Electronic Health Record Data Reveals Patterns of Care for Children with Cerebral Palsy
Published March 2021 | Developmental Medicine & Child Neurology
For children with cerebral palsy (CP)—the most commonly seen motor disability in children—care involves providers from many different specialties. Coordination of care among these groups is essential to improving outcomes. By characterizing the patterns of care for children with CP, Brad Kurowski, MD, MS, and colleagues are providing a foundation for understanding how care could be improved.
The team began by extracting electronic health record data for more than 6,000 children with CP over a 10-year period. Across 34 specialties, a total of nearly 4 million in-person visits and care coordination notes were identified. Per child, the duration of care averaged five years and five months, with five specialty interactions and 22 in-person visits each year. Overall, the ratio of in-person to care coordination visits was one to five, meaning that most interactions with care teams occurred outside of in-person visits.
Using hierarchical clustering, a machine learning algorithm that groups similar objects, the team also identified seven clusters of care—musculoskeletal and function, neurological, high-frequency/urgent care services, procedures, comorbid diagnoses, development and behavioral, and primary care.
“These care patterns help us gain a better understanding of where care can be optimized,” says Kurowski. “Medical informatics, machine learning, and big data approaches provide unique insights to inform the development of precision care models for individuals with CP.”
Next, researchers will use these insights to develop clinical dashboards and identify areas for intervention.
Care Clusters in Cerebral Palsy
This graph shows clusters of care for children with cerebral palsy. Clusters of specialties are shown on the x-axis and individuals with cerebral palsy are shown on the y-axis. Each row represents the visits of one child across all specialties at one tertiary medical center. Each column represents the number of visits to a specialty over 10 years. The number of visits was normalized using the mean of each column. Clusters were created using unsupervised hierarchical clustering applied to both rows and columns. PACU, post-anesthesia care unit; BMCP, Behavioral Medicine and Clinical Psychology; DDBP, Division of Developmental and Behavioral Pediatrics.
