Background/Aims: The serrated pathway accounts for 15–25% of sporadic colorectal cancer (CRC). In our study, we sought to accurately characterize sessile serrated polyps (SSP) in a population by electronically interrogating colonoscopy patients’ endoscopy and pathology reports using a rules-based text search of pre-defined SSP-related terms. To this aim, we compared a sample of putative SSP and hyperplastic polyps (HP) using our algorithm to a determination of SSP or HP by pathologist and molecular examination to determine the feasibility of large-scale identification of SSP in electronic medical records. Methods: In 23,990 endoscopy reports from colonoscopies with pathology performed at a University of Utah Healthcare facility in 2000–2012, we identified serrated lesions and categorized each as putative SSP or HP using a text search algorithm. We obtained 93 tissue samples for histologic and molecular analysis. Results: Serrated polyps were categorized as putative SSP (N = 920) and putative HP (N = 7159) by text search algorithm. Histologic examination of 93 samples identified 37 SSP, 11 probable SSP, and 45 HP. Of 26 putative SSP, 25 were SSP/probable SSP (96%) by histology. Of 67 putative HP, 44 were HP (66%) by histology. Reducing size criterion from ≥1 to ≥5 mm in the search algorithm caused improved sensitivity (77.1%) without decline in specificity (97.8%). Conclusions: A simple rules-based search to identify SSP provides “proof of principle” that SSP can be identified in a large electronic record set. Pilot data indicate defining large, right-sided polyps as ≥5 mm provides adequate sensitivity to detect SSP from electronic records while maintaining high specificity.
- Receiver operating characteristic curve
- Sessile serrated adenoma/polyp
- Text mining
ASJC Scopus subject areas