Photo: Thomas Barwick/Getty Images
Pickleball is among the fastest growing sports in this country. The number of players has ballooned by more than 200% over the last three years, according to Pickleheads.
The number of pickleball injuries has also risen. Most arise from falls and include serious bone fractures, ligament sprains and muscle strain, according to NBC News.
To get data on the injuries, Cedars-Sinai is using AI to extract data from chart notes. At HIMSS25, Kathy Bailey, principal data intelligence analyst for Cedars-Sinai, will explain what she and her team have learned and how that applies to clinical research during the session “AI-Powered Extraction of Pickleball Injury Information from Patient Notes,” on Tuesday, March 4.
Bailey has no comprehensive figures on the number of pickleball injuries, but Cedars-Sinai has been able to identify more patients than they could have using more traditional methods.
“We identified more than enough patients that the study had hoped for, with 80% correctly identified in our sample,” Bailey said.
GPT-4 is extracting discrete data elements from free-text chart notes to identify pickleball-related injuries. The relevant data gives information such as the likelihood of pickleball-related injury, the injury site, the severity level and date of injury, which significantly reduces manual review time.
Leveraging AI can improve data accuracy and operational efficiency while minimizing manual efforts and data preparation, Bailey said. It efficiently searches through clinical notes to find information that isn’t captured in structured medical codes or other discrete methods.
The rationale behind the project is to address the increasing demand for robust data extraction methods in clinical research. Traditional methods, such as REGEX in SQL, often fail to capture contextual information, necessitating extensive manual review and data cleansing, Bailey said.
Robust data extraction is essential, Bailey said, because critical patient information often lives in unstructured clinical notes rather than structured database fields.
Without it, while doctors and nurses write detailed observations and interpretations in their notes, this valuable information remains locked away and does nothing to support clinical research.
“In our pickleball injury study, the AI component – the large language model used – labeled relevant cases by scanning physician notes, providing rich contextual details about each injury, like how it happened, the severity and injury site,” Bailey said. “This depth of information isn’t available through traditional coding-based searches. Plus, AI can analyze thousands of charts in a fraction of the amount of time, compared to the weeks or months it would take for manual review, making large-scale studies more feasible.”
Kathy Bailey, principal data intelligence analyst for Cedars-Sinai, will speak at HIMSS25 on “AI-Powered Extraction of Pickleball Injury Information from Patient Notes,” on Tuesday, March 4 from 12:45-1:45 p.m., in Venetian | Level 5 | Palazzo M.
Email the writer: SMorse@himss.org
