Background Adequate visibility is an important factor for achieving successful endoscopic hemostasis for the treatment of upper GI bleeding (UGIB). The independent factors that affect visibility during endoscopic procedures have yet to be determined. Objective To determine the factors that affect endoscopic visibility and to create a model that can predict in which patients unacceptable visibility is suspected before emergent endoscopic procedures for UGIB. Design Prospective, observational study. Setting University-affiliated tertiary care hospital in South Korea. Patients A total of 121 patients admitted because of UGIB. Intervention Analysis of the visibility score of the emergency endoscopies for UGIB. Main Outcome Measurements Factors affecting the visibility score of endoscopy and a classification and regression tree (CART) model for predicting of visibility. Results The EGD time and the appearance of the nasogastric (NG) tube aspirate were independent factors that were significantly associated with visibility (EGD time, P <.001; red blood appearance in NG tube aspirate, P <.001; coffee grounds appearance of NG tube aspirate, P =.006). Based on these results, a CART model was developed by using 70 patients who had been allocated to the training set. The CART generated algorithms that proposed the use of the appearance of the NG tube aspirate and the EGD time (8.5 hours) to predict visibility. The sensitivity and specificity for predicting poor visibility were 71.4% and 86.4%, respectively. Conclusion The use of the CART model enables the prediction of which patients will have poor visibility during emergent endoscopy.
All Science Journal Classification (ASJC) codes
- Radiology Nuclear Medicine and imaging