A medical algorithm capable of real-time monitoring of electronic health records can predict whether a hospital patient will deteriorate hours in advance and alert physicians about their decline.
The algorithm, developed at the University of Chicago Medical Center and marketed by Quant HC, tracks 28 variables like respiratory rate, age, kidney function and length of hospital stay and provides a timeline of the patient’s risk.
It detects about 88% of patients who are at risk of deteriorating from organ failure, cardiac arrest, sepsis and other life-threatening conditions within a median 33-hour timeframe, improving clinical decision-making and resource management, says Rick Halton, co-founder and chief marketing officer of Apervita, an online marketplace and platform for more than 1,000 health algorithms, including Quant HC.
“This has a huge potential impact on reducing length of stay in hospital, because you’re catching that patient really early” instead of after their condition has deteriorated and they’ve been moved to intensive care, he adds. “It’s a lot more cost to the hospital, and clearly the patient’s life is at risk.”
What else is out there?
Increasingly, medical algorithms are working behind the scenes to aid in diagnosis, treatment, short- and long-term prognosis, care management, and claims processing. There are algorithms for managing chronic disease and for predicting frailty in patients, a major cause of rehospitalization, and ones that detect insufficient information, such as the need for a vaccination.
Vivify Health’s platform has close to 100 customized algorithms developed from evidence-based best practices across 600 hospitals and payers in the U.S. The algorithms drive patient-facing applications that improve care management after a patient leaves the hospital. Patients access it on their own mobile device or a kit provided by the firm.
The algorithms include personal data, video content, educational materials, and specific details like from input from a pharmacist on why a patient should continue taking a particular medication. They also include best practices for managing wide-ranging conditions, from congestive heart to liver disease and diabetes.
Through the algorithm, patients receive daily reminders to check their blood pressure or perform other health-related activities, explains Eric Rock, Vivify’s founder and CEO. More than 25 wireless devices for performing those tasks are integrated into the platform.
The algorithm also asks the patient a series of questions each day. When the patient’s condition or parameters or baselines change, so does their algorithm. Major changes to the algorithm require a physician’s okay.
As customers join the platform, they add their best practices and those are then incorporated into standardized content for all customers to use, Rock says. “It’s sort of like crowdsourcing, but with the best brands across the U.S. from a provider perspective.”
Medical algorithms yielding positive results
Hospitals using the remote care program are seeing over 50% reduction in acute utilization, Rock says. For example, Memorial Hermann in Houston reduced readmissions from 17% to just 5% after launching the program in 2013. The hospital also cut home health visits by 3.6 per episode and reduced home health length of stays from 82 to 48 days. The savings per patient totaled $8,500.
Some algorithms are so precise and so predictive that hospitals and providers need new care protocols to deal with them, because humans can’t see all of the subtle changes algorithm can pick up, Halton says.
However, older-generation algorithms that are still in use may not be as predictive, generating misinformation, such as false positives.
That concerns John Meigs, president-elect of the American Academy of Family Physicians. “The hope with an algorithm is that it will help standardize care and reduce errors, but sometimes the science may not be there to support the recommendations that you see,” he says. “And a lot of times an algorithm can oversimplify a condition.”
Because patients are different, they have different co-morbidities, different underlying conditions, different lifestyles and cultures and values, and different personal approaches to care, Meigs adds. “So it’s hard … You can’t always put a round peg in a square hole.”
For that reason, a lot of physicians are uncomfortable with strict algorithms, Halton says.
Teaming to succeed
Still, most physicians and administrators today realize the potential for algorithms to improve patient care. Ideally, the two strategies — physicians and algorithms — are complementary. For example, a consumer-facing algorithm that relies on current guidelines and presents information in plain language may help a person know when to seek input from their physicians.
But such algorithms, often an app or available using a search engine, should not take the place of a trusted doctor-patient relationship, the American Medical Association says. The group recommends that patients share information about algorithms or apps they are using with their physicians.
Another challenge, with physician-facing algorithms, is the risk that multiple algorithm-based alerts may pop up, leading to alert fatigue and clinicians missing alerts that provide new information or signal a real crisis.
Algorithms are also being used at the back end of care to improve financial results and ensure that physicians are applying various protocols appropriately. For example, General Electric is commercializing an algorithm that sifts through medical claims and flags those that are likely to be denied by the payer. It also does a root cause analysis of why the claim may be denied and looks at trend analysis across many claims.
Doing the math for population health management
There are also implications for population health. “The growth in remote monitoring 10-15 years ago used clunky devices, and we didn’t have effective apps either until recently,” says Rock. Today, “you put all those components together and there are solutions out there for care coordination and analytics that are all about population health.”
The missing component, what algorithms like Vivify offer, is engaging patients of any age and technical ability to use them. “For those using those population health tools, the data is automated, the algorithms are automated,” he adds.
It’s not just about big data, although algorithms are typically generated using big data, says Halton. “When you’re applying it to a patient or a population, it’s really small data that you’re putting into practice.”
At the end of the day, algorithms help save physicians, clinicians, and administrators the chore of having to search through reams of data and manually process it, helping to ensure a better outcome for the patient and the hospital.