Using machine learning in electronic medical records to save a child’s life in the hospital

Machine Learning

This article has been reviewed in accordance with Science X’s editorial processes and policies. The editors highlight the following attributes while ensuring content authenticity:


trusted source


Infographic. credit: Pediatric critical care (2023). DOI: 10.1097/PCC.0000000000003186

In a report published in Pediatric critical carea team at Nationwide Children’s Hospital describes a machine-learning tool for timely identification of hospitalized children at risk of exacerbation. It certainly outperforms existing situational awareness programs when it comes to identifying exposed children.

“Predictive algorithms focused on improving clinical care have been increasingly developed over the years, but are largely not operationalized. Moving algorithms from computers to the bedside can be a long process. It requires the involvement and collaboration of clinicians, data scientists, and clinicians – informatics,” says Laura Rust, M.D., Ph.D., emergency medicine physician and physician informatics at Nationwide Children’s and lead author of the paper. said. “This project has been a journey of over five years and we are truly proud of its successful integration into our safety culture and its impact on patient outcomes.”

The Deterioration Risk Index (DRI) was built from the foundation of the Watchstander situational awareness program already in use at Nationwide Children’s. To facilitate adoption, the team utilized the same response mechanism for alerts. That is, patient assessment and huddle with her bedside care team within 30 minutes, risk mitigation, and creating an escalation plan.

Three diagnostic groups (cardiac structural defect (cardiac), oncology (malignancy), general (neither cardiac nor malignant)) were used to train three separate predictive models and evaluate the implemented algorithm. Developed.

“One of the design features that helped build trust with the clinical team was that we didn’t necessarily identify new criteria. Our model identified the most important existing situational awareness criteria. and just weight it accordingly,” says Tyler Gorham. Co-author of his IT Research & Innovation and publications at Nationwide Children’s.

According to Dr. Rust, there can be an enormous amount of clinical data to process at once in electronic medical records, especially after handoffs and transitions of care. The model helps reduce this cognitive burden by automatically handling these risk criteria behind the scenes. It has the advantage of being integrated with the Electronic Medical Record (EMR), keeping all data from all previous points in time, not just the current shift.

DRI was 2.4 times more sensitive than existing situational awareness programs and required 2.3 times less alarms per detected event. Specifically, the team observed a 4-fold improvement in sensitivity in the heart group and a 3-fold improvement in the malignancy group. A post-implementation pilot study found a 77% reduction in degradative events in his first 18 months compared to the event rate expected in the previous years.

According to the developers, perhaps the most important aspect of the model is its transparency.

“This is not a black box. We show clinicians what goes in and how algorithms evaluate the data and trigger alarms,” says Gorham. “This tool helps support clinical decision-making by allowing the clinical team to see why an alarm was triggered.”

The team also conducted a road show, visited the clinical unit where the tool will be deployed, answered questions, ran simulations with the bedside care team, and incorporated feedback.

“At Nationwide Children’s, our team is committed to a zero hero safety culture,” says Dr. Rust. “This provided a foundation and a shared mission for our multidisciplinary team to see this across the finish line.”

Further information, including algorithmic details, is available in the publication.

“We shared the recipe in publications,” says Gorham. “If others are interested, they can use the data from their center to retrain the model for locals. Hopefully, we can support better outcomes for all children, including those who are not in our care.”

For more information:
Laura OH Rust et al, The Deterioration Risk Index: Development and Piloting a Machine Learning Algorithm to Reduce Pediatric Inpatient Deterioration, Pediatric critical care (2023). DOI: 10.1097/PCC.0000000000003186

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *