521 Developing and Validating a Prediction Model to Identify Patients at High Risk for Failing Standard Sedation for Routine GI Endoscopy

Abstract

INTRODUCTION: In recent years, use of anesthesia-assisted (AA) sedation for routine GI endoscopy has increased markedly. One factor contributing to this increased use is clinical uncertainty about which patients are most likely to benefit from AA. We aimed to develop a prediction model to help providers identify patients who require AA due to high-risk of failing standard sedation (SS). METHODS: This was an IRB-approved retrospective longitudinal study of national Veterans Affairs (VA) data of all patients who underwent an EGD and/or colonoscopy in 2009-13 (during which AA was used for 4.1% of routine GI endoscopy in VA). Using step-wise logistic regression, we developed a model to predict the primary outcome of failed sedation using potentially predictive patient and procedure-level variables (Table 1). Failed sedation cases were identified electronically as patients who had a routine procedure (without polypectomy, dilation, or injection) with AA within 3-months of a prior similar SS procedure. All cases were manually validated by chart review (index endoscopy report or other records) documenting patient discomfort/agitation and/or medical instability resulting in either an incomplete procedure or specific mention of need for AA for future procedures. 10-fold cross validation was used to validate the final model. RESULTS: Of 710,252 SS procedures, there were 398 cases of failed sedation (173 EGD, 225 colonoscopy; prevalence = 0.06%). None of the factors found to be significant predictors of failed sedation (e.g., high-dose opioid use, younger age) had an odds ratio > 3; factors NOT associated with failed sedation included body mass index, obstructive sleep apnea, functional GI disorders, and others (Table 1). 10-fold cross-validation yielded a mean area under the curve (AUC) of 0.67 (range 0.63-0.71). The model was well-calibrated, but patients in the top decile of risk had a failed sedation prevalence of only 0.16% (Table 2). CONCLUSION: Failed sedation among patients undergoing routine outpatient GI endoscopy with SS is very rare, even among patients in the highest decile of risk. This suggests that concerns regarding failed sedation due to commonly cited factors such as chronic opioid use, obesity, or increased comorbidity do not justify forgoing SS in favor of AA in most patients. It also suggests that use of AA is generally unnecessary, since failed sedation is very uncommon even in VA where overall use of AA is much lower than in the private sector.