Could artificial intelligence (AI) leverage nursing documentation to predict when a patient’s condition is about to deteriorate, up to 24 hours ahead of other early warning systems?

Brigham nurse-scientist Patti Dykes, PhD, RN, FAAN, FACMI, believes the answer is yes, and she and her team are launching a national study to investigate.

Through the CONCERN project (which stands for Communicating Narrative Concerns Entered by RNs), the team is developing evidence-based clinical decision support (CDS) that builds on more than 10 years of data science investigation. Their goal is to give clinicians advance notice of when to intervene and provide potentially lifesaving care before a major event, such as cardiac arrest or the onset of sepsis.

“Our team has used artificial intelligence and machine learning approaches to confirm patterns of nursing documentation that are predictive of patient deterioration,” said Dykes, principal investigator of the project, which is a collaborative effort with Columbia University and the National Institute of Nursing Research.

“Specifically, when nurses are concerned about a patient, they increase their surveillance, resulting in increased documentation in the electronic health record. We have taken these nursing patterns and built predictive models using patient outcomes, including rapid response calls, cardiac events, inpatient mortality and unexpected transfers to intensive care units.”

What differentiates the CONCERN CDS from other early-warning systems for patient deterioration is that it is not triggered by physiological changes, such as out-of-range vital signs that are known to be late indicators of deterioration, says Dykes.

“Instead, CONCERN detects the nurses’ expert clinical judgment when they see that there are slightly perceptible changes in the patient’s clinical state,” she said. “Because CONCERN is based on subtle changes that usually occur well before physiological alterations, it can predict patient deterioration five to 24 hours earlier than warning systems based on physiological data.”

The CONCERN CDS prototype was designed with the input of Brigham nurses.

“When Patti first mentioned this study, I was immediately interested, as I often take care of septic patients and wanted to be involved in improving the process of early detection,” said Julia Marvel, BSN, RN, of Braunwald Tower 14CD, who worked as an intern on the project along with colleague Jack O’Meara, BSN, RN.

They collected data and information regarding sepsis indicators, such as a patient’s blood pressure and heart rate, as well as nursing documentation and notes. “I focused largely on what a nurse does when he or she believes a patient’s condition is deteriorating, as we are finding that a patient’s vital signs and lab work may not be the only predictors of sepsis,” Marvel said.

Marvel and O’Meara also conducted patient chart reviews to find commonalties between patients with sepsis, such as low hematocrit, elevated lactate and elevated white blood cell count.

“My research demonstrated that the frequency of paging the responding clinician increases, as well as the frequency of vital sign documentation and nursing notes,” Marvel said.

Patti Dykes

Jack O’Meara and Julia Marvel

These kinds of patterns are built into the CDS so that it “knows” when to alert the care team that certain activities the nurse is doing may indicate the patient is at risk for developing sepsis or suffering a cardiac arrest.

“It’s exciting to see the importance being placed on sepsis from a research standpoint,” said O’Meara. “I hope the clinical experience of myself and Julia is helpful in furthering this project and improving patient care.”

The CONCERN CDS uses existing electronic health record documentation. Dykes is in the process of working with Brigham nurses to identify the best way to incorporate it into Partners eCare so that it can support nurses and other clinicians in early recognition of at-risk patients.

The system will be implemented and evaluated at the Brigham, Newton-Wellesley Hospital, New York Presbyterian Hospital-Columbia University Medical Center and The Allen Hospital, part of the New York Presbyterian Health System.

The study will likely begin recruiting patients in early 2020.

Home