Image by Getty Images
Image by Getty Images

COVID-19 spurs hospital tech

Rapid advances have been made in virtual rounds, chatbots, machine learning, and more.

As soon as the first U.S. patient to be hospitalized for COVID-19 was admitted to Providence Regional Medical Center in Everett, Wash., on Jan. 20, the hospital's technology team sprang into action. By the end of the next day, the centralized electronic medical record (EMR) had been updated to include a dedicated COVID-19 screening tool accessible to all 51 hospitals and 1,000 clinics throughout the health system.

“We had a disaster preparedness plan in place, a unified EMR, and we were already doing telehealth and virtual visits at all of our hospitals,” said Todd Czartoski, MD, chief medical technology officer for Providence Health, which operates in Alaska, Washington, Montana, Oregon, California, New Mexico, and Texas. “Prior outbreaks such as Ebola, MERS, and SARS had prepared us to handle this type of event.”

Having an established telehealth system—along with related tools like handheld devices, home monitoring software, and chatbots—was a huge plus in dealing with COVID-19, said Dr. Czartoski.

Other hospitals similarly picked up their adoption of existing virtual technology. “Prior to COVID, we were doing very few virtual ambulatory visits (video or phone) in a typical week,” said Craig Umscheid, MD, MS, FACP, chief quality and innovation officer and director of the Center for Healthcare Delivery Science and Innovation at University of Chicago Medicine. “With the emergence of COVID, virtual visits went from about zero in mid-March to more than 75,000 as of mid-July.”

Although many hospitals had invested in telemedicine in recent years, the pandemic forced them to implement and scale digital tools and virtual processes at a much faster pace than in the past, said Benji Mathews, MD, FACP, associate professor of medicine at the University of Minnesota and chief of hospital medicine at Regions Hospital, HealthPartners, in St. Paul, Minn.

“We're setting up virtual services in the smallest platforms possible, including iPhones and iPads,” he said. “We needed to create tools that are movement oriented, easily deployable, and simple to use for everyone.”

Video tools

When the virus struck, virtual visits became a way to limit physical contact between staff and infected patients within the hospital, said Dr. Czartoski. At Providence, clinicians use a mobile cart with a camera and computer to connect with patients from anywhere in the hospital, while patients log in on an iPad from their rooms.

Despite the physical distance, many patients say the system makes them feel more connected to their clinicians, he noted.

“Using iPads has helped overcome the sense of isolation patients feel when everyone who comes into their room is completely covered and masked,” he said. “Just being able to see someone's face and have a normal conversation on video tends to forge more of a connection than having providers come in all gowned up.”

The University of California, Los Angeles (UCLA), Medical Center was already using iPads in rooms before the pandemic, said Michael Pfeffer, MD, FACP, assistant vice chancellor and chief information officer for UCLA Health Sciences. In response to higher use during the pandemic, the hospital tagged iPads to rooms so patients would not have to set up accounts or log in to communicate with clinicians and hospital staff could easily and securely video chat with patients using their room numbers. Patients can also use the devices to initiate external video calls.

“A provider can make a video chat request to a patient in the hospital and the patient can accept or reject the request,” he said. “We've scaled that same process to local skilled nursing facilities for our patients to video chat with our providers.”

Video proved useful for education as well, noted Dr. Mathews. At Regions Hospital, medical students and residents round virtually. “We want students to learn about COVID-19, but we are also in a situation with limited personal protective equipment and rules around restricting exposure of medical students to COVID-19 patients,” he said. “As a compromise, we use telemedicine to allow residents and students to connect with patients remotely.”

During virtual rounds, one member of the care team enters the patient's room and connects via iPad with the rest of the team located in another area of the hospital. The remote clinician can even hear the sound of the patient's heartbeat via a digital stethoscope connected to the computer's speakers. A decision tree helps clinicians determine when virtual visits are appropriate, for example, not in cases of severe cognitive impairment, hemodynamic instability, or goals-of-care conversations.

“With telehealth, I'm able to provide supervision to a student or resident without seeing certain patients in person,” Dr. Mathews said. “It also helps with efficiency because I can easily check in with patients multiple times a day remotely versus spending a lot of time walking around the hospital to physically visit each room.”

Regions Hospital offers teleguidance to hospitalists in performing point-of-care-ultrasound, said Dr. Mathews, who also serves as ultrasound director in hospital medicine at HealthPartners and published a study on the applications for ultrasound in COVID-19 in the June Journal of Hospital Medicine. Cloud-based secure platforms allow instructors based anywhere to look into a clinician's handheld device at the bedside and offer feedback on what they're seeing or change the image settings when necessary.

“Teleguidance allows me to remotely log onto my phone or laptop, and help out instantly,” said Dr. Mathews. “As a result, we can perform a test and interpret the results within minutes.”

Extending outside the hospital

Regions Hospital expanded its use of telehealth to patients who've been treated and discharged for COVID-19, said Dr. Mathews. Patients receive remote devices, such as oximeters, at discharge to be used at home for self-monitoring. The following day, they receive a call from the hospitalist on duty to check on how they're feeling and answer any questions.

Dr. Mathews and others note that the federal government's move to waive restrictions around billing for remote visits was key to the rapid adoption of telehealth during the pandemic—something they hope will continue. In the early days of COVID-19, CMS issued temporary rules allowing clinicians to conduct remote video and audio visits, even across state lines, and get reimbursed at the same rates as in-person services.

As a result, more patients were able to self-monitor symptoms at home, avoiding visits to the ED and urgent care facilities. Hospitals were also able to provide virtual hospitalist services in other states during regional surges and staffing shortages.

“The changes made to reimbursement were very helpful to accelerating video visits—and we'll see a decline if those restrictions go back into place. . . . There's no doubt that the video visit will play a role for all kinds of patients in future if they continue to be reimbursed,” said Dr. Pfeffer. “However, a lot of research needs to be done on outcomes related to video visits, as well as when they are most appropriate and some proven techniques.”

Another challenge for hospitals is deciding how much to invest in home monitoring technology, said Dr. Czartoski. Given the uncertainty around reimbursement and the need for scalability, Providence opted for a $25 pulse oximeter and $5 digital thermometer—versus more costly Bluetooth-enabled tools linked to the EMR—to distribute to patients at discharge.

The hospital also invested in software integrated into its EMR that would allow it to scale tele-ICU services, said Dr. Czartoski. The software enables ICU staff to remotely monitor up to thousands of COVID-19 patients discharged to home quarantine by automatically sending secure texts three times a day asking about their symptoms and current temperature, oxygen level, and pulse.

“Those questions then guide our concern level about a patient and are put into an electronic dashboard that nurses monitor,” he explained. “Patients get a green, yellow, or red circle based on their answers, with red signaling the need for a call or video visit or possible admission coordinated through the ED. The sickest patients are handed off to our ED to safely get a higher level of care so they're not coming in on their own.”

As of June, the hospital had used the tele-ICU system to monitor almost 4,000 patients across five states, said Dr. Czartoski. Although data are still being collected and analyzed, it appears that using remote technology helped avoid overcrowding in the ED—the hospital never became overwhelmed or ran out of beds, he said—while still ensuring that patients received necessary care.

Chatbots and screens

Chatbots are another way that hospitals have been leveraging existing technology during the crisis. For example, when the virus struck, Providence fast-tracked an existing chatbot symptom checker that it had developed in partnership with Microsoft Corp., equipping it with a COVID-19-specific protocol linked to its virtual express care service.

The Coronavirus Assessment Tool, accessible on the hospital's website, guides users through a series of questions about risk factors and symptoms, then links them to an appropriate level of care, if needed. At the pandemic's peak, the service handled more than 1,100 patients per day and volume is still high—up to 300 to 400 users daily compared to 50 to 60 in the pre-pandemic era, said Dr. Czartoski.

The chatbot served as a model for other hospitals, including Brigham Health, part of Boston-based Partners Healthcare. The Partners version similarly provides patients with a simple, mobile-friendly application interface to facilitate triage, said Haipeng (Mark) Zhang, DO, MMSc, medical director of the Brigham and Women's Digital Innovation Hub.

“The service was originally designed to offload our hotline, which was becoming overwhelmed when COVID surged in our region,” said Dr. Zhang. “Now, when people call the hotline, they get a message to go to the chatbot.”

The chatbot provides a general platform that can be specified for various protocols, including but not limited to COVID-19 screening, he said. “We're actively refining it, adding features, and making it durable for the long haul so it can be used in a variety of scenarios.”

The health system also uses an online screener to make sure employees aren't infected. Tech experts created COVID Pass, a web-based tool that asks employees arriving at the hospital for a shift to log in and pass a screening test. Any with symptoms are instructed to return home and wait for information about next steps and additional screening from Partners' occupational health office, Dr. Zhang said. “Employee screening is critical to our ongoing COVID efforts beyond the surge,” he said.

Dr. Zhang's team is now working on improving the service by refining the application and implementing self-service kiosks for screening. COVID Pass can also be used to estimate how many staff are physically on the premises each day, he added. During surges, those numbers helped administrators determine how many face masks to have on hand, for example, or how many lunches to order from restaurants that were donating food during the crisis.

Machine learning models have also been used to improve identification of COVID-19. The models “learn” to distinguish between COVID-19 and other illnesses with similar symptoms by recognizing patterns in CT scans and X-rays—the more data that are fed into a model, the greater its diagnostic accuracy.

“The urgency and the sheer number of cases of COVID-19 have given us the ability to train these models on a large data set,” said Benson Babu, MD, MBA, chief of hospital medicine at Saint John's Episcopal Hospital in Far Rockaway, N.Y. “Once the model learns all the variations of the disease, it's able to diagnose COVID-19 and predict risk in individual patients with a high degree of accuracy based on radiographic data.”

In an April 22 study published in the journal Applied Intelligence, Dr. Babu and colleagues reviewed deep learning models developed by research teams around the world and showed how CT chest scan data could be used to differentiate between COVID-19 and pneumonia with upwards of 90% accuracy.

When linked to a hospital's EMR, these models help physicians make more accurate diagnoses by analyzing a patient's radiographic data and calculating the probability of COVID-19, said Dr. Babu. The model helps prevent missed diagnoses in the early stages of the virus, when some patients may have positive pulmonary imaging findings but still test negative for COVID-19 on nasal or throat swab tests.

In addition, the machine learning models assist with documentation by automatically suggesting the appropriate codes for reimbursement based on the information in the EMR, said Dr. Babu. The same data can be used to predict surges in patient volume and plan appropriate staffing levels. “Once these models learn all of the variations of the disease process patterns based on CT scans and X-rays from a large amount of patient data, they are can capture and predict who is at risk for COVID-19 with accuracy,” said Dr. Babu. “Using similar available data, we have been able to predict surges at our hospital three to four weeks in advance.”

While everyone hopes the need to predict COVID-19 surges will eventually decline, experts anticipate that many of the technological changes made for the pandemic are here to stay and could potentially transform practice. The rapid adoption of new technology by hospitals, clinicians, and patients over the past several months may emerge as a silver lining of the crisis. In fact, in retrospect, the pre-COVID-19 era seems like a pilot phase for telehealth, observed Dr. Umscheid.

“Virtual visits were there in the background, but not really adopted because nothing was forcing people to use it,” he said.