Background Free text narratives in the Electronic Medical Record (EMR) provide rich information, but extracting data is difficult. For emergency department (ED) surveillance and to inform a prevention program for beach-related injury and illness (BRII), we developed and tested the ‘Wipeout Method’ to query ED EMR narratives in EPIC.

Methods The first of this five-step process involved identifying a cohort of ED BRII cases via lifeguard reports and generating an initial set of search terms based on their EMR narrative. The next four iterative phases involved using the set of search terms, updated for each phase, to query ED EMR records from sequential sample time periods. In each phase, we manually verified BRII cases and analyzed true and false positives of the search using a combination of single word, bi-gram and tri-gram frequencies; gold standard review of high activity days; deep word search of false positive terms; and text classification regression. The set of terms was refined at the end of each stage with the goal of minimizing false positives without compromising precision.

Results The ‘Wipeout Method’ generated a set of 49 query terms with 75.2% precision over all available ED EPIC records in our hospital, a 19-month period. We verified 1,605 BRII cases from 2,134 flagged records.

Conclusion This novel method allowed identification of the majority of cases in medical records with the use of minimal resources. The technique is widely applicable to other injury and public health areas for case identification for surveillance and study purposes.