IDDF2020-ABS-0075 Optimizing the use of gastroscope for ICU patients based on machine learning model
Yi Yu · Abstracts · 2020
Background We aim to establish an objective and feasible pre-gastroscopic screening standard to solve the overuse of gastroscopy for ICU patients. Methods This study collected the demographic information, diet, lifestyle, medical history, symptoms, PGI, PGII, G-17 and Hp antibody of the patients from the MIMIC-III and Philips eICU collaboration databases. The decision tree model, logistic regression model, random forest model and support vector machine model were trained by the collected information. The accuracy and validity of the machine learning models predicting positive gastroscopic results were evaluated by comparing the efficiencies of different pre-gastroscopic screening ways. Results 1273 gastroscopic positive cases of a total of 720 cases were enrolled in this study. In the training set, support vector machine model fitted the highest degree (AUC=1.000), the random forest model (AUC=0.941), the decision tree model (AUC is 0.885), and the worst is the Logistic regression model (AUC=0.839). In the test set, four machine learning model has better prediction effect, AUC from high to low were random forest model (0.879), logistic regression model (0.842), the decision tree model (0.827) and support vector machine model (0.826). Assuming risk cut-off value was 0.85, the sensitivity of the model is 93.17%, as well as specificity is 15.70%, and only recommended gastroscopy in 89% of patients, the average 2.27 times gastroscopy can be found that the positive cases. Compared with direct gastroscopy, the efficiency of gastroscopy is increased by 3.57 times after using the screening model. Conclusions This study compared the variables in the model with single-factor analysis results, and proved that the history of upper gastrointestinal polyps, PG II, PG I, Hp antibody, smoking, drinking were important predicting variables for positive gastroscopic results, as well as the single alarm symptom is difficult to predict the results of gastroscope accurately. The model can predict positive gastroscopic risk effectively and provide objective criteria for optimizing the use of gastroscope, which may be a new way to decrease the overuse of gastroscopy for ICU patients. However, before being applied in clinical practice, the models need externally validated.