Home Icon
Search Icon

Anesthesiology Control Tower: Feedback Alerts to Supplement Treatment (ACTFAST)

Project Final Report (PDF, 243.65 KB) Disclaimer

Using algorithms can predict postoperative adverse outcomes with a high degree of accuracy leading to better outcomes for the highest risk patients.

Project Details -
Completed

Most patients undergo surgery to cure a condition or improve quality of life; however, an estimated 1 to 5 percent of surgical patients die within 1 month of the operation. Furthermore, 10 to 20 percent of these patients experience major complications such as heart attacks, unremitting pain, infections, and blood clots in the weeks to months following their procedures. Some of this morbidity and mortality may be preventable through early identification of risk factors and better communication regarding risk mitigation.

To address these issues, researchers at the Washington University School of Medicine developed and evaluated an air traffic control-like command center for operating rooms (ORs). The Anesthesiology Control Tower: Feedback Alerts to Supplement Treatments or ACTFAST study applied data mining and machine learning to forecast adverse patient outcomes. This study used data from perioperative electronic health records (EHR) and real-time physiological data including patient demographic characteristics, comorbid conditions, preoperative vital signs, selected preoperative laboratory values, intraoperative time series, and selected intraoperative medications. Expert clinicians outside of the OR used Anesthesiology Control Tower (ACT) software to monitor the real-time status of all ORs in the operating suite. Detected events triggered an alert on a clinical dashboard. If the outside clinician decided the alert is important, they provided the attending anesthesiologists with real-time decision support on their personal communication devices. The ACT was tested in a randomized controlled trial (RCT).

The specific aims of this projects were as follows:

  • Develop, refine, and validate forecasting algorithms for adverse outcomes. 
  • Assess the usability of an ACT for the operating suite. 
  • Assess whether the ACT improves clinician compliance with standards of care and surrogate measures of patient outcomes. 

An SQL database was developed from four electronic health data repositories. Novel machine learning algorithms were constructed to predict postoperative 30-day mortality, postoperative acute renal failure, and postoperative acute respiratory failure. Results indicated that the algorithms were able to predict postoperative adverse outcomes with a high degree of accuracy. Researchers began the RCT, but results of this RCT were not available at the end of the grant period because of mid-study EHR vendor changes and the COVID pandemic.

The research team feels that adoption of ACT should be a priority so that providers are made aware of the highest risk patients that would benefit from greater monitoring. The team is confident that the results from the completed RCT will provide value and insight.

The Anesthesiology Control Tower: Like Air Traffic Control for Operating Rooms

The Anesthesiology Control Tower: Like Air Traffic Control for Operating Rooms

Using algorithms for real-time monitoring during surgery can predict and prevent adverse outcomes, leading to better outcomes for patients.

Surgery is common, but still risky

Surgery is a big insult to the human body. A lot can go wrong. In fact, it does. An estimated 10 to 20 percent of patients who undergo major inpatient surgery experience major complications such as heart attacks, unremitting pain, infections, and blood clots in the weeks to months following their procedures; about two percent are dead within 30 days of surgery. Some of this morbidity and mortality may be preventable through early identification of risk factors and better communication to mitigate risks during the surgery. How can we leverage all the data that are collected during surgery—vital signs, heart rate, blood pressure, temperature, fluid administered, drugs given—to predict potential outcomes during surgery in a way that supports decision making?

Air traffic control concepts can predict high risks for healthcare complications and improve decision making

To address this, Dr. Michael Avidan looked to another high-risk industry—aviation—that has significantly increased flying safety by using air traffic control centers on the ground, monitoring all the nearby airplanes and helping them to coordinate their activity, and prioritizing what needs to happen next.

In the operating room, clinicians are inundated with so much information that they can't sort it out in real time and separate the wheat from the chaff. And during operations, the patient's risk isn't static. It changes with time. So, if you've got a lot of bleeding, your risk will go up. If you've got a long period with low blood pressure, your risk will go up.”
–Dr. Avidan

Dr. Avidan and a team of researchers and computer scientists at Washington University developed and evaluated an air traffic control-like command center for operating rooms (ORs). The Anesthesiology Control Tower: Feedback Alerts to Supplement Treatments (or ACTFAST) study applied data mining and machine learning to develop predictive algorithms, helping to predict patients who are at risk for specific complications, including respiratory failure, kidney failure, and death. This study used data from perioperative electronic health records and real-time physiological data that included patient demographic characteristics, comorbid conditions, preoperative vital signs, selected preoperative laboratory values, intraoperative time series, and selected intraoperative medications. Expert clinicians outside of the OR used Anesthesiology Control Tower (ACT) software to monitor the real-time status of all ORs in the operating suite, where detected events triggered an alert on a clinical dashboard. If clinicians in the ACT decided the alert was important, they provided real-time decision support to the anesthesia team working "on the ground" in the OR on their personal communication devices, and final decisions regarding clinical care were made by the "ground team."

In the modern world, we now have tools to try to help clinicians to care better for patients, and we should be leveraging technology to help us in this way. Planes are going to be safer when there are computers on board that help the pilots and air traffic control centers that are on the ground. I think that we should be striving to make the acute care of patients safer, using the same kind of thinking and paradigm.”
–Dr. Avidan

Integrating ACTFAST into operating rooms improves care

Machine learning algorithms for predicting postoperative death, acute kidney injury, and acute respiratory failure were successfully developed and validated using a database of approximately 110,000 surgical patients. Algorithms were able to predict postoperative adverse outcomes with a high degree of accuracy. The team found that the ACT improved clinician compliance with standards of care, as well as with surrogate measures of patient outcome, such as blood pressure control, temperature, or glucose control. The ACTFAST technology is now fully integrated into all Washington University School of Medicine’s operating rooms. Importantly, because the clinicians in the ACT regularly communicate with anesthesiologists in the operating room, these anesthesiologists have begun to view these clinicians as valuable collaborators. Such cultural acceptance of the ACT by clinicians in the ORs is necessary for the ACT intervention to have any impact on process measures or patient clinical outcomes.

The success of this AHRQ-funded study provided the foundation for a 5-year NINR-funded R01, the Telemedicine Control Tower for the OR Navigating Information, Care, and Safety (TECTONICS) trial, which will evaluate the impact of the ACT on clinical outcomes of patients post-surgery. This exciting and burgeoning research program would not have been possible without the foundational support for the ACTFAST study by AHRQ.