A new open-source model of patient deterioration is improving care in the University of Michigan healthcare system.
Now, a study published in the British Medical Journal shows that it is effective in 12 other hospitals in the United States, surpassing the accuracy of the widely used EPIC deterioration index by more than 21%.
In addition to the efficiency of the model, the way it was designed paves the way for considerably faster development of future models, bypassing the challenges of sharing sensitive patient data.
Models of patient deterioration help doctors and nurses make better decisions about care, such as proactively transferring the most at-risk patients to the ICU before they deteriorate. They can also help providers identify patients at lower risk of serious complications, i.e. patients who might benefit from earlier discharge from hospital or transfer to a care facility. of less intensity.
“We were able to develop the M-CURES model in a fraction of the time it took to build previous models through close collaboration between clinicians and data scientists and allowing other healthcare systems to validate the model without sharing any of their patient data,” mentioned Jenna Wiensassociate professor of electrical engineering and computer science at UM and senior author of the paper.
Wiens explains that the ability to quickly develop effective predictive models can be crucial in a situation like a pandemic, where quick action is essential and the threat is poorly understood. Development of the M-CURES model, she says, began in early 2020, at the onset of the COVID-19 pandemic. The UM, Michigan Medicine health system needed a better way to predict COVID-19 patient outcomes.
Predictive models use machine learning algorithms that sift through vast amounts of patient data, “training” themselves using correlations in past data to predict future outcomes. The prototype models are then validated using even more data from other patients.
The team’s first challenge was a lack of data to train the model, because COVID-19 was so new that the previous patient data normally used to train the models simply didn’t exist. So they used data from pre-pandemic patients, identifying a cluster of respiratory symptoms similar to COVID-19 and extracting five years of data from those patients.
Next, the team worked closely with Michigan Medicine clinicians to shorten the typically months-long process of summarizing thousands of data points into a handful of key predictors. They developed a hybrid approach where data scientists and clinicians worked together to eliminate potentially misleading variables. This allowed them to validate the effectiveness of M-CURES at Michigan Medicine in just a few weeks.
“Our hybrid approach to feature selection used both data-driven techniques and expert clinician knowledge,” said Fahad Kamran, co-first author of the study with Shengpu Tang. Both are PhD students in computer science and engineering at UM. “This type of collaboration gave us confidence in the final model, even though limited data was available for validation early in the process.”
To speed up the crucial step of validating their model in other healthcare systems, Weins’ team pioneered an approach to avoid the months-long process of accessing sensitive patient data. Instead, they simply sent their newly developed code to teams at other hospitals, who applied the model internally and reported the results. This allowed Wiens’ team to quickly validate the M-CURES model in a dozen hospitals in the United States with different structures and demographics, helping to ensure that the algorithm is accurate and fair.
The next step is for M-CURES to be used by Michigan Medicine rapid response teams as a real-time tool to identify patients at risk of deterioration.
“Rapid Response Teams are specialized clinical teams that can act quickly to intervene on patients before they experience poor outcomes,” said Michael Sjoding, Assistant Professor of Pulmonary and Critical Care Medicine at Michigan Medicine and clinical lead in the development of M-CURES. “We are delighted that the M-CURES model is supporting this effort.”
Perhaps the most important outcome of the project is the ability to use the tactics developed for the M-CURES model to more quickly develop predictive models for emerging health threats in the future.
“Models like M-CURES hold tremendous promise for improving both clinical care and resource allocation in healthcare settings,” said Erica Shenoy, an infectious disease specialist at Massachusetts General Hospital and author of the paper. . “They can provide important prognostic information to clinicians at the point of care.”
The collaboration also included researchers from Mass General Brigham, University of Texas Southwestern, University of California San Francisco, and Harvard School of Medicine.
The work was supported by the National Science Foundation; National Institutes of Health; National Library of Medicine; National Heart, Lung and Blood Institute; Agency for Research and Quality in Health; Centers for Control and Prevention of Disasters; National Center for Emerging and Zoonotic Infectious Diseases; Precision Health at UM; and Institute for Health Care Policy and Innovation at UM.