Skip to contents
Health

Paging Dr. Algorithm: How AI and other new tech are changing care in Colorado hospitals

UCHealth and HealthONE use advanced algorithms, remote monitoring to chase disease and help keep hospital patients safe

Safety View Technician Mekdes Daniel watches a series of patient cameras at UCHealth administrative offices on Sept. 25, 2019, in Aurora. (Seth McConnell, Special to the Colorado Sun)
  • Credibility:

A wobbly leg emerges from under a blanket and makes a tentative search for the polished floor below. The movement catches the eye of Lupita Rodriguez, who’s responsible for keeping this patient safe.

Rodriguez leans forward to talk to her patient, who immediately stops at the sound of the authoritative voice. “Mrs. Smith? Please call your nurse before you get up. Thank you.” 

Mrs. Smith, a bit confused and definitely unsteady on her feet amid her hospital stay, nods and sits back.

Rodriguez lets go of the talk button. She then spins her gaze across the 11 other TV screens arrayed in front of her, looking for the next looming risk of a traumatic and very expensive hospital room fall.

Rodriguez, as it turns out, is not sitting in the room with Mrs. Smith or any of the other 11 patients in her charge — she’s watching a dozen people at various hospitals in the UCHealth system from an Aurora office park near the Panera Bread and the Walgreens, with the patients often as far as 150 miles away. 

Three attendants like Rodriguez, and the bank of 36 TV screens, replace what until recently had to be 36 human monitors, one stationed in each room where a delirious or uncooperative patient might topple over getting out of bed or heading to the bathroom. The digitalsitters room is just one of the innovation centers run by UCHealth at the office on South Peoria Street, part of an expansive effort to put remote monitoring, video interaction, machine learning and powerful algorithms deeper into direct patient care. 

Other major hospital systems are laying down the same digital paths, with HealthONE/HCA, the largest hospital provider in metro Denver, using algorithm-based alerts and remote camera monitoring of vulnerable patients, among other technologies. Hospitals and large physician practices are also using machine learning algorithms to handle bureaucratic challenges, such as predicting optimal staffing levels for emergency departments or what financial arrangement is most likely to prompt patients to pay bills.

Privacy and public perception are two of the challenges facing hospital systems as they rapidly implement the new technologies. Patients and government overseers want to know when all the promises of innovation leading to savings will actually cut bills. Hospitals currently take up the largest share of health spending, at $1 out of every $3, and many in Colorado boast high profit margins

Launching new data initiatives for the good of patient care, meanwhile, makes the same data more vulnerable to the bad. Technology can already do a lot that patients and their human doctors may not be quite ready for, said Dr. Chris Davis, a former ER physician who now is medical director of UCHealth’s Virtual Health Center. So it’s a delicate navigation.

Genetic testing or advanced medical imaging, for example, might be able to warn people far in advance of a future malady — some may not want to know, others may not have the resources to do anything about it. Consumer advocates, meanwhile, worry that insurance companies or employers might use negative medical data to penalize individuals

“The privacy concern is very real,” Davis said. “We want to stay on the right side of the line between cool and creepy.”

Advanced algorithms trained to fight sepsis

The digital-sitting room for video monitoring of patients in danger of falling is one of the less controversial innovations and clearly represents a savings for UCHealth over hiring individual human monitors in every room. And the hospital has found remote monitoring to be just as effective as in-room minders, in part because a human team is dispatched at the same time the monitor is speaking to the patient on the speaker. In instances when a patient would be unresponsive to voice commands or questions, the video monitoring can be switched off and a person posted in the room. 

“This one’s a no-brainer,” Davis said. 

Registered nurse Adrian Ugarte looks at a patient’s vital signs in the Virtual ICU at UCHealth administrative offices in Aurora, Colorado. UCHealth systems are using an innovative new system of high tech monitoring to care for patients to help avoid falls in hospital rooms and detect sepsis earlier. (Seth McConnell, Special to the Colorado Sun)

This fall will mark a giant move forward for the innovations lab, which is expanding an algorithm-based attack on deadly hospital sepsis. 

Antibiotics are highly effective against hospital-acquired sepsis, but every minute that goes by before administering them raises the risk. An algorithm culling a complex set of screening criteria has cut sepsis treatment time by more than two hours for some UCHealth patients, already saving lives even before systemwide implementation.

“What’s hard about sepsis is the screening criteria are highly sensitive but not specific,” said Dr. Richard Zane, chair of Emergency Medicine and chief innovation officer for UCHealth. For example, Zane said, the early signs of sepsis are so general — higher heart rate, faster breathing, higher temperature, “if I caught you at the end of a workout, you’d meet the criteria for sepsis, with the glaring negative that you have no reason to be septic.”

The new protocol monitors vital signs and lab results, while adding in the patient’s recent history and comparing it to thousands of past cases. “So, an algorithm that maintains the sensitivity, but increases the specificity,” Zane said. “It’s a game changer.”

The data crunching combines with increased teamwork at UCHealth to combat sepsis faster. The inpatient pharmacies, for example, have sped up how they fill orders related to sepsis and get them to treatment units.

Over the past two years, UCHealth hospitals have reduced mortality from sepsis by about 14%, saving the lives of an additional 225 patients, the hospital said. Faster treatment also improves recovery and cuts the length of hospital stay, significantly cutting costs.  

At HealthONE/HCA’s regional hospitals, recent technology innovations also include a sepsis-related monitoring system. The protocol is based in part on learning derived from millions of past medical records that give indicators on when a hospital patient started “decompensating” — getting rapidly worse instead of better. 

The HCA system constantly monitors six lab results and six vital signs for a hospital patient and alerts the mobile devices of care team members– from nurses to supply chain managers to hospital CEOs — when thresholds are breached. 

HealthONE Chief Information Officer Andy Draper likened it to signing up for airline schedule changes. “If they don’t ping you that your flight is delayed, there’s no value in that.”

The patient-facing technology may look shiny and new, Draper said, but it’s based on years of back-shop work that has taken a long time to pay off. HCA had to create a massive data warehouse to reconcile and organize patient records from 184 hospitals nationwide, cleaning up differences in phraseology or numerical values so that researchers had comparable data. (HealthONE/HCA hospitals in the region include Presbyterian/St. Luke’s, Rose, SkyRidge, Medical Center of Aurora, Swedish and others.)

“You have to get physicians involved and get everyone to agree to the data and what is true and what is not true,” Draper said, which took HCA most of eight years. Then researchers started building data tools and algorithms to search across the records and find patterns and answers. 

Another addition to the HCA system is machine-learning assisted reading of radiology records “in the background” of acute medical care. For example, a patient from a car accident might have a full-body CT scan for internal injuries. A reading for acute injuries is negative. But an algorithm informed by natural-language processing — how a system like Alexa or Siri analyzes the way humans speak and turns it into data the system can process — discovers a description of a potential lung tumor in the scan. A case navigator follows up with the patient and schedules an oncology appointment.

“Are we replacing nurses with computers?” Draper observes. “Absolutely not. Are we making them more efficient by giving them alerts when a patient is deteriorating? Yes, and that’s the right thing to do. Are we eliminating radiologists? No. Are we giving them targets to look at? Yes.” 

Alexa! Am I experiencing congestive heart failure?

Zane, Davis and other UCHealth leaders are also taking deep dives into the possibilities of mobile technology and the “internet of things,” where even the most everyday device can be equipped cheaply with a form of communication. 

Diagnostics and algorithms that play in the background can continually synthesize information from a patient and the patient’s surroundings to give out alerts to a care team before things get unhealthy and expensive. 

Zane offered two examples. In one, the care team of a chronic asthma or emphysema patient has given them an inhaler with a monitoring device and transmitter. 

“If you’re using your inhaler more than two days in a row, you are likely going to have an acute incident that requires acute intervention,” Zane said. “Or if you’re not symptomatic yet, and you are using it five or six times a day when you usually use it four, that means something.”

In another case, a patient with a history of heart failure wears a connected step monitor at home. 

“If you average 8,000 steps a day and today you only took 5,000, or if you usually go to Starbucks once a day but today you didn’t, does that mean something?” Zane said. “I think it does. You get an alert, ‘Hey Mary, are you feeling OK?’ She texts, ‘No, I don’t feel so great today.’ And I can intervene before Mary has to come to the ER.”

Some of the technology, hospital technologists said, is ahead of public education and acceptance. The line between cool and creepy may vary by generation or circumstance — it’s creepy, for example, if Starbucks is the one sending you the message “We missed you today — go get your heart checked.” 

And the technology doesn’t always work as promised — health care is not the first industry to deal with the “Where’s my jetpack?” question.

Dr. Chris Davis, medical director of Virtual Health, works on a patients file in Virtual Surveillance at UCHealth administrative offices in Aurora, Colorado. (Seth McConnell, Special to the Colorado Sun)

Draper, 51, said he was a big fan of the futurist bible “Megatrends” when it came out in the early ’80s. 

“The intersection of big data and medicine because of computing power — we’re right in it. But we’re right in the very beginning of it,” he said. “There are a lot of tools that will pop up and we should embrace and love them all, and then over time we’ll see what their real potential is.” 

The evolution in technology also changes the demands on health care personnel. It can replace them, but technology is also providing new opportunity. Lupita Rodriguez, 34, commuted from Parker to UCHealth’s Colorado Springs Memorial Hospital to work as a certified nursing assistant, then found out online about the monitoring lab. Many current employees in the virtual health center have some form of health training, but it’s not required — training comes with the job. 

“I love technology,” Rodriguez said. “It’s not direct patient care, but it’s amazing how you can still make a difference using technology.” 

Some patients have been appearing on her monitors for close to a year. “We can develop a connection, and some recognize our voice,” she said.

Hospital patients these days are more intensively sick than in years past — anyone able enough is sent home or to a lower level of care. Some patients will die. 

Ashley Oakes sits in another non-video monitoring room at UCHealth’s Aurora office, looking at vital signs for up to 42 patients at a time. If a patient flatlines and a nurse is not in their room, Oakes gets on a hotline to help launch the response. 

“Being the eyes and ears for the nurses, you’re making an impact on their care and that’s rewarding,” said Oakes, who trained as an EMT and is now in a boot camp for data analysis in addition to her UCHealth job. When a patient’s vitals do crash, though, Oakes appreciates when one of the hospital nurses calls her to relate the outcome.

“My heart goes out, and I definitely want to know what happened,” she said. “I know they’re in the best care they can be because we were watching them.”


The Colorado Sun has no paywall, meaning readers do not have to pay to access stories. We believe vital information needs to be seen by the people impacted, whether it’s a public health crisis, investigative reporting or keeping lawmakers accountable.

This reporting depends on support from readers like you. For just $5/month, you can invest in an informed community.