One Minn. health care provider using AI to prioritize patients for COVID-19 shots
Go Deeper.
Create an account or log in to save stories.
Like this?
Thanks for liking this story! We have added it to a list of your favorite stories.
As the COVID-19 vaccine slowly becomes more widely available to people beyond health care workers, educators and long-term care residents, federal and state health leaders have set parameters around who is eligible to receive the vaccine next.
Right now, people 65 and older are eligible to be vaccinated in Minnesota. But that’s a large group of people — nearly a million in Minnesota alone — and for now, vaccine demand far outpaces the supply.
That has left health care providers tasked with administering the vaccine in a challenging spot, deciding how to prioritize who should get vaccinated first, within that group of eligible patients.
Most of the largest health care providers in Minnesota are using some sort of randomization to identify patients to be vaccinated within the eligible groups. Essentia Health, Allina Health, Health Partners and Mayo Clinic are all offering vaccines to patients who are 65 or 75 and older, and sub-prioritizing within those age groups by randomly identifying patients to make a vaccine appointment.
Turn Up Your Support
MPR News helps you turn down the noise and build shared understanding. Turn up your support for this public resource and keep trusted journalism accessible to all.
But South Dakota-based Sanford Health, which serves thousands of patients in western and northern Minnesota, has taken a different approach. Very early in the pandemic, before vaccines were even on the horizon, Sanford researchers dove deep into science to better understand who should be at the front of the system’s line, when the time came to vaccinate.
That early work is now being put into action, in the form of an algorithm that helps identify which of Sanford’s patients who are 65 and older are most at risk — and should be vaccinated first.
Using data to prioritize
When COVID-19 began spreading in China more than a year ago, a Sanford data analytics team started compiling available research on who was at greatest risk of having the most severe outcomes of the disease.
Sanford chief physician Jeremy Cauwels said that, as the virus has spread, the team has continued to expand the database it built, which now includes not only the latest research on COVID-19 risk factors, but data from the roughly 85,000 COVID-19 patients Sanford has treated across its system.
"With that 85,000 people, what we can do is take a real-time picture that evolves over time using computer learning, to tell us what patients or what people in the Midwest get the sickest from COVID-19,” Cauwels said. “And which ones spend more time in the hospital, which ones have the highest risk of death."
To prioritize vaccine doses, Sanford first narrows its pool of possible recipients to members of the groups state health officials say are eligible to be vaccinated — right now, that’s people 65 and older. Then they apply the computer algorithm to that group, to assign a patient’s level of risk for worse outcomes from COVID-19, based on a long list of health-related factors.
"If you were to just go with the three most most mathematically important for predicting whether you're going to do well or not [with a COVID-19 diagnoses], you could go just with age, obesity and kidney disease," said Cauwels.
These health conditions and others — including current or past cancer, diabetes or heart disease — are weighted based on the scientific data around how each worsens the outcome for COVID-19 patients.
While most major Minnesota health care providers are using some form of random choice to decide who within priority groups is first for a vaccine, Cauwels believes an artificial intelligence approach is more equitable than random choice to administer the vaccine.
“I think the reason for that is the difference between what we've seen here and what we've seen in South Florida, where people are lined up for blocks next to a hospital because they found out that doses are available,” said Cauwels. “Or you have people that are waiting hours on the phone in the Minneapolis area to try to get into a finite number of vaccines slots."
The task of prioritizing vaccine distribution is being complicated by scarce supply, desperately high demand and increased risk factors based on age, race, medical conditions and job exposure.
Deciding who gets a vaccine isn't just a medical decision. There are also ethical and moral choices, said Debra DeBruin, a University of Minnesota bioethicist and a member of the Minnesota Department of Health’s vaccine allocation advisory group.
"The ethics guidance that we have in the state of Minnesota is that we should prioritize individuals who are at highest risk of poor outcomes from COVID," she said.
That’s part of the Health Department’s framework for vaccine distribution, which is based on federal guidance from the Centers for Disease Control and Prevention and input from the state’s own advisory group.
But deciding which factors to use in assessing risk is where things get complicated.
The federal government allocates vaccine supply to the state. The state, in turn, distributes vaccine to providers — health care systems, clinics and public health departments, for example — across Minnesota.
The state’s vaccine allocation plan takes into account two factors in determining how much vaccine to send to different regions: The population of the targeted group in the region — for instance, people over 65 — and the region’s Social Vulnerability Index score.
That metric, which the CDC recommends using, assigns a number to areas through the lens of 15 social factors about the population, from poverty and unemployment to race and ethnicity.
The state Health Department didn’t respond to questions about that assessment process, so it’s unclear how MDH uses that index to distribute vaccine doses.
‘You have to take into account disparities’
Distributing vaccines across the state is one decision. Prioritizing who gets a shot first presents a whole new set of questions, said DeBruin.
“You have to take into account disparities,” she said.
There are notable disparities in COVID-19 outcomes related to patients’ race and ethnicity. When adjusted for age, people of color and Indigenous people in Minnesota die of COVID-19 at significantly higher rates than their white counterparts.
Those communities statistically have higher rates of chronic disease, and less access to health care.
Sanford has chosen not to use race or ethnicity as a factor in its algorithm. By focusing on medical risk factors, Cauwels said, patients who are members of racial and ethnic groups with higher rates of chronic disease will be elevated in Sanford’s risk prioritization.
“We have not used ethnicity, because the medical literature at this point is not clear if being Hispanic causes you to have poor outcomes, or if it's because of the place you live, the diseases you have,” he explained.
In Wisconsin, the University of Wisconsin-Madison’s medical center used an algorithm based on race to prioritize vaccines for employees in its first round of distribution, said Shiva Bidar-Sielaff, chief diversity officer at UW Health.
“It's incredibly important to realize that all data points to the fact that, unfortunately, race and ethnicity have been shown to create a much higher risk of hospitalization and death for COVID-19,” she said.
“So when we looked at our algorithm, we saw that if you add age and SVI, which has that component of race and ethnicity, it's a multiplier effect in how much higher-risk an individual is at for hospitalization and death.”
Bidar-Sielaff said UW Health is not applying the algorithm as it distributes vaccines to patients in the 65-plus age group.
“What we did for our 65-plus population is we actually directly used race, and prioritized sending invitations to our 65-plus patients that are Black, Latinx and Native American to receive the vaccine,” she said.
UW Health’s vaccine advisory group, made up of physicians and bioethicists, is still discussing which metrics to use as distribution of vaccines moves to other age groups or categories of patients.
While race correlates strongly with more severe COVID-19 cases and increased risk of death from the disease, DeBruin is hesitant to embrace it as a unique factor, because prioritizing health conditions will also prioritize racial and ethnic groups with higher rates of chronic diseases.
“So if you can prioritize people based on those underlying health conditions — risk factors — you can address, I think, some of those health disparities,” she said.
Another important factor to consider, she said, is patients’ jobs — particularly if they are essential workers at higher risk of exposure to COVID-19, whether directly or through regular contact with the general public.
“As it turns out, prioritizing based on these work roles helps to promote equity, too, because many of these jobs are low-wage positions, so there are issues regarding socioeconomic status, race and ethnicity,” she said.
Many health care systems are constantly evaluating the best way to get the most impact from scarce vaccines, and the process is likely to evolve going forward as vaccine supply increases, said DeBruin, who adds some bumps in the road are to be expected as the state and health care providers navigate a pandemic.
The ethical way to vaccinate
As health systems work through the complex choices of deciding how to effectively and ethically distribute what can be a life-saving vaccine, they must consider individual patients’ needs alongside the broader public health impacts of vaccine prioritization.
DeBruin said the most ethical process is to base those choices on the best available science.
"If there is a really well designed scientifically sound algorithm, the best thing would be if that can be shared across the state,” she said. “If we could adopt an approach like that consistently across the state, then you wouldn't have those inconsistencies between systems."
State health leaders have said repeatedly in recent weeks that the state Health Department provides guidance to its partner providers, but ultimately, the choices about vaccine priorities within eligible groups are up to the providers administering the vaccine to the public.
Cauwels said Sanford has offered its algorithm to officials in Minnesota and other states where it operates.
A spokesperson for the Minnesota Department of Health could not confirm details of such an offer, but the agency’s infectious disease director, Kris Ehresmann, said the tool could be useful on a broader scale.
“Sanford’s tool sounds like one that other health systems could benefit from,” Ehresmann said in a statement. “We appreciate Sanford’s commitment to immunize for impact. They are doing a thoughtful job.”
Immunizing for impact means efficiently getting vaccines to those at greatest risk from COVID-19, a key part of the state’s framework for vaccine distribution.
But when private health systems take on the bulk of on-the-ground vaccine distribution, it runs the risk of allowing some of the state’s most vulnerable patients to fall through the cracks: people who might not be part of the health care system.
Many of the people in Minnesota’s most vulnerable populations don't have a regular doctor or get regular health care.
Sanford has tried to address that gap with a website where anyone can sign up to be part of the patient database from which the system prioritizes vaccine distribution. Officials said nearly 90,000 people have signed up, providing basic medical and personal information, since vaccines became available.
"We use as much as we know about their medical information to put them on the list, right in between the people that we've known about for 20 years, to make sure that we're doing the best thing we can to distribute equitably between all populations," said Cauwels.
That still leaves out a segment of the population without access to technology or the skills to use it.
Otherwise, people who are eligible to get the vaccine in Minnesota, but are unaffiliated with the major health systems, community clinics, pharmacies, employers or local public health departments that are distributing it, have one primary avenue available for connecting to it: The state’s vaccine lottery system.
Since January, the state Health Department has reserved a fraction of its weekly allotment of doses to be randomly administered to people over 65, regardless of health conditions.
During a series of sign-up windows over the past few weeks, people 65 or older were able to register with the state for a chance to be randomly selected for vaccination in a two-week pilot program. The program has since evolved — and registration is closed, for the time being — into three vaccine clinics in the state’s largest population centers.
DeBruin said there is a place for lotteries, or random choice, to decide who gets a vaccine, but only if everyone in the lottery group is at the same risk from the virus.
“A lottery across groups that are not similarly situated in terms of risk undermines the moral recommendation that we try to protect people who are most vulnerable, because if you're running a lottery against people at varying risks, then people who are at lower risk have just as much chance at getting access to vaccine as people who are at much, much higher risk,” she said.
“But the very best thing to do, from an ethics point of view, is to allocate based on risk.”
As she looks ahead to the possibility of increased vaccine availability and more widespread distribution, DeBruin said her greatest fear is that organizations will default to speed as their top priority, a rush to just get as many shots in arms as quickly as possible.
"People who are already privileged and who already have great access to care will get vaccinated,” she said. “And people who are not privileged and do not have great access to care will be left behind."
COVID-19 in Minnesota
Data in these graphs are based on the Minnesota Department of Health's cumulative totals released at 11 a.m. daily. You can find more detailed statistics on COVID-19 at the Health Department website.
The coronavirus is transmitted through respiratory droplets, coughs and sneezes, similar to the way the flu can spread.