Augusta Health has saved 282 lives with AI-infused sepsis early warning system

The hospital developed the system that gathers data from the EHR, runs it through artificial intelligence, and delivers real-time warnings to clinicians.
By Bill Siwicki
12:01 PM

In Virginia, the statewide mortality rate for sepsis was 13.2% in 2016. Sepsis is the body’s life-threatening response to infection that can lead to tissue damage and organ failure. In the U.S., 1.5 million people develop sepsis each year, and about 17% of those die. Early detection of sepsis is critical to decrease mortality.

THE PROBLEM

Clinical and IT staffs at Augusta Health, an independent, community-owned, not-for-profit hospital in Virginia, knew that studies have shown that though treatments are available in a general hospital setting, they are rarely completed in a timely manner.

“Our nurses are highly trained and are skilled at detecting early symptoms of sepsis based on standard indicators, but they are also very busy,” said Penny Cooper, a data scientist at Augusta Health. “Aware of how many patients our nurses care for and the many tasks nurses juggle at once, leadership formed a sepsis taskforce with the goal of providing staff with the resources to identify symptoms of sepsis sooner.”

PROPOSAL

For sepsis, it’s all about early detection. Mortality from sepsis increases by as much as 8% for every hour that treatment is delayed.

“By identifying sepsis early in the process, we have a much better chance of treating the infection before it goes too far,” Cooper said. “In addition, early identification of sepsis remains the greatest barrier to compliance with recommended evidenced based bundles.”

"By using artificial intelligence capabilities, we are able to screen 100% of our inpatient population and deliver results directly to the caregiver wherever they are in the hospital without any manual intervention."

Penny Cooper, Augusta Health

So Augusta Health decided to develop a system with Vocera Communications that provides an extra set of eyes that automatically reviews data. By reviewing the data electronically, staff is able to recognize symptoms earlier and provide automated alerts to the bedside nurses.

MARKETPLACE

On the clinical communications technology front, vendors include Avaya, Halo, HipLink Software, Mobile Heartbeat, PatientSafe Solutions, PerfectServe, Spok, Telmediq and Vocera.

MEETING THE CHALLENGE

Clinicians are familiar with the standard SIRS (Systemic Inflammatory Response Syndrome) criteria: Temp >38°C; heart rate >90; respiratory rate >20; abnormal white blood cells.

“But we wanted to see if we could increase the sensitivity by adding additional variables, so we began with the inpatient population and a retrospective study,” Cooper explained.

“We started with standard SIRS criteria but also added Mean Arterial Pressure and Shock Index,” she added. “By adding additional variables – MAP and Shock Index as categorical variable – to the Logistic Regression Analysis, we were able to increase the overall c statistic by .07. The c stat is a measurement of how well your test performs.”

Staff run an automated process that collects information from the Rauland Bed System and clinical data from the EHR for all current inpatients.

“The data then is compiled and analyzed and a score is assigned based on the results of the retrospective study,” Cooper noted. “This occurs every hour for every inpatient. Then for patients with a score greater than the cutoff, an alert is sent to the attending and charge nurse on their Vocera devices.”

The sepsis communication system was developed in-house and is an example of interoperability between different healthcare information systems. The process runs in the background so no one actually interacts with the process unless their patient is alerted and requires screening.

“By using artificial intelligence capabilities, we are able to screen 100% of our inpatient population and deliver results directly to the caregiver wherever they are in the hospital without any manual intervention,” Cooper explained. “Core measure requirements along with patient impact and historical volume trends also makes the introduction of this tool relevant and timely.”

RESULTS

The sepsis mortality rate at August Health now is 4.8%, compared with 13.2% at the state level.

“The work done by our teams at Augusta Health to reduce mortality rates from sepsis has been a collaborative effort,” Cooper said.

“We automated a process to screen all of our patients without any manual intervention but most importantly we have saved lives. By subtracting the actual mortality from the expected mortality rate, we estimate a total of 282 lives have been saved.”

Health Quality Innovators named the hospital the Health Quality Innovator for Virginia in the category of Data-driven Care in 2018. U.S. News & World Report recognized Augusta Health as a Best Hospital in the Shenandoah Valley for 2019-20.

ADVICE FOR OTHERS

“System interoperability is key to ensure that the right data gets to the right clinicians at the right time and without any manual effort,” Cooper advised. “While many EHRs may alert the clinician of sepsis within the medical record, this process delivers the alert to the communications device of the bedside nurse.”

With the input from clinicians and quality staff, the process was not overly difficult to accomplish, she added. Staff developed the system using off-the-shelf tools, including SQL Server and SQL Server Integration Services.

“The use of the model within the study facility has resulted in a culture change,” she said. “A review at a daily safety huddle prompts proactive rounding by the nursing directors. This process provides additional support that nursing staff may need to get patients to the appropriate level of care. For the immediate future, we plan to continue the use of the model within our facility, re-evaluate it from both an operational and clinical standpoint, and modify as appropriate.”

Twitter: @SiwickiHealthIT
Email the writer: bill.siwicki@himssmedia.com
Healthcare IT News is a HIMSS Media publication.