Risk Factors for Gastrointestinal Bleeding in Patients With Acute Myocardial Infarction: Multicenter Retrospective Cohort Study

Abstract

Background: Gastrointestinal bleeding (GIB) is a severe and potentially life-threatening complication in patients with acute myocardial infarction (AMI), significantly affecting prognosis during hospitalization. Early identification of high-risk patients is essential to reduce complications, improve outcomes, and guide clinical decision-making.

Objective: This study aimed to develop and validate a machine learning (ML)-based model for predicting in-hospital GIB in patients with AMI, identify key risk factors, and evaluate the clinical applicability of the model for risk stratification and decision support.

Methods: A multicenter retrospective cohort study was conducted, including 1910 patients with AMI from the Affiliated Hospital of Guangdong Medical University (2005-2024). Patients were divided into training (n=1575) and testing (n=335) cohorts based on admission dates. For external validation, 1746 patients with AMI were included in the publicly available MIMIC-IV (Medical Information Mart for Intensive Care IV) database. Propensity score matching was adjusted for demographics, and the Boruta algorithm identified key predictors. A total of 7 ML algorithms-logistic regression, k-nearest neighbors, support vector machine, decision tree, random forest (RF), extreme gradient boosting, and neural networks-were trained using 10-fold cross-validation. The models were evaluated for the area under the receiver operating characteristic curve, accuracy, sensitivity, specificity, recall, F_1-score, and decision curve analysis. Shapley additive explanations analysis ranked variable importance. Kaplan-Meier survival analysis evaluated the impact of GIB on short-term survival. Multivariate logistic regression assessed the relationship between coronary heart disease (CHD) and in-hospital GIB after adjusting for clinical variables.

Results: The RF model outperformed other ML models, achieving an area under the receiver operating characteristic curve of 0.77 in the training cohort, 0.77 in the testing cohort, and 0.75 in the validation cohort. Key predictors included red blood cell count, hemoglobin, maximal myoglobin, hematocrit, CHD, and other variables, all of which were strongly associated with GIB risk. Decision curve analysis demonstrated the clinical use of the RF model for early risk stratification. Kaplan-Meier survival analysis showed no significant differences in 7- and 15-day survival rates between patients with AMI with and without GIB (P=.83 for 7-day survival and P=.87 for 15-day survival). Multivariate logistic regression showed that CHD was an independent risk factor for in-hospital GIB (odds ratio 2.79, 95% CI 2.09-3.74). Stratified analyses by sex, age, occupation, marital status, and other subgroups consistently showed that the association between CHD and GIB remained robust across all subgroups.

Conclusions: The ML-based RF model provides a robust and clinically applicable tool for predicting in-hospital GIB in patients with AMI. By leveraging routinely available clinical and laboratory data, the model supports early risk stratification and personalized preventive strategies.

Keywords: acute myocardial infarction; gastrointestinal bleeding; in-hospital; machine learning; prediction model.

Figure 1

Patient selection flowchart for the training, testing, and validation cohorts. A total of 10,046 local patients and 1801 patients from the MIMIC-IV database were screened. After applying exclusion criteria, 1910 patients were included in the local cohort and 1746 patients in the external validation cohort. AMI: acute myocardial infarction; GIB: gastrointestinal bleeding; MIMIC-IV: Medical Information Mart for Intensive Care IV.

Figure 2

Feature selection for predicting gastrointestinal bleeding in patients with acute myocardial infarction using the Boruta algorithm. The horizontal axis displays the names of each variable, while the vertical axis displays the corresponding z scores of each variable. The box plot shows the z score of each variable during model calculation. The blue boxes represent important variables, the red represents tentative variables, and the green represents rejected variables. AG: anion gap; ALB: albumin; ALB/GLO: albumin-to-globulin ratio; ALP: alkaline phosphatase; ALT: alanine aminotransferase; Apob: apolipoprotein B; Apoai: apolipoprotein A-I; APTT: activated partial thromboplastin time; APTTR: activated partial thromboplastin time ratio; Baso: basophils; Ca: calcium; CHD: coronary heart disease; ChE: cholinesterase; Cl: chloride; Crea: creatinine; CysC: cystatin C; DBIL: direct bilirubin; Eos: eosinophils; Fbg: fibrinogen; Glu: glucose; GGT: gamma-glutamyl transferase; GLO: globulin; HDLC: high-density lipoprotein cholesterol; Hb: hemoglobin; HCT: hematocrit; HCY: homocysteine; IBIL: indirect bilirubin; K: potassium; LDH: lactate dehydrogenase; LDLC: low-density lipoprotein cholesterol; Lymph: lymphocytes; Mb max: myoglobin maximum; MCH: mean corpuscular hemoglobin; MCHC: mean corpuscular hemoglobin concentration; MCV: mean corpuscular volume; Mono: monocytes; Na: sodium; Neut: neutrophils; NT proBNP max: N-terminal pro b-type natriuretic peptide maximum; P: phosphorus; PA: prealbumin; PCT: plateletcrit; PLT: platelets; PTR: prothrombin time ratio; PT: prothrombin time; PTINR: prothrombin time international normalized ratio; RBC: red blood cell; RDWCV: red cell distribution width; TG: triglycerides; TC: total cholesterol; TBIL: total bilirubin; TBA: total bile acids; TT: thrombin time; UA: uric acid; WBC: white blood cell.

Figure 3

(A) Receiver operating characteristic curve and (B) decision curve analyses for 7 machine learning models predicting gastrointestinal bleeding in patients with acute myocardial infarction (training cohort). DCtree: decision tree; KNN: k-nearest neighbors; LR: logistic regression; NNET: neural network; RF: random forest; SVM: support vector machine; XGBoost: extreme gradient boosting.

Figure 4

(A) SHAP (Shapley additive explanations) analysis of the top-15 predictors for gastrointestinal bleeding (GIB) in patients with acute myocardial infarction (AMI) using a random forest model ranked by mean absolute SHAP value, and (B) the impact of SHAP values on the occurrence of GIB in patients with AMI. ALP: alkaline phosphatase; CHD: coronary heart disease; DBIL: direct bilirubin; Fbg: fibrinogen; GGT: gamma-glutamyl transferase; GLO: globulin; HCT: hematocrit; Hb: hemoglobin; Hs cTnT Max: high-sensitivity cardiac troponin T maximum; Mb max: myoglobin maximum; NT proBNP max: N-terminal pro b-type natriuretic peptide maximum; P: phosphorus; RBC: red blood cell; TBIL: total bilirubin; UA: uric acid.

Figure 5

Subgroup analysis of coronary heart disease (CHD) impact on gastrointestinal bleeding (GIB) risk in patients with acute myocardial infarction (AMI). These results confirm that CHD is a consistent and significant risk factor for in-hospital GIB in patients with AMI, reinforcing its importance in risk stratification across diverse patient subgroups. OR: odds ratio.