Every day, doctors and nurses in the cardiac intensive care unit at Boston Children’s Hospital confront a quiet cacophony. Their tiny patients are surrounded by the soft beeping of dozens of bedside monitors, each emitting its own torrent of data.
Too often, staffers at Children’s and other hospitals are forced to respond to an emergency introduced by the shrill alarm of a monitor that indicates a patient’s condition has suddenly worsened.
“There are so many pieces of data coming at the physician. With almost 30 ICU patients, we’re just inundated,” said Dr. Joshua Salvin, a pediatric cardiologist at Boston Children’s. “If we had something that could tell us where the hot spots are on the floor, we could direct resources to the most sick patients.”
Now the hospital has such a tool.
Drawing from a field of data science called predictive analytics, Boston Children’s Hospital has joined with a startup named Etiometry to develop a system that aims to forecast changes in a patient’s condition before the alarms go off.
The crux of the approach is to analyze all the data the monitors put out — heart and respiratory rates, for example, along with other vital signs — and assesses whether the patient’s condition is at risk of worsening.
And because the system accumulates a long record of information, it can detect subtle changes in the relationships between vital signs to identify potential problems.
“When you think about all the data elements, every single one reveals something new about the patient,” said Evan Butler, Etiometry’s chief executive. “The algorithms we have incorporate every data point to create a collage of the patient’s status.”
Named the Stability Index, Etiometry’s system attempts to take some of the guesswork out of the myriad data coming from as many as a dozen machines.
A highly trained physician can read those blips and beeps like a sailor scanning the sky for signs of a storm. Yet two doctors might interpret the mass of data from the same patient slightly differently.
The Stability Index pops up on computer screens that Children’s Hospital physicians monitor at the cardiac ICU. Doctors choose different parameters to measure, then the Etiometry system renders its risk assessment on a simple numerical scale, with 0 being most stable and 4 the least.
For now, the system is still being tested internally, and only attending physicians have access to the index.
The unit’s medical director, Dr. Melvin Almodovar, uses it to double-check his own clinical assessment of patients. Etiometry’s founders are careful to note that physicians will always be the ultimate bedside decision makers, using the Stability Index to confirm or inform their own diagnoses.
Butler said that an information-overload environment like the intensive care unit is ideal for a data-driven risk assessment tool, because the patients teeter between life and death. A predictive model can act as an early warning system, pointing out risky changes in multiple vital signs in a more sophisticated way than bedside alarms.
Creating a health forecast for each patient through predictive modeling is the “holy grail” of ICU care, said Dr. Peter Laussen, the hospital’s former chief of cardiovascular critical care who forged the initial relationship with Etiometry.
Laussen laid the groundwork for the patient modeling system in 2010 when he and Almodovar moved to create one well-organized place to store all data from bedside machines. At the time, the cardiac unit relied on a combination of nurses’ notes and static data points gathered during rounds.
Laussen and Almodovar hired Arcadia Solutions, a health care IT consulting firm, to build visualization software that would combine the bedside data in one interface with a single view of the patient’s many vital signs. Dubbed T3, it was such a dramatic improvement that Laussen brought the system with him when he left Children’s to become the head of critical care medicine at the Hospital for Sick Children in Toronto last year.
The T3 made it possible for doctors at Children’s Hospital to view all patient information on one screen instead of scanning a dozen monitors. Still, doctors had to rely on training and intuition to gauge how the patient would fare over the coming hours and days.
Laussen and Almodovar next sought out a system that would give doctors a computer-based risk evaluation derived from each patient’s data set. Children’s Hospital launched a search for a software company with the chops to handle such a sophisticated assignment and considered the obvious names — big, established medical technology companies such as General Electric Co.
Instead, Children’s elected to work with Etiometry, then a young company started by three Boston University graduates. Butler and Dr. Dimitar Baronov are both aerospace engineers. The third Etiometry founder, Dr. Jesse Lock, is a biomedical engineer.
“We needed to find a partner to continue to develop T3 as a tool for data capture storage and visualization, but also a partner that was going to be able to use the data very wisely and take it to the next levels in terms of predictive algorithms,” Laussen said. “I realized very quickly that Jesse and Evan and Dimitar were able to utilize this data to do mathematical modeling and I realized they had a unique way to think about the data.”
Butler and Baronov had originally worked on mathematical algorithms to predict the flight paths of autonomous aircraft, or drones. The pair and Lock would meet over beers and talk about their research, and eventually decided medicine would be a better field for their skills than the military.
“For Dimitar and my perspective, I don’t think going to make missiles or anything really appealed to us,” Butler said. “And what Jesse pointed out is that health care is at least 30 years behind in terms of data analysis.”
Predictive modeling uses sophisticated mathematical equations to analyze a set of data and then project a future outcome — whether the flight path of a drone or the health of a child in intensive care. The science is widely used in commerce — Sears and Amazon.com use big data sets to make predictions about customers’ potential interests, while credit rating agencies use predictive modeling to determine credit scores.
Etiometry’s technology belongs to a field called machine learning, in which mathematical models can “learn,” or continuously sharpen predictions, as more information is analyzed. So, the more data fed into the model, the more accurate the prediction becomes.
The company is already refining its tool, by developing a color-coded interface for tablets and other computers that can pinpoint high-risk patients on the floor, instead of using the current number-based rating system. Etiometry and the medical staff also hope to expand the system to other units within Children’s and to other hospitals.
Laussen, now in Toronto, hopes that tools such as T3 and the Stability Index will help usher in the era of Big Data in hospitals.
“I attend many meetings, and I hear people talk about the ‘Era of Big Data,’ capturing it, utilizing it,” Laussen said.
“Everybody agrees that it’s the right thing to do, but we don’t have a uniform tool by which to do it. Within hospitals, we can’t do it. But we can if we partner with really smart people outside.”