Prediction of Adverse Events in Stable Non-Variceal Gastrointestinal Bleeding Using Machine Learning

Abstract

Clinical risk-scoring systems are important for identifying patients with upper gastrointestinal bleeding (UGIB) who are at a high risk of hemodynamic instability. We developed an algorithm that predicts adverse events in patients with initially stable non-variceal UGIB using machine learning (ML). Using prospective observational registry, 1439 out of 3363 consecutive patients were enrolled. Primary outcomes included adverse events such as mortality, hypotension, and rebleeding within 7 days. Four machine learning algorithms, namely, logistic regression with regularization (LR), random forest classifier (RF), gradient boosting classifier (GB), and voting classifier (VC), were compared with the Glasgow–Blatchford score (GBS) and Rockall scores. The RF model showed the highest accuracies and significant improvement over conventional methods for predicting mortality (area under the curve: RF 0.917 vs. GBS 0.710), but the performance of the VC model was best in hypotension (VC 0.757 vs. GBS 0.668) and rebleeding within 7 days (VC 0.733 vs. GBS 0.694). Clinically significant variables including blood urea nitrogen, albumin, hemoglobin, platelet, prothrombin time, age, and lactate were identified by the global feature importance analysis. These results suggest that ML models will be useful early predictive tools for identifying high-risk patients with initially stable non-variceal UGIB admitted at an emergency department.

Current Status and Future Perspective of Artificial Intelligence in the Management of Peptic Ulcer Bleeding: A Review of Recent Literature

“…In 2020, Seo et al [17] prospectively analyzed 1439 PUB cases to compare the accuracy of ML and conventional scores for PUB patient instability including hypotension, rebleeding, and mortality. Four ML algorithms, namely, logistic regression with regularization, random forest classifier (RF), gradient boosting classifier (GB), and voting classifier (VC), were compared using the GBS and Rockall scores.…”

Feasibility of electronic patient-reported outcome monitoring with self-management advice and warning system in aplastic anemia and paroxysmal nocturnal hemoglobinuria – a pilot study (ePRO-AA-PNH)

“…For inexperienced physicians the tool might be especially helpful to estimate the severity and monitor the disease course, as it is di cult to gain extensive clinical experience in these rare diseases. The multidimensional recording of symptoms, blood values and therapies could also form the basis for an automatic prediction of adverse events and disease course as has been attempted in other elds (34,35), in particular when considering the future bene ts of arti cial intelligence and machine learning.…”

Role of artificial intelligence in multidisciplinary imaging diagnosis of gastrointestinal diseases

“…For example, clinical risk-scoring systems offer invaluable, but not always practical, help for better stratification of patients at risk of UGIB and hemodynamic instability. Seo et al [ 69 ] developed an ML algorithm that predicts adverse events in patients with initially stable non-variceal UGIB[ 69 ]. Primary outcomes analyzed in this study included adverse events such as mortality, low blood pressure and rebleeding within 7 d. The authors compared four ML algorithms (logistic regression with regularization, RF classifier, gradient boosting classifier and voting classifier) with clinical Glasgow–Blatchford and Rockall scores.…”