Vanderbilt doctors Jason Martin, left, Michael Hooper and Lisa Weavind are testing the new sepsis detection system in the ICU. When the system determines that a patient may have developed sepsis, a red light flashes in the electronic dashboard that displays patient information on ICU workstations. Photo by Steve Green,
When Jason Martin gives a talk about his research, he begins with the dramatic story of Mariana Bridi da Costa: The young Brazilian supermodel died from severe sepsis in January after amputation of both her hands and feet failed to stop its spread.
Martin, who is a fellow in allergy, pulmonary and critical care medicine, is part of an interdisciplinary team at Vanderbilt University that has come up with a high-tech approach to combat this deadly illness, which is one of the top 10 causes of death in the United States and kills more than half a million people worldwide every year.
The team is made up of clinicians and informatics experts from the
“This is an effort to use the power of informatics to move from reactive to proactive medical treatment by creating tools to support the use of evidence-based clinical guidelines,” said Peter Miller, director of the Vanderbilt HealthTech Laboratory, who oversees the project.
Sepsis, or systemic inflammatory response syndrome (SIRS), causes the body to attack itself. It is triggered when bacteria invade the body from outside through wounds or IV lines. The bacterial infection overstimulates the body’s immune system, setting off a cascade of inflammatory and abnormal clotting responses that can lead to organ failure and death. Infections that cause sepsis can be acquired outside the hospital, but those acquired in the hospital are more difficult to manage because many patients are already weak and the invading bacteria are more likely to be drug-resistant strains.
According to a study conducted by the Emory University School of Medicine and the Centers for Disease Control and Prevention (CDC) in 2003, the incidence of sepsis increased by an average of 8.7 percent a year over the prior 22 years. Today, sepsis treatment accounts for 40 percent of all ICU costs – more than $17 billion annually, according to a 2001 study.
In 2006, Miller came to Vanderbilt from the Department of Homeland Security with the mission of fostering “the revolutionary transformation of healthcare” by the application of informatics technologies. “For a while, we were a solution looking for a problem,” Miller recalled. In consultation with Gordon Bernard, the associate vice chancellor for research at VUMC, and with Associate Professor Arthur Wheeler and Assistant Professor Todd Rice in allergy, pulmonary and critical care medicine, Miller decided to focus on the problem of sepsis because it is common, deadly, expensive and treatable.
At the same time, ISIS computer scientists were investigating the issue of security and privacy of electronic patient records under the aegis of the National Science Foundation’s TRUST (Team for Research in Ubiquitous Secure Technology) Science and
Two years ago, Miller assembled his team. The clinical members were Martin; Liza Weavind, associate professor of anesthesiology; and David Maron, associate professor of medicine and emergency medicine. The informatics experts included Ed Shultz, director of information technology integration, and health system engineer Daniel Albert. Joining them from
The first step in the $360,000-plus project was creating a common vocabulary and knowledge base among the team members. The
“Working with the docs is a pleasure,” Mathe added. “They are very busy, but they are so smart it is scary. They pick up new ideas very quickly.”
The first part of the project was the development of an automated early detection system that can alert doctors that a patient may be developing sepsis. The doctors came up with a formula involving patient temperature, heart rate, respiration rate and white blood count that they felt was indicative of the onset of SIRS. Currently, the alerts appear on “patient dashboards” displayed on ICU workstations. (In the future, they hope to add the capability of displaying the alert on doctors’ cell phones.) When a doctor gets an alert, she checks out the patient. If she decides it was a false alarm, the system goes to sleep for 48 hours before resuming operation. If she decides the system was right and begins treatment, the alert system turns off for a week.
“For the last 15 years, we have been storing more and more of our information electronically,” Shultz said. “But even a few years ago we couldn’t have done a project like this because it makes decisions based on information stored on different systems that could not communicate effectively in real time.” Patient temperature and respiratory data are handled by one system, for instance, while another handles laboratory test results. So a major technical challenge was building pipelines between the different systems and getting them to “play nicely” with each other.
Creating the decision management system presented a different kind of problem. The doctors wanted to base it on a set of guidelines developed by the Surviving Sepsis Campaign, a worldwide consortium of critical care professional societies, based on peer-reviewed studies, an approach broadly referred to as evidence-based medicine.
“I tried to organize the protocol into a flow chart, but it was a mess,” said Martin. “It’s not easy to convert medical protocols into ones and zeros because there is a lot of nuance and judgment involved.”
After a period of “beating our heads against the wall trying to make this work,” the
Using this approach, Mathe and his colleagues developed a special modeling language specifically for clinical decision-making. “Although the language is specific to sepsis management, we made the underlying technical infrastructure so general that it can model virtually any medical protocol,” Mathe said. The team has already begun applying it to a second problem, treatment of chronic heart failure.
According to the researchers, it will take six to 12 months of operation before they can begin to judge the system’s effectiveness. They began running the detection system in the background in March. During this time they averaged three to four alerts per day in the 25-bed ICU. When the system was deployed, the very first alert led an attending physician to begin antibiotic therapy. They will run the detection system for several months before implementing the decision management system. This sequential implementation will allow them to independently assess the effectiveness of the alert and management parts of the system.
“A key message of this project is that collaboration is very important in addressing these kinds of problems,” Miller said. “When people from different disciplines come together they produce positive outcomes.”