Annual Conference: Houston Methodist uses AI in virtual models for critical care, patient observation and more

Artificial intelligence (AI) is driving innovation in clinical operations at the Houston Methodist hospital system, as explained during a presentation this week at HFMA’s Annual Conference.

Among other virtual-care programs developed by the organization’s innovation center, Houston Methodist has incorporated a virtual ICU and virtual observation in the inpatient setting. Both programs use AI to enhance processes and help guide clinical decision-making, and the technology could phase out the need for human operators altogether in the virtual observation model.

Such programs are “really a joint initiative between operations and IT working together to try to figure out what we need to do and how we can disrupt ourselves instead of having someone else do it for us,” said Michelle Stansbury, the health system’s vice president for innovation and IT applications.

Improving critical care capacity

The virtual ICU program was implemented specifically for overnight hours as the health system struggled to find enough intensivists to staff its hospitals. AI feeds predictive algorithms that allow for preemptive interventions, tracking 18 key measures, among them heart rate and respiratory rate.

“Not only do we have the nighttime coverage now with less individuals than what we had trying to staff for them being on-site, but we’re able to intervene much faster and quicker with these patients based on the algorithms that were built within this program,” Stansbury said.

Other proven benefits, she said, include improved ICU throughput, a reduction in burnout among ICU clinicians and improvement in clinical measures such as severity-adjusted outcomes, length of stay and hospital-acquired conditions.

Experienced intensivists, some of whom are off-site, are available for immediate consultation on issues related to patient management. They work collaboratively with bedside teams and offer care-plan support. Virtual RNs work in the on-site virtual operations center and can request that bedside nurses or virtual intensivists attend to a particular issue with a patient.

Clinical feedback on the virtual ICU concept was not always positive, with one physician going as far as to say, “You are going to kill people.”

“Eventually, they came around,” Stansbury said.

Keeping an eye on at-risk patients

The genesis of the virtual observation program (a.k.a., tele-sitting) was the installation of camera technology in rooms to monitor patients who were at risk of harming themselves or others, with nurses gathering in the virtual operations center to use the technology. The organization soon saw an opportunity to utilize the space in other ways, with one patient observation technician able to keep tabs on a dozen patients at once.

The techs can interact with patients and providers, and redirect and escalate as necessary, to ensure a safe environment. As the observation program grew over a recent 11-month period, the monthly number of monitored hours increased from 449 to 32,005.

“For fall patients, you don’t absolutely have somebody sitting in the room,” Stansbury said. “[Techs] can talk two-way with the patients themselves when they’re trying to get out of bed, and notify the clinician.”

An ongoing pilot is testing a system that could be operated with no techs.

Cameras “can detect whenever a patient is trying to get out of bed,” Stansbury said. “And as the detection is happening, the system can automatically tell them, ‘Please stay in your bed. Nursing will be here very soon.’ It also alerts the nurse that now a patient is trying to get out of the bed.”

Streamlining admissions and discharges

In general, a health system that finds a good vendor for such technology should consider what other use cases can be addressed.

For example, the two-way video technology used for the virtual ICU program, along with the iPads that had been installed in patient rooms during the pandemic, laid the foundation for a telenursing program to help make up for the staffing shortage.

A time study of the organization’s nurses had found they took 30 to 45 minutes to complete an admission or discharge, in large part because they kept getting interrupted by the need to attend to other tasks.

“Is that really nursing care you would want them doing? We said, ‘Let’s try to do virtual,’” Stansbury said.

The idea evolved to placing dedicated one-on-one nurses in the operations center.

As a result, “HCAHPS scores have gone through the roof,” Stansbury said. “Our patients are very satisfied.”

Next frontiers

A partnership with a vendor is allowing Houston Methodist to leverage 22 years of Medicare data via a risk stratification tool to gain a holistic view of patients when they’re away from the facility.

Also, an AI tool is being integrated with the electronic health record to leverage all information on a patient’s chart to reduce deterioration and, in turn, length of stay.

“What are those clinical barriers that are happening? What are the complex-care things that are happening, discharge barriers, everything else that you can look [at]?” Stansbury said. “On top of that, what are the social vulnerabilities that are going on with that patient?”

The tool has identified the leading barriers to discharge: pending consults, IV medication, pending tests and procedures, post-discharge placement and pending test results. Stansbury said such information has been a boon for case managers.