A PICTURE-perfect response to patient deterioration

ANN ARBOR, MI – While in the hospital, a patient’s condition may deteriorate due to various factors such as difficulty breathing, altered mental status or an unexpected drop in blood pressure. During deterioration events, early intervention is key to a positive outcome, yet their unanticipated and often sudden onset makes them difficult to identify in advance. At U-M Health, a predictive analytic tool developed by the University of Michigan Max Harry Weil Institute for Critical Care Research and Innovation is helping members of the hospital’s Rapid Response Team (RRT) stop the spark of deterioration before it starts.

Comprised of nurses from the surgical intensive care unit (SICU), the RRT functions like an ambulance inside the hospital, quickly bringing lifesaving interventions to patients whenever bedside teams make the call. RRT members Tiffany Watts, RN, and Ernie Saxton, BSN, RN, CCRN have been working alongside the Weil Institute’s Data Science core to deploy Weil’s “PICTURE” analytic within the RRT.

Short for “Predicting Intensive Care Transfers and other UnfoReseen Events,” PICTURE uses electronic health record data to passively and accurately predict a patient’s risk of deteriorating up to an average of 30 hours in advance of symptoms. While it has only been a few months since the official launch, the team is already reporting improved response times thanks, in part, to these predictions. 

“PICTURE’s alerts have been very consistent. It seems like we’re getting to these patients faster now,” said Watts, who serves as Clinical Lead for the Adult RRT. “It’s still early, but we’re seeing exciting things.”

“As a floor nurse before joining the Rapid Response Team, all I ever wanted was to be able to give the reason for my gut feeling when everything about a patient looked technically fine but something ‘felt bad’. PICTURE provides a concrete number that we can use with our own critical thinking to determine how these patients are really doing.”

Tiffany Watts, RN
Clinical Lead, Adult Rapid Response Team,
U-M Health

In addition to responding to emergencies when they happen, a chief goal of the RRT is intervening before these emergencies can begin. To help them get ahead of potential events, the RRT members will round at the start of their shifts, meeting with care teams and patients on each floor to learn in advance where they may be needed and why.

“Rounding can be a time-consuming process,” said Brandon Cummings, Senior Data Scientist at the Weil Institute and a lead on the PICTURE team. “The RRT might get a range of different feedback based on the experiences of who is on staff that day. Having a system like PICTURE—something that helps them prioritize patients—saves a lot of time and legwork.”  

“With PICTURE, we’re getting the knowledge we need earlier in the day so we can go to the patients that it flags as high-risk and potentially catch them before a situation elevates,” said Saxton, who serves as Educational Nurse Coordinator.

According to Saxton, PICTURE’s advanced warnings also provide key information once the team is at the bedside. “It can be difficult to focus while we’re also listening to the bedside nurse and the physician, while we’re putting a fluid IV in or putting some oxygen on. PICTURE helps us know what the situation is before we even get there.”

PICTURE is designed to model patient physiology as opposed to clinician behavior to ensure it always provides novel information. To generate its predictions, the system automatically and continuously processes an array of data routinely collected as part of patient care including lab results and vital signs. It takes this information and consolidates it into a number—a risk score—that allows care teams to see and understand a patient’s status at a glance. PICTURE also provides a list of the factors that go into each score to help guide clinical decision-making. To the RRT, such explanations are invaluable.

“As a floor nurse before joining the Rapid Response Team, all I ever wanted was to be able to give the reason for my gut feeling when everything about a patient looked technically fine but something ‘felt bad’.” said Watts. “PICTURE provides a concrete number that we can use with our own critical thinking to determine how these patients are really doing.”

“Because we have this list of potential events, and because we can see what exactly is going into each score, we can consult ahead of time with the floor nurses and staff as well as within the RRT,” said Saxton. “If we notice a patient who initially had a lower risk score start trending higher, we can bring that up.”

Over the next year, the Weil Institute Data Science Team will be working with Saxton, Watts and the RRT as a whole to further optimize PICTURE. The system currently includes versions for Adult and Pediatric All-Cause Deterioration Prediction with additional models in development.

“We used to say rapid response systems generally would stop a spark from becoming a fire,” said Saxton. “Now, with PICTURE, it seems like we’re able to look at cases where someone was getting ready to strike the match, and we’re able to stop the spark on some of them.”