UChicago Medicine receives $2.4M grant to prevent diagnostic errors for stroke
November 12, 2019
The University of Chicago Medicine has been awarded a $2.4M grant over four years from the Agency for Healthcare Research and Quality (AHRQ) to study and prevent diagnostic errors when treating stroke patients in the emergency department.
The project, called Targeted Healthcare Engineering for Systematic Interventions in Stroke (THESIS), is a collaboration among six Chicago-area hospitals that will use modern predictive analytics and artificial intelligence tools to analyze patient data and medical scenarios that can lead to misdiagnosis of stroke, and then develop tools and procedures to help emergency providers limit such errors.
“Diagnostic error is a major source of patient harm and poor quality of care in the U.S.,” said Shyam Prabhakaran, MD, Chair of the Department of Neurology at UChicago Medicine and co-principal investigator for the project. “If you make the wrong diagnosis, then downstream treatment decisions will be wrong too.”
The AHRQ is a division of the United States Department of Health & Human Services that supports research to improve the safety, quality and accessibility of health care. One area of focus is studying diagnostic errors, or mistakes that can lead to missed diagnoses, under- or over-diagnosing the severity of a problem, or making a false positive diagnosis.
We may not be able to predict the optimal intervention for every patient, but if we can be more alert in those cases it’s a big step forward.
A review of three large observational studies in 2014 found that about 5% of adults experienced diagnostic errors each year, and over half of these errors had the potential for severe harm. Proper diagnosis is crucial for stroke, where time is of the essence to begin appropriate treatment before the patient suffers irreversible neurological damage. Prabhakaran estimates that the vast majority of stroke patients present with classic symptoms that are straightforward and easy to diagnose quickly, but the small remaining subset of borderline or atypical cases can lead to mistakes.
Prabhakaran and UChicago health services researcher Jane Holl, MD, are leading the collaboration of three Chicago-area academic medical centers and their affiliates: UChicago Medicine and Ingalls Memorial Hospital, Northwestern Medicine and Lake Forest Hospital, and Rush University Medical Center and Rush Oak Park Hospital.
Analytics, modeling and prediction
During the first year and a half of the project, the research teams will collect data from each hospital about patients who came to the emergency department and were incorrectly diagnosed. Then they will apply advanced analytical tools and artificial intelligence software to analyze scenarios and identify patterns or variables that could lead to misdiagnosis. The research team also will conduct retrospective interviews with staff to learn more about specific interactions with patients, or what they recall about the cases that may have influenced their decision making.
“We’re looking for everything that goes on during the physician-patient interaction, including everything around them, the medical record system and the environment in the emergency department at the time,” Holl said. “We want to try to identify what leads to this error occurring, how doctors are thinking or getting information from the patient, and how they are processing that information to come to the diagnosis — or to miss it.”
Once this data is analyzed, the research team can then use predictive modeling tools to simulate possible solutions and improvements and see which steps are most effective. This will allow them to evaluate different approaches and fine-tune the improvements before actually implementing them in the clinic.
Prabhakaran and Holl said that these improvements will take the form of enforced prompts and procedures, rather than additional training that relies on staff to respond differently in a chaotic and fast-paced ER. For example, the electronic health record system might prompt a physician to ask more detailed questions if they type imprecise notes such as “the patient’s wife says he fainted” or “the patient has slurred speech.” Either of those symptoms could be a sign of other conditions, like diabetic low blood sugar, and not a stroke.
At the end of the four years, the team will develop a complete pilot implementation of recommendations to avoid diagnostic errors for stroke, including updated standard questions for evaluating patients or additional tests that are automatically prompted by certain symptoms or lab values.
“There are a lot of patients where it’s very cut and dried, but for that small cohort where it’s not, we want to be able to identify when you have a patient that you should be concerned about making the right diagnosis,” Holl said. “We may not be able to predict the optimal intervention for every patient, but if we can be more alert in those cases it’s a big step forward.”
Shyam Prabhakaran, MD
Shyam Prabhakaran, MD, is the Chair of the Department of Neurology at UChicago Medicine. He is an internationally recognized leader in vascular neurology and stroke research and treatment, and has led projects focused on uncovering the underlying causes of recurrent strokes, improving stroke care, and optimizing patient outcomes and recovery.Learn more about Dr. Prabhakaran