Abstract
BACKGROUND/AIMS: Screening colonoscopy is inefficient in that precancerous polyps (adenomas, advanced adenomas) or early carcinoma will be detected in only a small proportion of the patients undergoing the procedure. If predictive models could be developed to predict which types of patients are most likely to have advanced adenomas, screening colonoscopies could be targeted to these groups. This 1-year pilot project, funded by CRN-3, is collecting the data needed to construct an adenoma prediction model. METHODS: Subjects are Health Partners Medical Group primary care or specialty care patients who had an initial screening colonoscopy at a Health Partners outpatient endoscopy center between 2008 and 2010. Eligible subjects will be identified using a 2-step process: Step 1 will select potential patients based on age and availability of referral information. Step 2 will restrict the analysis to initial procedures performed for screening indication. The narrative pathology reports of these patients will be extracted from the central surgical pathology database, and information extracted to ascertain the polypectomy outcome (carcinoma, advanced adenoma, hyperplastic adenoma, adenoma number and size). Demographic and risk factor data will be collected from the primary care records of these patients, including age, gender, race, family history, body mass, smoking status, and diabetes diagnosis. RESULTS: To date, the study has identified 22,088 unique subjects (step 1) who had a procedure during the study interval. Of this number, 10,794 had polypectomies (48.9% polyp yield). We have successfully constructed and tested a polyp classification algorithm using the pathology data of 130 subjects. Polyp pathology records have been extracted from the electronic pathology database. Risk factor data for age, gender, and race have been collected. Race is 96.3% complete; 2002 subjects (9.1%) are African American. Retrieval of risk factor data, including family history, body mass, smoking status, and diabetes diagnosis is underway. An R03 proposal for data analysis is being drafted for Dec. 2011 submittal. DISCUSSION: This study will develop a model to predict the probability that a subject having a combination of risk factors will have an advanced adenoma detected during screening colonoscopy.