162 Development of a New Bowel Preparation Prediction Model Based on Patient's Questionnaire: To Proceed or Not to Proceed With Colonoscopy? A Study From Veterans Affairs Hospital to Stratify High Risk Population

Abstract

INTRODUCTION: Bowel preparation (BP) is one of the intra-procedural quality indicators. The percentage of poor BP leading to repeat colonoscopy in = or < 1 year should not exceed more than 15%. The challenge remains that before doing the colonoscopy, we cannot reliably predict, whether a BP will be adequate for a good quality examination? The aim of our study is to propose a model based on patient's questionnaire that can identify patients with poor BP in pre-operative (pre-op) area prior to their colonoscopy. METHODS: From November 2018 to May 2019, 461 consecutive patients of 18 years of age or older who presented to Veterans Affairs (VA) Nebraska-Western Iowa Health Care System for colonoscopy each completed a voluntary questionnaire in the pre-op prior to their colonoscopies. This study was approved by VA institutional review board (IRB#00163). Information was collected regarding gender, age, ethnicity, education level, first colonoscopy, days of BP, BP instructions, amount of BP completed, last solid meal, timing of last bowel movement (BM), travel time to VA, nausea/vomiting and abdominal pain/bloating while taking BP, Bristol stool scale and last BM stool character (A-clear liquid stool, B-turbid, C-soft stools). Information regarding the Boston bowel preparation scale (BBPS) was calculated during the colonoscopy. Accuracy of the model was measured by the area under ROC curve. Statistical analysis was performed using STATA version 14.2. RESULTS: 461 patients were included (94% males, 92% Caucasians). BBPS score of 5 or greater was noted in 95% of the patients. We used a multivariable logistic regression model using most of the variables depicted in Table 1 to predict the outcome of a patient having a BBPS score of 5 or greater. Area under the ROC curve was 0.7675 (Figure 1). With a cut point for the predicted probability of having a Boston Prep Score of 5 or greater of ≥0.8935, we achieved a sensitivity of 87.4% and a specificity of 50%. The coefficients included in the predicted model are depicted in Table 2. CONCLUSION: Our new bowel preparation prediction model to predict probability of having a BBPS of 5 or greater in pre-op area achieved sensitivity of 87.4% and specificity of 50% with area under ROC = 0.7675. Only 5% of our patient population had BBPS < 5 which could explain the low specificity of our predicted model. Multi-center studies are needed to further validate our model.