[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to main content
Download PDF

Risk factors and prediction model for acute ischemic stroke after off-pump coronary artery bypass grafting based on Bayesian network

BMC Medical Informatics and Decision Making volume 24, Article number: 349 (2024) Cite this article

Abstract

Background

This study aimed to identify the risk factors of acute ischemic stroke (AIS) occurring during hospitalization in patients following off-pump coronary artery bypass grafting (OPCABG) and utilize Bayesian network (BN) methods to establish predictive models for this disease.

Methods

Data were collected from the electronic health records of adult patients who underwent OPCABG at Beijing Anzhen Hospital from January 2018 to December 2022. Patients were allocated to the training and test sets in an 8:2 ratio according to the principle of randomness. Subsequently, a BN model was established using the training dataset and validated against the testing dataset. The BN model was developed using a tabu search algorithm. Finally, receiver operating characteristic (ROC) and calibration curves were plotted to assess the extent of disparity in predictive performance between the BN and logistic models.

Results

A total of 10,184 patients (mean (SD) age, 62.45 (8.7) years; 2524 (24.7%) females) were enrolled, including 151 (1.5%) with AIS and 10,033 (98.5%) without AIS. Female sex, history of ischemic stroke, severe carotid artery stenosis, high glycated albumin (GA) levels, high D-dimer levels, high erythrocyte distribution width (RDW), and high blood urea nitrogen (BUN) levels were strongly associated with AIS. Type 2 diabetes mellitus (T2DM) was indirectly linked to AIS through GA and BUN. The BN models exhibited superior performance to logistic regression in both the training and testing sets, achieving accuracies of 72.64% and 71.48%, area under the curve (AUC) of 0.899 (95% confidence interval (CI), 0.876–0.921) and 0.852 (95% CI, 0.769–0.935), sensitivities of 91.87% and 89.29%, and specificities of 72.35% and 71.24% (using the optimal cut-off), respectively.

Conclusion

Female gender, IS history, carotid stenosis (> 70%), RDW-CV, GA, D-dimer, BUN, and T2DM are potential predictors of IS in our Chinese cohort. The BN model demonstrated greater efficiency than the logistic regression model. Hence, employing BN models could be conducive to the early diagnosis and prevention of AIS after OPCABG.

Peer Review reports

Background

Coronary artery bypass grafting (CABG) is a well-established method for myocardial revascularization in patients with multivessel disease and diabetes [1]. Notably, off-pump coronary artery bypass grafting (OPCABG) has emerged as the predominant surgical approach because it avoids cardiac arrest and mitigates liver and kidney function impairment. However, acute ischemic stroke (AIS) remains one of the most serious complications of OPCABG, increasing the length of stay and mortality during hospitalization [2, 3], with an incidence ranging from 1.1 to 5.7% [4, 5]. This complication significantly impacts patients’ quality of life and survival time post-surgery [6, 7]. The occurrence of AIS imposes a significant healthcare burden and affects the allocation of medical resources. Therefore, developing effective predictive tools to identify high-risk patients is crucial, as they can facilitate targeted preoperative interventions and assist clinicians in making informed postoperative decisions, ultimately reducing complications.

Several clinical tools, such as the Society of Thoracic Surgeons (STS) models [8] and the European System for Cardiac Operative Risk Evaluation (EuroSCORE) [9], have been used to assess the risk of postoperative complications of CABG. These models rely primarily on preoperative baseline patient characteristics and accessory examinations to guide clinical decision-making. However, these models are not specifically designed to predict AIS following OPCABG. The EuroSCORE primarily focuses on mortality rates, whereas the STS model only assesses the overall probability of stroke without distinguishing between ischemic and hemorrhagic types [8]. Additionally, most of these traditional models are constructed using logistic regression, which assumes that features are independent, but the risk factors associated with AIS after CABG are diverse, and interactions between these factors cannot be ruled out [10]. This highlights a critical gap in the existing predictive models, as they may overlook the complex interdependencies among risk factors, thereby suggesting the limitations of these traditional scoring systems in predicting OPCABG outcomes. So far, several studies have been carried out in order to establish machine learning models for predicting complications such as acute kidney injury and atrial fibrillation following CABG [11, 12], but there is still a relative paucity of research focused on predicting AIS. Therefore, it is imperative to use machine learning methods to develop a more effective clinical prediction model specifically for AIS after OPCABG.

In this context, Bayesian Networks (BN) are particularly suitable for our study as they can effectively manage missing data, model complex interdependencies among multiple risk factors, and provide interpretable results [13], while many other machine learning models have structural or tuning parameters that cannot be directly estimated from the data. Moreover, BN’s probabilistic framework allows for the exploration of variable relationships that traditional models may overlook, thereby offering valuable insights for clinical decision-making. To address the gap in predictive modeling for AIS in post-OPCABG patients, we established a BN model and compared its performance with traditional logistic regression approaches.

Methods

Setting and data sources

This retrospective study included adult patients who underwent CABG at the Beijing Anzhen Hospital between January 2018 and December 2022. The inclusion criteria were as follows: (1) age ≥ 18 years; (2) patients who had received CABG surgery. The exclusion criteria were as follows: (1) CABG under cardiopulmonary bypass; (2) CABG combined with other operations; (3) intraoperative suspension without completion of the operation; (4) incomplete relevant data.

Data were obtained from electronic health records and included demographic information, medical history, comorbidities, vascular ultrasound, electrocardiography, laboratory data, surgical information, and discharge status. This study adhered to the principles of the Declaration of Helsinki and was approved by the Ethics Committee of the Beijing Anzhen Hospital, Capital Medical University.

Outcome variables

AIS was defined as the sudden onset of global or focal neurological deficits persisting for more than 24 h, as confirmed by computed tomography and/or magnetic resonance imaging upon initial presentation [14]. Postoperative diagnosis of AIS was initially established by the surgical team and subsequently validated by two neurologists according to clinical presentation and imaging findings. To minimize variability in imaging interpretation, a standardized protocol based on established practice recommendations or guidelines [15, 16] was followed for the radiologist’s review, including predefined criteria for identifying ischemic lesions and structured reporting formats. Imaging reports were thoroughly completed and reviewed by a radiologist.

Quality control

The data collection process adhered to customized standardized protocols, and the researchers were well acquainted with these procedures prior to commencing data retrieval. Data entry adhered to the double entry method, and the accuracy of the double-entry process was assessed by comparing the two independent entries for inconsistencies. In case of discrepancies during the review process, consultation with medical records was performed, and necessary data corrections were made. Additionally, post-entry quality checks were conducted, including random audits of the dataset to ensure accuracy and completeness.

Data processing for predictive variables

To establish the predictive model, we processed the data for predictive variables following the criteria that combines relevant literature review and exclusion rule for missing data. Through a comprehensive review of the recent literature, old age, stroke history, unstable angina pectoris, peripheral vascular disease, carotid artery stenosis, and atrial fibrillation are shown to be independent risk factors for postoperative ischemic stroke (IS). By contrast, the results for chronic renal failure, previous hypertension, congestive heart failure, diabetes, and gender are not consistent in the literature [7, 17]. A recent retrospective cohort study of 36,898 participants revealed that factors such as female sex, advanced age, prior cardiac surgery, and history of stroke, peripheral vascular disease, heart failure, atrial fibrillation, and diabetes were significantly associated with acute IS within 30 days after CABG [6]. Furthermore, Tatlisuluoglu et al. reported that lower hemoglobin and lymphocyte count and higher red cell distribution width (RDW), platelet-lymphocyte ratio (PLR), and neutrophil to lymphocyte ratio (NLR) were associated with perioperative IS [18]. Ultimately, we assembled a dataset of 47 variables encompassing traditional AIS risk factors, such as age, sex, and IS history, as well as additional clinically relevant variables used in cardiac surgery evaluations. Regarding the approach to handle missing data, we adhered to guidelines from biostatistical literature, which suggest that if the rate of missing data in a dataset exceeds 30%, it compromises the data’s confidence level. Thus, in our study, instances were excluded from the dataset once their missing attributes > 15 (15 of 47).

We screened for potential risk factors for AIS using univariate logistic regression analysis. Variables significantly associated with the outcome (P < 0.05) were subsequently utilized in least absolute shrinkage and selection operator (LASSO) regression algorithm [19]. Dummy variables were generated for categorical features. Cross-validation determined optimal tuning parameters (λ) for LASSO regression. The data were randomly divided, resulting in the creation of an 80% training set and 20% testing set. AIS prediction was conducted using both BN and logistic regression models as the primary analytical tools. The construction of the BN model involved the use of a Tabu Search algorithm. The model performance was assessed using the test sets. AIS-related variables were quantified and encoded before constructing the BN model.

To construct a binary classification prediction model, it is advisable to ensure that the sample size is at least ten times greater than the number of independent variables. We incorporated 13 independent variables into the multivariate analysis, resulting in a minimum sample size of 130 individuals per group. Notably, 151 patients with AIS and 10,033 patients without AIS underwent OPCABG surgery were included in this study, indicating that our sample size was sufficient to develop the prediction model.

Bayesian network

The BN model is illustrated in the form of a directed acyclic graph [20]. The nodes within the graph symbolize random variables and the presence of directed edges indicates the existence of probabilistic relationships among the variables encompassed by the model. If a directed arc exists from X1 to X2, it represents X1 leading to X2, where X1 is the parent node, and X2 is the child node. The conditional probability distribution table linked to each parent node illustrates this state. BN represents the joint probability distributions of random variables X = {X1,.,Xn}, allowing for the derivation of probability expressions as follows:

$$\eqalign{& P\left( {{X_1}, \ldots ,{X_n}} \right) = \cr & P\left( {{X_1}} \right)P\left( {\left. {{X_2}} \right|{X_1}} \right) \ldots P\left( {\left. {{X_n}} \right|{X_1},{X_2}, \ldots {X_{n - 1}}} \right) = \cr & \mathop \prod \limits^n P\left( {\left. {{X_i}} \right|\pi \left( {{X_i}} \right)} \right) \cr}$$

where π (Xi) denotes the set of parent nodes of Xi; π (Xi) ⊆ {X1,...,Xi−1} [21].

In this study, a gathered dataset was used to formulate a BN model aimed at forecasting the incidence of AIS. A total of 47 random variables were extracted from patient data for each instance. Initially, we employed univariate logistic regression to screen for potential risk factors associated with AIS. This approach allowed us to evaluate the relationship between each independent variable and the outcome individually. We then applied the LASSO regression to filter the nodes, in order to reduce complexity in the network structure. The optimal model was subsequently established using the tabu search algorithm.

Statistical analysis

Statistical analyses were conducted using SPSS Statistics for Windows (version 23.0). Continuous variables were delineated through the utilization of either mean ± standard deviation (SD) or median (interquartile range), whereas categorical variables were depicted as counts (n) and percentages (%). The assessment of categorical variables involved the utilization of the χ2 test, while for normally and non-normally distributed data, Student’s t-test and Mann–Whitney U-test were applied, respectively. Variables correlated with AIS were evaluated using univariate binary logistic regression and LASSO regression analyses. The filtered variables were regarded as potential candidates for further investigation in subsequent multivariate analyses. The predictive models were evaluated and the area under the curve (AUC) was determined using receiver operating characteristic (ROC) curve analysis. The AUC was calculated using the roc function from the pROC package in R, which provided both the AUC values and their corresponding confidence intervals. The performance of the developed models was assessed using Youden analysis, a statistical measure often adopted to select the best classification threshold and assess the best cut-off [22]. The optimal cutoffs were established using the training set, and sensitivity and specificity were subsequently calculated. A separate validation set was employed to confirm the robustness of the cutoffs and model performance. The DeLong test was used to evaluate the statistical significance of the disparity among the AUC values. The Hosmer–Lemeshow (HL) test was utilized to evaluate the disparity between the predicted and true values and generate a calibration curve. The BN model was constructed using the Tabu Search algorithm, whereas the parameter estimation for the BN model utilized the maximum-likelihood estimation method. The BN model was established using RStudio 4.3.2. LASSO regression analysis was performed using the “glmnet” package in R. To visualize the topology of the BN, we employed Netica32 software (Norsys Software Corp, Vancouver, Canada).

Results

Patient selection

Of the 22,558 patients who underwent CABG, 12,374 were ineligible for the study based on the eligibility criteria and were excluded. A total of 10,184 patients (mean (SD) age, 62.45 (8.7) years; 2,524 (24.7%) females) were enrolled, of which 151 (1.5%) had AIS and 10,033 (98.5%) did not (Fig. 1). Table 1 shows the characteristics of the two groups.

Fig. 1
figure 1

Flowchart describing the screening of patients underwent CABG. CABG, coronary artery bypass grafting

Full size image
Table 1 Baseline data of patients undergoing isolated OPCABG
Full size table

Risk factors for AIS

Of the 47 variables, 18 were associated with AIS by univariate logistic regression, namely: age (odds ratio (OR), 1.035; 95% confidence interval (CI), 1.015–1.056; P = 0.001), female sex (OR, 2.265; 95% CI, 1.635–3.137; P < 0.001), T2DM (OR, 2.252; 95% CI, 1.621–3.127; P < 0.001), hypertension (OR, 1.874; 95% CI, 1.271–2.763; P = 0.002), history of IS (OR, 17.563; 95% CI, 10.585–29.142; P < 0.001), carotid stenosis > 70% (OR, 4.234; 95% CI, 2.813–6.372; P < 0.001), erythrocyte (OR, 0.723; 95% CI, 0.581–0.900; P = 0.004), hemoglobin (OR, 0.986; 95% CI, 0.979–0.993; P < 0.001), red blood cell distribution width-coefficient of variation (RDW-CV) (OR, 1.241; 95% CI, 1.167–1.320; P < 0.001), BUN (OR, 1.079; 95% CI, 1.029–1.131; P = 0.002), fasting blood glucose (FBG) (OR, 1.099; 95% CI, 1.055–1.145; P < 0.001), glycated albumin (GA) (OR, 1.062; 95% CI, 1.030–1.095; P < 0.001), total protein (TP) (OR, 1.045; 95% CI, 1.017–1.074; P = 0.002), AGR (OR, 0.291; 95% CI, 0.152–0.555; P < 0.001), B-type natriuretic peptide (BNP) (OR, 1.001; 95% CI, 1.000–1.001; P = 0.001), fibrin degradation products (FDP) (OR, 1.023; 95% CI, 1.010–1.037; P < 0.001), plasma fibrinogen (PF) (OR, 1.478; 95% CI, 1.276–1.712; P < 0.001), and D-dimer (OR, 1.000; 95% CI, 1.000–1.000; P < 0.001).

Variable selection in LASSO regression

LASSO regression was used to further refine the initial set of 18 variables. The determination of the tuning parameter λ was conducted via 10-fold cross-validation, as illustrated in Fig. 2B. The optimal λ value was chosen to minimize cross-validation error. At log (λ) = ‒7.366, 13 variables from the original 18 were selected, which included T2DM, female sex, IS history, carotid stenosis > 70%, erythrocyte count, hemoglobin level, RDW-CV, BUN, FBG, GA, TP, PF, and D-dimer. Figure 2A illustrates the coefficient profiles of the candidate variables obtained from the LASSO regression. We then incorporated these variables into multivariate logistic regression models. Seven variables demonstrating statistical significance were integrated, including female sex (OR, 1.866; 95% CI, 1.240–2.809; P = 0.003), IS history (OR, 17.966; 95% CI, 9.995–32.294; P < 0.001), RDW-CV (OR, 1.261; 95% CI, 1.158–1.373; P < 0.001), TP (OR, 1.058; 95% CI, 1.023–1.094; P = 0.001), PF (OR, 1.442; 95% CI, 1.188–1.751; P < 0.001), D-dimer (OR, 1.000; 95% CI, 1.000–1.000; P = 0.001), and carotid stenosis > 70% (OR, 3.529; 95% CI, 2.194–5.674; P < 0.001) (Table 2).

Fig. 2
figure 2

Variable selection by using the LASSO regression. A coefficient profile plot was produced against the log(lambda) sequence (A). Thirteen variables with nonzero coefficients were selected by optimal lambda. The optimal parameter (lambda) in the LASSO model was verified by plotting the binomial deviance curve against log(lambda), with dotted vertical lines indicating the minimum criteria and one standard error from the minimum(B). LASSO regression, least absolute shrinkage and selection operator regression

Full size image
Table 2 Risk factors of AIS in patients underwent isolated OPCABG: multivariate binary logistic regression analysis
Full size table

BN structure

Using BN analysis, we delineated the probabilistic dependencies between the AIS and its predictors in a complex network (Fig. 3). The nodes directly linked to AIS included female sex, IS history, carotid stenosis (> 70%), RDW-CV, GA, D-dimer, and BUN, whereas the other variables were indirectly related to AIS. For example, T2DM was indirectly linked to AIS via GA. The conditional probabilities for each node in the network (Table S1) were estimated using maximum likelihood estimation. Common variables predicting AIS included female sex, IS history, carotid stenosis > 70%, RDW-CV, GA, D-dimer, and BUN.

Fig. 3
figure 3

BN for prediction of occurrence of AIS in patients underwent isolated OPCABG. BN, Bayesian network; OPCABG, off-pump coronary artery bypass grafting; AIS, acute ischemic stroke; IS, ischemic stroke; T2DM, type 2 diabetes mellitus; RDW-CV, red cell distribution width coefficient of variation; BUN, blood urea nitrogen; FBG, fasting blood glucose; GA, glycated albumin; TP, total protein; FDP, fibrin degradation products; PF, plasma fibrinogen

Full size image

Model performance evaluation

The performance of the two models was evaluated by assessing multiple metrics, including accuracy, AUC, specificity, sensitivity, and calibration curve. The logistic regression predictive model exhibited accuracies of 72.44% and 72.21%, AUCs of 0.876 (95% CI, 0.846–0.907) and 0.847 (95% CI, 0.799–0.895), sensitivities of 87.80% and 85.71%, and specificities of 72.21% and 72.03% (using an optimal risk cutoff of 0.016) in the individual training and testing datasets, respectively. However, the BN model exhibited accuracies of 72.64% and 71.48%, AUCs of 0.899 (95% CI, 0.876–0.921) and 0.852 (95% CI, 0.769–0.935), sensitivities of 91.87% and 89.29%, and specificities of 72.35% and 71.24% (using the optimal risk cut-off of 0.013). Table 3 presents the results of the two models. No statistically significant discrepancies were observed in the AUC values between the logistic regression and BN models (Fig. 4). The Delong test was conducted in the training and testing cohorts, yielding P > 0.05 (P = 0.2465 and P = 0.9229, respectively). Additionally, the calibration curves illustrate the superior performance of the BN model over the logistic model in aligning the actual and predicted probabilities (Fig. 5). The logistic regression and BN models exhibited satisfactory calibration in both the training (BN: P = 0.999, χ2 < 0.001, degrees of freedom (df) = 5; logistic regression: P = 0.3008, χ2 = 9.515, df = 8) and testing (BN: P = 0.2279, χ2 = 6.904, df = 5; logistic regression: P = 0.6470, χ2 = 6.002, df = 8) datasets, as confirmed by the Hosmer–Lemeshow test.

Fig. 4
figure 4

ROC curves of BN model and LR model for prediction of in patients undergoing isolated OPCABG. The AUCs of the BN model for prediction of AIS was 0.899 (95% CI, 0.876–0.921) and 0.852 (95% CI, 0.769–0.935), as well as 0.876 (95% CI, 0.846–0.907) and 0.847 (95% CI, 0.799–0.895) of the LR model in training and test sets, respectively. ROC, Receiver operating characteristic; BN, Bayesian network; LR, logistic regression; OPCABG, off-pump coronary artery bypass grafting; AUCs areas under the curve; AIS, acute ischemic stroke

Full size image
Fig. 5
figure 5

Calibration of the predictive models. The calibration plots demonstrated that the predictive risk of AIS was in good agreement with the actually observed results, in either BN model of (A) train and (B) test sets, or in LR model of (C) train and (D) test sets. AIS, acute ischemic stroke; BN, Bayesian network; LR logistic regression

Full size image
Table 3 The performance of different predictive models
Full size table

Discussion

This study employed univariate logistic regression and LASSO regression to identify the primary risk factors for AIS in patients who underwent OPCABG during hospitalization. Subsequently, a BN model was developed to estimate the conditional probability of each node using the tabu search algorithm. Our study highlighted that risk factors such as female sex, IS history, carotid stenosis (> 70%), RDW-CV, GA, D-dimer, and BUN were directly associated with the occurrence of postoperative AIS. Additionally, T2DM was indirectly associated with AIS through GA and BUN. Furthermore, the BN model exhibited superior or comparable predictive performance for AIS compared to the logistic regression model.

Our study revealed that sex significantly influenced perioperative stroke occurrence, with females exhibiting a significantly higher risk of AIS than males, consistent with previous research findings [6, 7, 23,24,25,26]. William et al. observed a clear association between female sex and higher 30-day mortality rates [24]. It has been suggested that this difference could be attributed to anatomical and physiological factors. Therefore, adequate postoperative attention should be paid to female patients to mitigate these risks.

Patients’ baseline characteristics, such as a history of IS and severe carotid stenosis (stenosis rate > 70%), two well-known traditional risk factors [6, 17, 26,27,28,29], were directly associated with AIS. Current pre-and intraoperative interventions for carotid artery stenosis mainly consist of drug treatment and revascularization procedures, including carotid artery stenting (CAS) and carotid endarterectomy (CEA). However, the effectiveness of these revascularization techniques in mitigating postoperative stroke in patients with severe carotid artery stenosis remains uncertain [28, 30]. Hence, a deeper understanding of these techniques’ efficacy could enhance risk assessments for AIS, inform clinical practice, and guide future research directions.

Consistent with previous studies [7, 17], T2DM was indirectly associated with postoperative AIS. Glycated albumin reflects the average blood glucose level over the previous 2–3 weeks, making it a valuable indicator for diabetes evaluation and diagnosis [31]. Multiple studies indicate a correlation between elevated glycemic levels and both stroke incidence and unfavorable short-term outcomes post-stroke [32,33,34]. Limited research has investigated the impact of GA levels on post-CABG stroke incidence. Our BN analysis revealed a direct relationship between GA and AIS. Notably, two studies found that patients with elevated GA levels exhibited reduced sensitivity to antiplatelet therapy, suggesting this is a potential cause of increased AIS risk associated with high GA levels [35, 36]. Therefore, managing preoperative blood glucose levels is crucial for reducing postoperative AIS probability. Based on the above findings, specific recommendations were proposed that serum GA levels should be controlled within the normal range (< 17.1%) to mitigate this risk.

In our study, high BUN levels were associated with an increased incidence of AIS. Peng et al. similarly found that elevated BUN levels correlated with a higher risk of IS [37]. The mechanism by which BUN affects AIS is generally considered to involve increased insulin resistance and inhibited insulin secretion, leading to diabetes. Additional mechanisms, such as increased islet protein O-GlcNAcylation and impaired glycolysis, also warrant further exploration [37,38,39]. Our model indicates that BUN could indirectly influence AIS occurrence through the diabetes-GA pathway, potentially supporting these mechanisms.

RDW has been identified as an independent risk factor for predicting AIS after CABG [18, 40]. Our study corroborates this finding, showing a strong association between RDW and AIS risk. As a fast, simple, and inexpensive biomarker, RDW has significant predictive value not only for AIS following CABG but also for patients with transient ischemic attack (TIA) [41]. Although the pathological mechanism by which elevated RDW affects AIS is largely unknown, it is speculated to be related to the inflammatory state and oxidative stress damage reflected by RDW [41, 42]. Additionally, the other related variables like hemoglobin and erythrocyte count have ORs close to 1, indicative of their limited direct associations with AIS. Although the individual contributions of these variables may appear minimal, they were deemed relevant in the context of the overall risk prediction model. Their inclusion helps control for potential confounding effects and ensures a more comprehensive evaluation of risk factors for AIS. Nevertheless, future studies with larger sample sizes may further elucidate their role and significance in predicting AIS in diverse populations.

Currently, there is limited data on the influence of D-dimer levels on AIS following CABG. Given that patients requiring CABG often present with heart failure preoperatively, we reviewed data showing that elevated plasma D-dimer levels in heart failure patients upon admission were significantly correlated with an increased incidence of IS during hospitalization and within 30 days post-admission [43]. Our study also demonstrated a direct correlation between D-dimer and postoperative AIS. Additionally, D-dimer has implications for screening patients for acute cerebral infarction post-surgery due to subtle embolic causes, such as occult atrial fibrillation, malignant tumors, and venous thromboembolism [44, 45]. Therefore, D-dimer could serve as a potential predictive biomarker for AIS after CABG.

BN models offer numerous advantages in the medical field. Unlike logistic regression models, BN can handle missing data, a common issue in clinical practice, enhancing their suitability for developing diagnostic models. Additionally, BN models typically represent domain structures, making the results easily interpretable to healthcare professionals. BN models are crucial for quantitatively assessing clinically relevant outcomes. For example, data presented in Table S1 show that the probability of AIS in female patients with a history of stroke, normal RDW, BUN, GA, and D-dimer levels, and no severe carotid artery stenosis was 0.036. This probability increased to 0.154 with severe carotid stenosis and further to 0.2 with high BUN levels. These findings can be readily applied in clinical settings, contributing to early detection and diagnosis of IS.

As shown in Table 4, an example application is provided to illustrate the practical application of this BN model. In Case 1, a 67-year-old male patient with a history of stroke and severe carotid stenosis have a 0.11 probability of developing AIS. Based on this risk assessment, the clinical team took proactive measures, such as monitoring blood pressure and ensuring adequate blood volume, finally preventing the occurrence of AIS. In Case 2, a 62-year-old female patient with elevated D-dimer levels was predicted to have a lower probability (0.01) of developing AIS. Thus, the clinical team reduced the duration of antithrombotic discontinuation, and consequently, no AIS occurred. In Case 3, a 70-year-old female patient with a history of stroke and severe carotid stenosis had a 0.15 probability of developing AIS. Based on the prediction information, a closer postoperative monitoring was implemented to facilitate the timely detection of AIS. Afterwards, the patient was transferred to a specialized intensive care unit for further therapy, and consequently, her neurological function was significantly improved after stroke treatment. In a word, providing these real-world examples will better illustrate the model’s practical value and reliability in identifying high-risk cases in clinical settings, ultimately leading to improved patient outcomes.

Table 4 BN model related application examples
Full size table

To enhance the clinical utility of our findings, we propose several actionable interventions for clinicians: (1)strengthen preoperative assessments, focusing on high-risk factors such as female sex, a history of IS, and severe carotid stenosis to develop personalized surgical plans; (2)optimize blood glucose management, particularly in diabetic patients, to lower the risk of postoperative AIS; (3)encourage multidisciplinary collaboration among surgical teams, internists, neurologists, and anesthesiologists to create comprehensive preoperative and postoperative care strategies. (4)implement rigorous monitoring for high-risk patients post-surgery and timely intervention upon early symptom detection. Additionally, surgeon experience, technique, and decision-making can also impact the risk of postoperative stroke, but currently these surgeon-related factors are determined by many unknowns and therein lies the opportunity to study. More evidence is needed to address this issue in the future.

The calibration results of this study demonstrate consistency between the BN and logistic regression models. Furthermore, the analysis shows that the BN model exhibited superior or comparable performance to traditional logistic regression models in terms of accuracy, AUC, specificity, and sensitivity. This is noteworthy because logistic regression models assume independence among variables, a limitation that overlooks intervariable relationships. In contrast, BN models construct network models by extensively mining data to uncover variable interactions, effectively utilizing available data and revealing valuable insights into complex interactions among multiple factors.

Conclusion

In conclusion, this study represents the first BN model aimed at forecasting AIS occurrence in patients following isolated OPCABG, demonstrating enhanced efficiency compared to the logistic regression model. T2DM, female sex, history of IS, carotid stenosis (> 70%), RDW-CV, GA, D-dimer, and BUN were potential predictors of AIS in a Chinese study cohort. Further research is warranted to explore the risk factors for AIS in patients who have undergone isolated OPCABG and to elucidate their causal relationships. Such investigations can aid in refining disease prevention strategies, facilitating early detection and treatment of patients with AIS, thereby improving patient prognosis and reducing economic burdens for this patient population.

Limitations

This study has several limitations. First, the incidence of perioperative stroke following CABG was very low; thus, despite a total sample size exceeding 10,000 cases, the limited number of positive cases may hinder robust statistical analysis and affect the overall statistical power and reliability of the findings. Second, the study was a retrospective analysis conducted at a single center, relying exclusively on data from this institution, primarily in northern China, which may have introduced selection bias and limited the generalizability of the results. Third, due to constraints in data extraction from hospital electronic systems, our primary outcome was the occurrence of in-hospital AIS after OPCABG, rather than within 30 days post-OPCABG. Finally, the arcs in BNs indicate probabilistic dependence rather than causality, suggesting that causal relationships require further validation in future prospective studies.

Future work

Future research should focus on validating the causal relationships identified in our BN analysis through prospective studies. Expanding the research to include multicenter studies across diverse regions, external validation will enhance the generalizability of our findings to other populations or clinical settings. Furthermore, there is a need for a comprehensive comparison between the BN model and traditional models, such as the STS model, to assess their relative predictive performance. Investigating the effectiveness of targeted interventions, such as preoperative blood glucose management, and exploring other potential risk factors—such as family history, environmental influences, and lifestyle choices—will provide a more comprehensive understanding of the factors contributing to the occurrence of AIS in patients following OPCABG.

Data availability

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Abbreviations

AIS:

Acute ischemic stroke

CABG:

Coronary artery bypass grafting

OPCABG:

off-pump coronary artery bypass grafting

T2DM:

Type 2 diabetes mellitus

BN:

Bayesian network

MI:

Myocardial infarction

AMI:

Acute myocardial infarction

IS:

Ischemic stroke

LVEF:

Left ventricular ejection fraction

BMI:

Body mass index

RDW-CV:

Red cell distribution width coefficient of variation

Hs-CRP:

High-sensitivity C-reactive protein

ALT:

Alanine aminotransferase

AST:

Aspartate aminotransferase

TG:

Triglyceride

TC:

Total cholesterol

HDL-C:

High-density lipoprotein cholesterol

LDL-C:

Low-density lipoprotein cholesterol

BUN:

Blood urea nitrogen

UA:

Uric acid

FBG:

Fasting blood glucose

GA:

Glycated albumin

HCY:

Homocysteine

CK:

Creatine kinase

CK-MB:

Creatine kinase-MB

LDH:

Lactate dehydrogenase

Mb:

Myoglobin

TP:

Total protein

AGR:

Albumin-globulin ratio

BNP:

B-type natriuretic peptide

PT:

Prothrombin time

APTT:

Activated partial thromboplastin time

INR:

International normalized ratio

FDP:

Fibrin degradation products

PF:

Plasma fibrinogen

References

  1. Sun LY, Gaudino M, Chen RJ, Bader Eddeen A, Ruel M. Long-term outcomes in patients with severely reduced left ventricular ejection fraction undergoing percutaneous coronary intervention vs coronary artery bypass grafting. JAMA Cardiol. 2020;5:631–41.

    Article  PubMed  PubMed Central  Google Scholar 

  2. Bakaeen FG, Gaudino M, Whitman G, Doenst T, Ruel M, Taggart DP, Stulak JM, Benedetto U, Anyanwu A, Chikwe J, Bozkurt B, Puskas JD, Silvestry SC, Velazquez E, Slaughter MS, McCarthy PM, Soltesz EG, Moon MR. 2021: The american association for thoracic surgery expert consensus document: Coronary artery bypass grafting in patients with ischemic cardiomyopathy and heart failure. J Thorac Cardiovasc Surg. 2021; 162: 829 – 50.e1.

  3. Head SJ, Milojevic M, Domanski MJ, Farkouh ME, Kappetein AP: Reply. Stroke risk following anaortic off-pump cabg versus pci. J Am Coll Cardiol. 2018;72:2681.

    Article  PubMed  Google Scholar 

  4. Lorusso R, Moscarelli M, Di Franco A, Grazioli V, Nicolini F, Gherli T, De Bonis M, Taramasso M, Villa E, Troise G, Scrofani R, Antona C, Mariscalco G, Beghi C, Miceli A, Glauber M, Ranucci M, De Vincentiis C, Gaudino M. Association between coronary artery bypass surgical techniques and postoperative stroke. J Am Heart Assoc. 2019;8:e013650.

    Article  PubMed  PubMed Central  Google Scholar 

  5. Mohamed MO, Hirji S, Mohamed W, Percy E, Braidley P, Chung J, Aranki S, Mamas MA. Incidence and predictors of postoperative ischemic stroke after coronary artery bypass grafting. Int J Clin Pract. 2021;75:e14067.

    Article  PubMed  Google Scholar 

  6. Jonsson K, Barbu M, Nielsen SJ, Hafsteinsdottir B, Gudbjartsson T, Jensen EM, Silverborn M, Jeppsson A. Perioperative stroke and survival in coronary artery bypass grafting patients: a swedeheart study. Eur J Cardiothorac Surg. 2022;62:ezac025.

  7. Sultan I, Bianco V, Kilic A, Jovin T, Jadhav A, Jankowitz B, Aranda-Michel E, D’Angelo MP, Navid F, Wang Y, Thoma F, Gleason TG. Predictors and outcomes of ischemic stroke after cardiac surgery. Ann Thorac Surg. 2020;110:448–56.

    Article  PubMed  Google Scholar 

  8. Shahian DM, Jacobs JP, Badhwar V, Kurlansky PA, Furnary AP, Cleveland JC Jr., Lobdell KW, Vassileva C, Wyler von Ballmoos MC, Thourani VH, Rankin JS, Edgerton JR, D’Agostino RS, Desai ND, Feng L, He X, O’Brien SM. The society of thoracic surgeons 2018 adult cardiac surgery risk models: part 1-background, design considerations, and model development. Ann Thorac Surg. 2018;105:1411–8.

    Article  PubMed  Google Scholar 

  9. Nashef SA, Roques F, Sharples LD, Nilsson J, Smith C, Goldstone AR, Lockowandt U. Euroscore ii. Eur J Cardiothorac Surg. 2012; 41: 734 – 44; discussion 44 – 5.

  10. Choi JW, Hwang HY, Kim KB. The performance of the current risk prediction scoring systems in patients undergoing anaortic off-pump coronary artery bypass grafting. J Korean Med Sci. 2021;36:e163.

    Article  PubMed  PubMed Central  Google Scholar 

  11. Song Y, Zhai W, Ma S, Wu Y, Ren M, Van den Eynde J, Nardi P, Pang PYK, Ali JM, Han J, Guo Z. Machine learning-based prediction of off-pump coronary artery bypass grafting-associated acute kidney injury. J Thorac Dis. 2024;16:4535–42.

    Article  PubMed  PubMed Central  Google Scholar 

  12. Parise O, Parise G, Vaidyanathan A, Occhipinti M, Gharaviri A, Tetta C, Bidar E, Maesen B, Maessen JG, La Meir M, Gelsomino S. Machine learning to identify patients at risk of developing new-onset atrial fibrillation after coronary artery bypass. J Cardiovasc Dev Dis. 2023; 10:82.

  13. Song W, Qiu L, Qing J, Zhi W, Zha Z, Hu X, Qin Z, Gong H, Li Y. Using bayesian network model with mmhc algorithm to detect risk factors for stroke. Math Biosci Eng. 2022;19:13660–74.

    Article  PubMed  Google Scholar 

  14. Lee S, Song IU, Na SH, Jeong DS, Chung SW. Association between long-term functional outcome and change in Hs-Crp level in patients with acute ischemic stroke. Neurologist. 2020;25:122–5.

    Article  PubMed  Google Scholar 

  15. Latchaw RE, Alberts MJ, Lev MH, Connors JJ, Harbaugh RE, Higashida RT, Hobson R, Kidwell CS, Koroshetz WJ, Mathews V, Villablanca P, Warach S, Walters B. Recommendations for imaging of acute ischemic stroke: a scientific statement from the American heart association. Stroke. 2009;40:3646–78.

    Article  PubMed  Google Scholar 

  16. Wintermark M, Sanelli PC, Albers GW, Bello JA, Derdeyn CP, Hetts SW, Johnson MH, Kidwell CS, Lev MH, Liebeskind DS, Rowley HA, Schaefer PW, Sunshine JL, Zaharchuk G, Meltzer CC. Imaging recommendations for acute stroke and transient ischemic attack patients: a joint statement by the American society of neuroradiology, the American college of radiology and the society of neurointerventional surgery. J Am Coll Radiol. 2013;10:828–32.

    Article  PubMed  PubMed Central  Google Scholar 

  17. Mao Z, Zhong X, Yin J, Zhao Z, Hu X, Hackett ML. Predictors associated with stroke after coronary artery bypass grafting: a systematic review. J Neurol Sci. 2015;357:1–7.

    Article  PubMed  Google Scholar 

  18. Tatlisuluoglu D, Tezcan B, Mungan İ, Çakirli YA, Tümer NB, Taşoğlu İ. Predicting postoperative ischemic stroke problems in patients following coronary bypass surgery using neutrophil-lymphocyte ratio, platelet-lymphocyte ratio, and red blood cell distribution width values. Kardiochir Torakochirurgia Pol. 2022;19:90–5.

    PubMed  PubMed Central  Google Scholar 

  19. Tibshirani R. Regression shrinkage and selection via the lasso. J Royal Stat Soc Ser B-Statistical Methodol. 1996;58:267–88.

    Article  Google Scholar 

  20. Arora P, Boyne D, Slater JJ, Gupta A, Brenner DR, Druzdzel MJ. Bayesian networks for risk prediction using real-world data: a tool for precision medicine. Value Health. 2019;22:439–45.

    Article  PubMed  Google Scholar 

  21. Song W, Qin Z, Hu X, Han H, Li A, Zhou X, Li Y, Li R. Using bayesian networks with tabu-search algorithm to explore risk factors for hyperhomocysteinemia. Sci Rep. 2023;13:1610.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  22. Hajian-Tilaki K. The choice of methods in determining the optimal cut-off value for quantitative diagnostic test evaluation. Stat Methods Med Res. 2018;27:2374–83.

    Article  PubMed  Google Scholar 

  23. Rubens FD, Fremes SE, Grubic N, Fergusson D, Taljaard M, van Walraven C. Outcomes following coronary artery bypass grafting with multiple arterial grafting by pump status in men and women. J Thorac Cardiovasc Surg. 2024; 167: 1796 – 807.e15.

  24. Qu WW, Wei JW, Binongo JN, Keeling WB. Sex differences in failure-to-rescue after coronary artery bypass grafting. Ann Thorac Surg. 2022;114:1596–602.

    Article  PubMed  Google Scholar 

  25. Bryce Robinson N, Naik A, Rahouma M, Morsi M, Wright D, Hameed I, Di Franco A, Girardi LN, Gaudino M. Sex differences in outcomes following coronary artery bypass grafting: a meta-analysis. Interact Cardiovasc Thorac Surg. 2021;33:841–7.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  26. Mehta A, Choxi R, Gleason T, Wechsler L, Jovin T, Thirumala PD. Carotid artery disease as a predictor of in-hospital postoperative stroke after coronary artery bypass grafting from 1999 to 2011. J Cardiothorac Vasc Anesth. 2018;32:1587–96.

    Article  PubMed  Google Scholar 

  27. Wang S, Ran Y, Cheng S, Lyu Y, Liu J. Determinants and clinical outcomes of stroke following revascularization among patients with reduced ejection fraction. Brain Behav. 2023;13:e2927.

    Article  PubMed  PubMed Central  Google Scholar 

  28. Chen S, Mi C, Zhang S, Li Y, Yun Y, Zhang X, Chen J, Li Y, Zhang H, Gao T, Zou C, Ma X. The role of carotid artery stenosis in predicting stroke after coronary artery bypass grafting in a Chinese cohort study. Sci Rep. 2023;13:21536.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  29. Masabni K, Raza S, Blackstone EH, Gornik HL, Sabik JF 3. Does preoperative carotid stenosis screening reduce perioperative stroke in patients undergoing coronary artery bypass grafting? J Thorac Cardiovasc Surg. 2015;149:1253–60.

  30. Klarin D, Patel VI, Zhang S, Xian Y, Kosinski A, Yerokun B, Badhwar V, Thourani VH, Sundt TM, Shahian D, Melnitchouk S. Concomitant carotid endarterectomy and cardiac surgery does not decrease postoperative stroke rates. J Vasc Surg. 2020;72:589–e963.

    Article  PubMed  Google Scholar 

  31. Yazdanpanah S, Rabiee M, Tahriri M, Abdolrahim M, Rajab A, Jazayeri HE, Tayebi L. Evaluation of glycated albumin (Ga) and ga/hba1c ratio for diagnosis of diabetes and glycemic control: a comprehensive review. Crit Rev Clin Lab Sci. 2017;54:219–32.

    Article  CAS  PubMed  Google Scholar 

  32. Kim Y, Lee SH, Kang MK, Kim TJ, Jeong HY, Lee EJ, Bae J, Jeon K, Nam KW, Yoon BW. Glycated albumin, a novel biomarker for short-term functional outcomes in acute ischemic stroke. Brain Sci. 2021;11:337.

  33. Selvin E, Rawlings AM, Lutsey PL, Maruthur N, Pankow JS, Steffes M, Coresh J. Fructosamine and glycated albumin and the risk of cardiovascular outcomes and death. Circulation. 2015;132:269–77.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  34. Yang ZK, Shen Y, Shen WF, Pu LJ, Meng H, Zhang RY, Zhang Q, Chen QJ, De Caterina R, Lu L. Elevated glycated albumin and reduced endogenous secretory receptor for advanced glycation endproducts levels in serum predict major adverse cardio-cerebral events in patients with type 2 diabetes and stable coronary artery disease. Int J Cardiol. 2015;197:241–7.

    Article  PubMed  Google Scholar 

  35. Li J, Wang Y, Wang D, Lin J, Wang A, Zhao X, Liu L, Wang C, Wang Y. Glycated albumin predicts the effect of dual and single antiplatelet therapy on recurrent stroke. Neurology. 2015;84:1330–6.

    Article  CAS  PubMed  Google Scholar 

  36. Zhao X, Li Q, Tu C, Zeng Y, Ye Y. High glycated albumin is an independent predictor of low response to clopidogrel in acs patients: a cross-sectional study. Cardiovasc Diabetol. 2020;19:171.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  37. Peng R, Liu K, Li W, Yuan Y, Niu R, Zhou L, Xiao Y, Gao H, Yang H, Zhang C, Zhang X, He M, Wu T. Blood urea nitrogen, blood urea nitrogen to creatinine ratio and incident stroke: the dongfeng-tongji cohort. Atherosclerosis. 2021;333:1–8.

    Article  CAS  PubMed  Google Scholar 

  38. Xie Y, Bowe B, Li T, Xian H, Yan Y, Al-Aly Z. Higher blood urea nitrogen is associated with increased risk of incident diabetes mellitus. Kidney Int. 2018;93:741–52.

    Article  CAS  PubMed  Google Scholar 

  39. Koppe L, Nyam E, Vivot K, Manning Fox JE, Dai XQ, Nguyen BN, Trudel D, Attané C, Moullé VS, MacDonald PE, Ghislain J, Poitout V. Urea impairs β cell glycolysis and insulin secretion in chronic kidney disease. J Clin Invest. 2016;126:3598–612.

    Article  PubMed  PubMed Central  Google Scholar 

  40. Feng GH, Li HP, Li QL, Fu Y, Huang RB. Red blood cell distribution width and ischaemic stroke. Stroke Vasc Neurol. 2017;2:172–5.

    Article  PubMed  PubMed Central  Google Scholar 

  41. Xie KH, Liu LL, Liang YR, Su CY, Li H, Liu RN, Chen QQ, He JS, Ruan YK, He WK. Red cell distribution width: a novel predictive biomarker for stroke risk after transient ischaemic attack. Ann Med. 2022;54:1167–77.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  42. Xue J, Zhang D, Zhang XG, Zhu XQ, Xu XS, Yue YH. Red cell distribution width is associated with stroke severity and unfavorable functional outcomes in ischemic stroke. Front Neurol. 2022;13:938515.

    Article  PubMed  PubMed Central  Google Scholar 

  43. Hamatani Y, Nagai T, Nakai M, Nishimura K, Honda Y, Nakano H, Honda S, Iwakami N, Sugano Y, Asaumi Y, Aiba T, Noguchi T, Kusano K, Toyoda K, Yasuda S, Yokoyama H, Ogawa H, Anzai T. Elevated plasma d-dimer level is associated with short-term risk of ischemic stroke in patients with acute heart failure. Stroke. 2018;49:1737–40.

    Article  CAS  PubMed  Google Scholar 

  44. Nahab F, Sharashidze V, Liu M, Rathakrishnan P, El Jamal S, Duncan A, Hoskins M, Marmarchi F, Belagaje S, Bianchi N, Belair T, Henriquez L, Monah K, Rangaraju S. Markers of coagulation and hemostatic activation aid in identifying causes of cryptogenic stroke. Neurology. 2020;94:e1892–9.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  45. Ohara T, Farhoudi M, Bang OY, Koga M, Demchuk AM. The emerging value of serum d-dimer measurement in the work-up and management of ischemic stroke. Int J Stroke. 2020;15:122–31.

    Article  PubMed  Google Scholar 

Download references

Acknowledgements

We express our sincere gratitude to all the participants, research assistants, and outcome assessors involved in this study. Our thanks also go to Editage (www.editage.com) for English language editing.

Funding

The present work was supported by China National Natural Science Foundation (82071342).

Author information

Author notes

  1. Wenlong Zou, Haipeng Zhao and Ming Ren contributed equally to this work.

Authors and Affiliations

  1. Department of Neurology, Beijing Anzhen Hospital, Capital Medical University, Beijing, China

    Wenlong Zou, Guobin Yuan, Boyi Yuan, Zeyu Ji, Chao Wu, Bin Cai, Tingting Yang & Guangzhi Liu

  2. Department of Cardiovascular surgery, Beijing Chaoyang Hospital, Capital Medical University, Beijing, China

    Haipeng Zhao

  3. Department of Neurology, Shanghai Blue Cross Brain Hospital, Shanghai, China

    Ming Ren

  4. Department of Ophthalmology, Qingdao Central Hospital, University of Health and Rehabilitation Sciences, Qingdao, China

    Chaoxiong Cui

  5. Department of Ultrasonography, Qingdao Endocrine & Diabetes Hospital, Qingdao, China

    Jinjun Zou

Authors
  1. Wenlong Zou
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Haipeng Zhao
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Ming Ren
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Chaoxiong Cui
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Guobin Yuan
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Boyi Yuan
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. Zeyu Ji
    View author publications

    You can also search for this author in PubMed Google Scholar

  8. Chao Wu
    View author publications

    You can also search for this author in PubMed Google Scholar

  9. Bin Cai
    View author publications

    You can also search for this author in PubMed Google Scholar

  10. Tingting Yang
    View author publications

    You can also search for this author in PubMed Google Scholar

  11. Jinjun Zou
    View author publications

    You can also search for this author in PubMed Google Scholar

  12. Guangzhi Liu
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

GZL and JJZ conceived the experiment(s), WLZ, HPZ, and MR conducted the experiment(s), CXC, GBY, BYY, ZYJ, CW, BC, and TTY analyzed the results. All authors contributed to the article and approved the submitted version.

Corresponding authors

Correspondence to Jinjun Zou or Guangzhi Liu.

Ethics declarations

Ethics approval and consent to participate

Our research was conducted in accordance with the Declaration of Helsinki, and was approved by the ethical review board of Beijing Anzhen Hospital. Each participating patient in this study recruited written informed consent.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Electronic supplementary material

Below is the link to the electronic supplementary material.

12911_2024_2762_MOESM1_ESM.docx

Supplementary Material 1: Table S1. The conditional probability table of the training set based on acute ischemic stroke (AIS) as the target node

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Zou, W., Zhao, H., Ren, M. et al. Risk factors and prediction model for acute ischemic stroke after off-pump coronary artery bypass grafting based on Bayesian network. BMC Med Inform Decis Mak 24, 349 (2024). https://doi.org/10.1186/s12911-024-02762-2

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s12911-024-02762-2

Keywords

BMC Medical Informatics and Decision Making

ISSN: 1472-6947

Contact us