations, giving visibility into where that patient has been seen in the past and what their treatment preferences are. The result is a streamlined process that works to boost patient satisfaction from the first interaction.
The technology is also proving useful in business operations. Based upon the COVID-19 use case, where hospitals were linked with labs to speed the transmission of test results, the IT team is expanding this connectivity across facility types. For example, by using patient matching to support prescription drug monitoring, a Fairview clinician can rapidly verify whether an existing prescription holder is the same person as the patient they’re currently seeing.
“That typically would have gone into a work queue that would cost the company money to work down, and also potentially delay getting that prescription to the patient. Our ability to put this technology in the middle has reduced that workflow,” Murray explains.
The work done in diagnosing COVID-19 patients has also built a strong foundation for using patient matching technology to identify patterns in other conditions.
“We’re able to use this foundation and apply the same methodology, including the patient matching, across multiple systems to several different disease states,” Murray says. “So when we look at sepsis, for example, we utilize this success story and these lessons learned to build this concept of a learning health system where you take data, test and validate models, put them into practice, and then continue to monitor to make them better and smarter and faster.”