Interpretable Prognostic Modeling for Long-Term Survival of Type A Aortic Dissection Patients Using Support Vector Machine Algorithm

Interpretable Prognostic Modeling for Long-Term Survival of Type A Aortic Dissection Patients Using Support Vector Machine Algorithm | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Interpretable Prognostic Modeling for Long-Term Survival of Type A Aortic Dissection Patients Using Support Vector Machine Algorithm Hao Cai, Yue Shao, Xuan-yu Liu, Chang-ying Li, Hao-yu Ran, Hao-ming Shi, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5786813/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 15 Apr, 2025 Read the published version in European Journal of Medical Research → Version 1 posted 11 You are reading this latest preprint version Abstract Objective This study aims to develop a reliable and interpretable predictive model for the risk of long-term survival in type A aortic dissection (TAAD) patients, utilizing machine learning (ML) algorithms. Methods We retrospectively reviewed the clinical data diagnosed with Type A Aortic Dissection (TAAD) who underwent open surgical repair at our institution between September 2017 and December 2020. Cases with less than 20% missing data were imputed using random forest algorithms. To identify significant prognostic factors, we performed LASSO (Least Absolute Shrinkage and Selection Operator) Cox regression analysis, including preoperative blood markers, previous medical history and perioperative condition. Based on the advantages of the model and the characteristics of the dataset, we subsequently developed a machine learning-based prognostic model using Support Vector Machine (SVM) and evaluated its performance across key metrics. To further explain the decision-making process of the SVM model, we employed SHapley Additive exPlanation (SHAP) values for model interpretation. Results A total of 175 patients with TAAD were included in the study. Through LASSO Cox regression, univariate analysis, and clinical relevance assessment, eight feature variables were selected for modeling. Performance evaluation revealed that the SVM model showed excellent performance in both the training and test sets, with no significant overfitting, indicating strong clinical applicability. In the training set, the model achieved an AUC of 0.9247 (95% CI: 0.9200–0.9279), and in the testing set, 0.8800 (95% CI: 0.8492–0.9396). The accuracy was 0.8663 and 0.8857, precision was 0.8627 and 1.0000, recall was 0.8713 and 0.7333, F1 score was 0.8670 and 0.8462, Brier score was 0.1068 and 0.1070, average precision (AP) was 0.9266 and 0.9086, and C-index was 0.8901 and 0.8700, respectively. SHAP analysis identified that longer ICU hospital stay, abdominal pain, plasma transfusion volume, creatinine, white blood cell count, operation time, and systemic immune-inflammation index (SII) had significant positive impact on the model's predictions. Conclusion This study developed an interpretable predictive model based on the SVM algorithm to assess long-term survival in TAAD patients. The model demonstrated accuracy, precision, and robustness in identifying high-risk patients, providing clinicians with reliable evidence for prognosis management. type A aortic dissection machine learning long-term survival predictive model Support Vector Machine (SVM) Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction Aortic dissection is a life-threatening condition with an increasing incidence worldwide [1]. According to the Stanford classification system, established in the last century, Type A aortic dissection (TAAD) involves the ascending aorta and may extend to the aortic arch and descending aorta. Unlike Type B aortic dissection, TAAD has well-defined emergency indications for surgical repair [3]. Early studies indicate that the mortality rate for TAAD increases by 1–2% per hour during the first 48 hours [2]. Despite advancements in surgical techniques and life support systems, the prognosis for TAAD patients remains poor due to the time-critical nature of the condition and a high risk of complications in the early stages. Recent data reveal that in-hospital and 30-day mortality rates for TAAD surgery exceed 20%, underscoring the urgent need for improved prognostic strategies [4,5]. Identifying high-risk patients is therefore crucial for optimizing prognosis management and guiding clinical decisions. Perioperative conditions, such as ICU stay duration, and specific biomarkers, including C-reactive protein (CRP) levels, have been identified as significant prognostic factors. For instance, Liu et al. found that TAAD patients with ICU stays longer than 7 days face increased risks of adverse outcomes, including severe organ damage and respiratory complications [6]. Similarly, Tang et al. demonstrated that elevated preoperative CRP levels are independent predictors of in-hospital mortality, renal dysfunction, and stroke in TAAD patients [7]. Moreover, the Systemic Immune-Inflammation Index (SII), which combines neutrophil, lymphocyte, and platelet counts, has shown promise in predicting postoperative complications and three-year survival [8,9]. Despite their statistical significance, these indicators are not widely implemented in clinical practice, primarily due to their limited ability to account for the complexity of TAAD prognosis in real-world settings. The limited clinical application of these indicators may stem from their narrow focus on either short-term outcomes or long-term survival, which fails to address the broader, multi-organ impact of TAAD. Moreover, an exclusive focus on significantly abnormal markers may overlook the interrelationships between variables, leading to missing key information and underestimating the prognostic value of certain factors. Most of the prognostic indicators identified in previous studies have predominantly been used for preoperative risk stratification to guide surgical timing. Unlike other acute or malignant conditions, early surgical intervention is critical to mitigate risks of aortic rupture and mortality[42]. Thus, relying solely on preoperative assessments is insufficient for comprehensive prognostic management. Addressing this gap necessitates the development of models that integrate multiple perioperative, biomarker, and clinical factors to improve risk stratification accuracy and guide both pre- and postoperative decisions. Traditional regression methods, such as logistic regression and Cox regression, provide valuable insights but struggle to manage high-dimensional data and complex interactions, potentially overlooking critical prognostic factors [43,44]. In contrast, machine learning (ML) techniques excel in identifying complex patterns within datasets, offering advantages in model optimization and predictive accuracy [10]. ML has demonstrated significant potential in healthcare, particularly in predicting health outcomes and disease progression in cardiovascular conditions [10–13]. However, the application of ML to TAAD prognosis research, especially for long-term survival prediction, remains underexplored. This study aims to develop and validate a practical ML model utilizing SHapley Additive exPlanations (SHAP)-based visualization to identify key prognostic factors and stratify TAAD patients at high risk of poor outcomes. By enhancing the interpretability and accuracy of prognostic assessments, this model seeks to support clinicians in devising effective treatment strategies and improving patient outcomes. Material and Methods Patients In this retrospective cohort study, we analyzed the clinical data of patients with Stanford Type A Aortic Dissection (TAAD) who consecutively underwent surgical repair at the Department of Thoracic and Cardiovascular Surgery, First Affiliated Hospital of Chongqing Medical University, from September 2017 to December 2020. The inclusion criteria for this study were as follows: (1) patients diagnosed with Stanford TAAD following aortic computed tomography angiography (CTA); (2) patients who underwent surgical repair; (3) patients aged 18 years or older. The exclusion criteria included: (1) patients with more than 20% absence of clinical or laboratory data; (2) patients with preoperative comorbidities such as malignant tumors, hematological disorders, infections, systemic inflammatory diseases, or undergoing treatments that could influence biomarker levels and survival; (3) patients who died directly or indirectly from causes other than TAAD. Ultimately, 175 patients were included in the study ( Fig. 1A ). The study was approved by the Independent Ethics Committee of the First Affiliated Hospital of Chongqing Medical University (approval number: 2024-583-01), and the study was conducted in compliance with the ethical standards of the World Medical Association Declaration of Helsinki. Data Collection Demographic characteristics, clinical symptoms, and hemodynamic profiles at admission—including gender, age, smoking history, hypertension (HTN), diabetes mellitus (DM), cardiovascular disease (CVD), blood pressure, and heart rate (HR)—and intraoperative details such as blood loss, operation time, cardiopulmonary bypass time (CPB), aortic cross-clamp time (ACC), red blood cell transfusion volume, and plasma transfusion volume were retrospectively reviewed and extracted from electronic medical records. Blood indicators were assessed within 24 hours of hospital admission and prior to surgery. The biomarkers measured included red blood cell (RBC) count, creatinine, absolute neutrophil count (ANC), white blood cell (WBC) count, hemoglobin (Hb), platelet count, absolute lymphocyte count (ALC), monocyte count, serum albumin, uric acid, urea nitrogen, alanine aminotransferase (ALT), aspartate aminotransferase (AST), cardiac troponin T (cTnT), myoglobin, fibrinogen, and D-dimer, etc. A composite metric, the Systemic Immune-Inflammation Index (SII), was calculated using the following formula: platelet count × neutrophil count / lymphocyte count.​ Follow-up and Treatments All patients included in this study underwent surgical repair. The primary outcome was patient mortality. Overall survival (OS) was defined as the time from surgery to either death or the last follow-up. Follow-up started three months after discharge and concluded on August 20, 2024. Telephone follow-ups, conducted by trained interviewers, were performed at 3, 6, and 12 months after discharge, and subsequently every six months to support post-discharge care. The maximum OS observed was 2,445 days, with a median OS of 1,190 days. Data Preprocessing and Feature Variable Selection For features with a missing data rate below 20%, random forest imputation was employed [14]. The dataset was then randomly divided into an 80% training set and a 20% test set. To reduce the risk of overfitting, feature selection was conducted using LASSO Cox regression analysis. Features were excluded based on regression results and clinical relevance, yielding an initial model with 16 features, including 2 categorical and 14 continuous variables. Subsequently, univariate analysis and correlation analysis were performed on the preliminarily selected variables. Features were finalized for inclusion in the machine learning model based on statistical significance, clinical relevance, correlation coefficients and LASSO Cox regression coefficients (Fig. 1B). To address data imbalance, data normalization was performed first, followed by the application of the Synthetic Minority Over-sampling Technique (SMOTE) to the training set [15,16]. Model Selection and Performance Evaluation The Support Vector Machine (SVM) stands out among machine learning models for its suitability in small-sample datasets, offering significant theoretical and algorithmic advantages. Based on statistical learning theory, SVM utilizes the Structural Risk Minimization (SRM) principle, which prioritizes minimizing generalization error over merely optimizing training error. This approach significantly enhances the model's ability to generalize, especially when working with small sample sizes. Unlike traditional models, such as neural networks, which depend on large datasets to approximate the target function, SVM can effectively predict unseen data by maximizing the margin between classes, even with limited sample sizes [20]. SVM is particularly effective in handling small sample, high-dimensional datasets, offering a robust solution to the "curse of dimensionality." By employing kernel functions, SVM maps the original feature space into a higher-dimensional space, enabling the identification of an optimal, linearly separable hyperplane without the need to explicitly compute the high-dimensional mappings [21]. Additionally, when combined with LASSO Cox regression analysis, SVM incorporates regularization techniques to control model complexity and reduce the risk of overfitting. These features make SVM a widely adopted approach in various fields, including disease classification (e.g., early cancer diagnosis), survival analysis, and treatment recommendation [22–24]. Given the distinct advantages of Support Vector Machines (SVM), their widespread application in medicine and bioinformatics, and the characteristics of the dataset in this study, the SVM model was chosen for subsequent analysis. A 10-fold cross-validation strategy was employed to ensure a comprehensive evaluation of model performance. The training data was divided into 10 subsets, with each iteration using 9 subsets for model training and the remaining subset for validation. This process was repeated 10 times, allowing each subset to serve as the validation set once. Key performance metrics used to evaluate the model's predictive and generalization ability included the area under the receiver operating characteristic curve (AUC), accuracy, precision, recall, F1 score, Brier score, and the area under the precision-recall curve (AP) [18]. To enhance model interpretability, this study incorporated the SHAP (SHapley Additive exPlanations) method. SHAP quantifies the marginal contribution of each feature to the model’s predicted outcomes, offering both global and local insights into the decision-making process. By assigning precise attribution values to each variable, SHAP effectively identifies the most influential features for model predictions [19]. Statistical Analysis This study included the demographic characteristics, risk factors, admission status, preoperative blood markers, and intraoperative conditions of TAAD patients. Data analysis was performed using SPSS 27.0 and Python 3.1. Continuous variables are presented as mean ± standard deviation (SD), with comparisons made using Student's t-test or Mann-Whitney U test. Categorical variables are presented as frequencies and percentages (%) and analyzed using chi-square or Fisher's exact test. Pearson’s chi-squared test was used to evaluate correlations among continuous variables, visualized through a heatmap. The point-biserial correlation coefficient assessed correlation between categorical and continuous variables, with a bubble plot highlighting their strength and significance. The model was constructed using the SVM algorithm, and performance was evaluated using AUC, accuracy, precision, recall, F1 score, Brier score, and AP. Finally, the SHAP method was used for model interpretation. A two-tailed p-value of < 0.05 was considered statistically significant. Results Patient characteristics A total of 175 TAAD patients were included in the study ( Table S1 ), with 54 deaths observed during the follow-up period. The average age of the cohort was 48.87 ± 9.57 years, and the majority were male (139 cases, 79.43%). While only a small proportion of patients (6.29%, 11 cases) experienced concomitant shock, this group had a markedly higher mortality rate compared to those without shock (63.64% vs. 28.66%). At admission, abdominal pain was reported in 9.71% of patients (17 cases) and was more prevalent in the outcome group, with 18.52% (10 cases) reporting this symptom. This observation suggests a potential association between abdominal pain and the severity of TAAD. Additionally, patients with endpoint events had significantly shorter postoperative hospital stays compared to survivors (10.77 ± 10.29 days vs. 23.13 ± 14.42 days). This likely reflects the fact that critically ill patients are more prone to early mortality and fail to benefit from postoperative care. 175 TAAD patients were randomly assigned to the training (140 patients, 80%) and test (35 patients, 20%) groups. No significant differences were found between the two groups in terms of demographic data, risk factors, admission conditions, hospital stays, preoperative laboratory results, and intraoperative conditions ( Table S2 ). Table 1. Comparison of Perioperative Conditions: Variables Selected through LASSO Cox Regression for Initial Screening in the Training Set Variables No. of outcomes (%) No. of survivors (%) P value Shock 0.007 No 33 (84.6%) 97 (96.0%) Yes 6 (15.4%) 4 (4.0%) Abdominal pain 0.043 No 31 (79.5%) 94 (93.1%) Yes 8 (20.5%) 7 (6.9%) POHS (days) 12.41 ± 10.53 22.96 ± 14.9 <0.001 ICUHS (days) 11.51 ± 10.42 12.66 ± 8.63 0.542 WBC (×10ˆ9/L) 13.99 ± 5.15 11.55 ± 4.08 0.01 AST (IU/L) 109.44 ± 322.92 39.35 ± 68.05 0.187 Creatinine (μmol/L) 113.41 ± 52.5 85.63 ± 34.32 0.004 Fibrinogen (g/L) 3.14 ± 2.06 3.05 ± 1.66 0.801 BNP (pg/mL) 971.81 ± 1186.24 953.77 ± 1535.51 0.941 SII 3746.52 ± 6199.33 2340.46 ± 2270.78 0.075 Operation time (min) 624.69 ± 175.27 523.07 ± 112.79 0.001 Intraoperative bleeding (ml) 930.98 ± 1148.21 582.0 ± 467.63 0.073 Transfusion of SRBC (U) 4.13 ± 3.46 3.32 ± 2.95 0.202 Transfusion of Plasma (ml) 944.65 ± 419.55 750.3 ± 371.83 0.014 CPB time (min) 305.9 ± 86.56 264.15 ± 63.4 0.008 Abbreviations: POHS Postoperative Hospital Stay, ICUHS Intensive Care Unit Hospital Stay, WBC White Blood Cell count, AST Aspartate Aminotransferase, BNP B-type Natriuretic Peptide, SII Systemic Immune-Inflammation Index, SRBC: Suspended Red Blood Cells, CPB Cardiopulmonary Bypass. Feature Variable Selection LASSO (Least Absolute Shrinkage and Selection Operator) Cox regression was used to preliminarily screen variables associated with long-term survival, yielding 16 perioperative characteristic variables. The optimal lambda value, determined through cross-validation, was 25.9502 ( Fig. 2A, B ). Based on these findings, univariate analysis was performed on perioperative characteristics between survivors and patients who experienced endpoint events in the training set ( Table 1 ). To avoid information leakage from the test set, this analysis was limited to the training set. Variables with p-values < 0.05 were considered statistically significant. Among them, CBP and shock had small coefficients in the LASSO Cox regression, suggesting its limited contribution to the model, and was thus excluded. In the training group, the difference in preoperative SII values between survivor and outcome group did not reach statistical significance (p = 0.075, Table 1 ), likely due to large fluctuations in the SII data. However, given its relatively high regression coefficient in the LASSO Cox regression model, as well as its role as an important marker of inflammation and immune function, SII may still have significant prognostic value for TAAD patients. Therefore, we decided to include SII in further analysis. Correlation Among Selected Variables Included in the Model After further screening, eight variables were selected for model training based on the results of LASSO Cox regression analysis, univariate analysis, and clinical relevance. These variables included seven continuous variables and one categorical variable: ICU hospital stay, postoperative hospital stay, plasma transfusion, creatinine, operation time, SII, WBC, and abdominal pain. The heatmap presented Pearson’s correlation coefficients among continuous variables, providing a visual depiction of the strength and direction of these correlations. The color gradient, ranging from dark blue to dark red, represented the spectrum from weak to strong correlations. Each cell in the heatmap was labeled with a Pearson’s correlation coefficient, quantifying the linear relationship between two variables. A coefficient of 1 indicated a perfect positive correlation, -1 signified a perfect negative correlation, and 0 reflects no correlation. Similarly, the bubble plot illustrated the Point-Biserial correlation coefficients between the categorical variable (abdominal pain) and the continuous variables, with the size and position of the bubbles representing the strength and direction of the correlations, respectively. Asterisks were used to denote statistical significance, with a single asterisk (*) representing a p-value of less than 0.05. Pearson's chi-squared test demonstrated significant correlations between WBC and creatinine, operation time, and SII (p<0.001, p<0.05, p<0.001, respectively), with Pearson correlation coefficients of 0.29, 0.19, and 0.44, indicating weak to moderate relationships ( Fig.3 ). Conversely, no significant correlations were identified between WBC and ICU hospital stay, postoperative hospital stay, or plasma transfusion (p>0.05). Notably, a strong correlation was observed between ICU hospital stay and postoperative hospital stay (Pearson correlation coefficient: 0.74, p<0.001), suggesting that prolonged ICU stays are associated with extended postoperative hospitalizations. This strong relationship likely reflects the complexity of postoperative recovery and the severity of the patient's condition. Hospital stay, as a metric, may also serve as an indirect indicator of the effectiveness of comprehensive management during hospitalization and the patient’s overall risk profile. Consequently, ICU stay and postoperative hospital stay were included for further analysis. For the remaining variables, the correlation strengths were generally weak or lacked statistical significance ( Fig. 3 and Fig.S1 ). The Performance of the SVM Model After selecting the feature variables, the SMOTE method was applied to the training set to address data imbalance. Subsequently, the SVM algorithm was implemented, and model generalization was enhanced using 10-fold cross-validation, with final validation performed on the test set. The model demonstrated excellent performance, achieving AUC values exceeding 0.85 for both the training and test sets, indicating strong discriminatory ability ( Table 2, Fig. 4A ). Specifically, the training set achieved an AUC of 0.9247 (95% CI: 0.9200-0.9279), while the test set achieved an AUC of 0.8800 (95% CI: 0.8492-0.9396). The slightly lower AUC for the test set suggests some performance decline, but the difference remains within a reasonable range, with no significant signs of overfitting. The model demonstrated consistent accuracy across both the training and test sets, with the test set achieving slightly higher accuracy (training set: 0.8663, test set: 0.8857). Notably, the test set achieved 100% precision, underscoring the model’s reliability in predicting TAAD mortality ( Table 2, Fig. 4C and D ). Precision-Recall curves showed similar shapes and AP values above 0.9 for both sets, further confirming these results ( Fig. 4B ). The Brier Scores for the training and test sets were notably low, at 0.1068 and 0.1070, respectively, indicating well-calibrated probabilistic predictions. The C-index values were 0.8901 and 0.8700 for the training and test sets, respectively, both close to 0.9, indicating robust predictive performance and strong generalizability. These results align with prior analyses, further supporting the reliability of the model. In the Decision Curve Analysis (DCA), the model effectively distinguished risk levels and provided high net benefits across a wide range of thresholds. Although the test set curve was slightly lower than the training set, the overall trends were consistent, reflecting strong generalization ability ( Fig. 4E ). Performance metrics, including accuracy, precision, AUC, AP, Brier Score, F1 Score, and C-index, showed comparable results between the training and test sets, demonstrating the model’s robustness, reliability, and practical value in clinical applications ( Fig. S2 ). Table 2. Performance of the SVM Model between Training Set and Test Set. Metrics SVM (training) SVM (test) AUC 0.9247 0.8800 95%CI (0.9200-0.9279) ( 0.8492-0.9396) Accuracy 0.8663 0.8857 Precision 0.8627 1.0000 Recall 0.8713 0.7333 F1 score 0.8670 0.8462 Brier score 0.1068 0.1070 AP 0.9266 0.9086 C-index 0.8901 0.8700 Abbreviations: AUC Area Under the Curve, SVM Support Vector Machine, CI Confidence Interval, AP Average Precision, C-index Concordance Index. SHAP Interpretation and Feature Importance Visualization The SHAP method was employed to interpret the SVM model's predictions and evaluate its clinical relevance. By quantifying the contribution of each feature, SHAP values provided insights into their impact on the model's outputs. Feature importance analysis identified POHS as the most critical predictor of long-term survival in TAAD patients, highlighting the potential benefits of comprehensive postoperative management for survivors ( Fig. 5A ). Furthermore, the SHAP summary plot revealed that higher ICUHS, Plasma transfusion volume, creatinine, WBC, operation time, SII and abdominal pain positively influenced the model’s predictions, indicating their significant roles in survival outcomes ( Fig. 5B ). To validate the model's interpretability, decision curves were used to illustrate individualized predictions of long-term survival. The gray vertical line at 0 on the horizontal axis represented the model's baseline. Fig. 6A visualizes the decision-making process for TAAD survivors, while Fig. 6B illustrates it for patients with endpoint events. Discussion TAAD is a life-threatening vascular emergency for which surgical repair remains the primary treatment. However, its management poses significant challenges, particularly in early diagnosis, perioperative complications, and mortality risk. The nonspecific TAAD symptoms may delay timely diagnosis, while timely imaging is constrained by both time and resource availability, allowing the condition to deteriorate rapidly. The severity of the disease, coupled with the risks associated with emergency surgery, contributes to substantial perioperative mortality and the potential for serious complications. As a result, prognostic management for TAAD patients has become a critical focus of ongoing research. Despite advancements in prognostic assessment for TAAD, several challenges continue to limit the practical applicability of existing findings. For instance, Zhang et al. identified systolic blood pressure at admission, NT-proBNP, and white blood cell count as independent factors affecting in-hospital mortality among TAAD patients [25]. Similarly, numerous studies have highlighted preoperative indicators such as fibrinogen, BUN, NLR, PLR, D-dimer, UA, and CRP as prognostic markers for both short-term and long-term survival in TAAD patients [26-31]. However, relying solely on a limited set of preoperative indicators often fails to capture the full complexity of changes in a patient’s condition throughout the perioperative period, potentially overlooking critical factors. This limitation reduces the practical utility of these markers, even when statistically significant differences are observed. In addition, traditional Cox and logistic regression methods, while widely used, have inherent limitations. These methods struggle to manage complex, high-dimensional data and fail to fully capture intricate relationships between variables. Furthermore, their reliance on model assumptions can hinder their ability to generalize effectively to new datasets, thereby reducing their predictive performance and clinical applicability [32,33]. In recent years, ML methods have gained increasing attention for prognostic assessment in TAAD. However, research in this area remains in its early stages. For example, Zhang et al. developed a Treebag model to predict one-year mortality in TAAD patients, using 51 clinical characteristics, including blood markers at admission [37]. Key survival predictors identified included surgical intervention, beta-blocker use during hospitalization, systolic blood pressure at admission, lymphocyte count, carbon dioxide combining capacity, eosinophil count, and white blood cell count. Nevertheless, relying exclusively on preoperative indicators and a single type of postoperative medication may fail to account for the dynamic fluctuations in TAAD patients’ conditions over the perioperative period. This approach may also underestimate the potential benefits of a comprehensive perioperative management strategy. Moreover, Most recent ML studies have primarily focused on short-term mortality and postoperative complications, with limited attention given to long-term survival prediction in TAAD patients [34-36]. Of note, studies relying solely on public database may exclude certain patient subgroups due to the database's inclusion criteria. As a result, while statistically significant, these datasets may lack broad representativeness, limiting the model's generalizability. Addressing these gap is crucial for enhancing the prognostic utility of ML approaches and improving their practical application in real-world clinical settings. In this study, we used an interpretable machine learning approach to explore the relationship between preoperative vital signs, blood markers, clinical history, demographic factors, and long-term survival in TAAD patients. We specifically examined hospital stays to evaluate patients' overall clinical condition, hospitalization benefits, and perioperative risks. ICU stay reflected condition stability and critical care complexity, while postoperative hospital stay captured recovery progress and the effectiveness of comprehensive management. Together, these variables provided a multidimensional assessment of the balance between perioperative risks and long-term survival benefits, offering valuable insights into hospitalization outcomes in TAAD. Based on the data characteristics and the advantages of the SVM algorithm, we selected SVM for analysis. The SHAP method was applied to interpret feature importance and its relationship with long-term mortality. Our results showed strong performance of the SVM model in both training and test datasets, with postoperative hospital stay, ICU stay, and abdominal pain at admission identified as the most significant predictors. SHAP decision plots were used to further illustrate the model's decision-making process. Feature importance ranking and scatter plots identified postoperative hospital stay as the most critical factor, with a protective effect on long-term mortality in TAAD patients. This finding suggests that severely ill patients who are unable to tolerate prolonged hospital stays are more likely to experience early mortality, consistent with the observation that outcome group had shorter hospital stays. In contrast, patients with more stable conditions appeared to benefit from comprehensive postoperative management, resulting in improved long-term survival. ICU hospital stay was identified as the second most significant feature. Scatter plots indicated that longer ICU stay were associated with higher long-term mortality, which aligns with previous studies [38]. Prolonged ICU stay did not improve outcomes but instead increased the risk of complications and adverse events. Additionally, while severe chest pain is commonly reported at admission, patients presenting with abdominal pain had a higher incidence of endpoint events compared to survivors. Scatter plots further highlighted abdominal pain at admission as a key predictor of long-term mortality, potentially signaling disease progression. Other important features in the model included operation time, plasma transfusion volume, WBC, creatinine, and SII. Severe TAAD cases often require longer surgeries due to increased procedural complexity and coagulopathy, leading to higher plasma transfusion volumes—observations consistent with the outcome group. Similar findings have been reported in prior studies, indicating that prolonged surgery and higher transfusion requirements are associated with poor clinical outcomes [39-42]. Furthermore, SII, a composite marker of inflammation and immune function, along with WBC and creatinine, reflects preoperative immune status, renal function, and inflammatory levels, all of which are linked to adverse prognoses in TAAD patients. In this study, we developed an SVM-based machine learning model that exhibited strong predictive performance on both training and test datasets. This model serves as a valuable clinical tool for assessing the long-term prognosis of TAAD patients, enabling more informed decision-making regarding critical aspects of care, such as surgical timing, ICU transfer, transition to general wards, and discharge planning. By providing a comprehensive evaluation of perioperative factors, the model supports physicians in optimizing short-term postoperative management while maximizing long-term survival outcomes, ultimately enhancing the quality of patient care. Despite these contributions, this study has several limitations. First, although the SVM model showed good generalization with a small dataset, the limited sample size remains a concern. As a single-center retrospective study, the potential for selection bias may exist, underscoring the need for multi-center, large-scale studies to validate these findings. Second, while this study captured the impact of comprehensive perioperative management through hospital stay durations, it did not fully account for individual patient variability. Future research should aim to integrate fluctuations in perioperative blood markers and other individualized factors to refine and enhance the prognostic model. Conclusion This study is the first to develop and validate a machine learning-based prognostic model for long-term survival in TAAD patients. The model demonstrated strong predictive performance in both training and testing groups, supporting its clinical potential. It offers clinicians a tool for dynamically assessing long-term outcomes in TAAD patients. Abbreviations - TAAD: Type A Aortic Dissection - ML: Machine Learning - SVM: Support Vector Machine - SHAP: SHapley Additive exPlanations - LASSO: Least Absolute Shrinkage and Selection Operator - Cox: Cox regression model - CTA: Computed Tomography Angiography - CPB: Cardiopulmonary Bypass - ACC: Aortic Cross-Clamp - RBC: Red Blood Cell - WBC: White Blood Cell - Hb: Hemoglobin - Plt: Platelet - ANC: Absolute Neutrophil Count - ALC: Absolute Lymphocyte Count - ALB: Albumin - ALT: Alanine Aminotransferase - AST: Aspartate Aminotransferase - BUN: Blood Urea Nitrogen - LDH: Lactate Dehydrogenase - FDP: Fibrin/Fibrinogen Degradation Products - CK-MB: Creatine Kinase-Muscle/Brain - BNP: B-Type Natriuretic Peptide - cTnT: Cardiac Troponin T - SII: Systemic Immune-Inflammation Index - SRBC: Suspended Red Blood Cells - POHS: Postoperative Hospital Stay - ICUHS: Intensive Care Unit Hospital Stay - AUC: Area Under the Curve - AP: Average Precision - CI: Confidence Interval - PR: Precision-Recall - DCA: Decision Curve Analysis - NT-proBNP: N-terminal pro-B-type Natriuretic Peptide - NLR: Neutrophil to Lymphocyte Ratio - PLR: Platelet to Lymphocyte Ratio - ECG: Electrocardiogram Declarations Authors’ contribution HC was responsible for analyzing data, drafting manuscripts. YS was responsible for making critical revisions and data collection.XYL, CYL and HYR were responsible for patient follow-up and related clinicopathological data collection. HMS and CZ were responsible for technical support for data analysis, article grammar proofreading. QCW were responsible for the conception, design, and review of selected topics. This manuscript was read and approved by all credited authors. All authors read and approved the final manuscript. Funding This work was supported by grants from National Natural Science Foundation of China (82270506), and Project of innovation team for Graduate Teaching (CYYY-YJSJXCX-202318). Acknowledgement None. Availability of data and materials The datasets used in this study are available on request from the corresponding author. Ethics approval and consent to participate. This study was approved by the ethics committee of Chongqing Medical University (2024-583-01) and followed the ethical standards of the Helsinki Declaration. Informed consent was obtained from all the participants. Consent to participate. All patients provided written informed consent. Consent for publication Not applicable. Competing interests The authors have declared that no competing interest exists. References Moriyama S, Hara M, Hirota T, Nakata K, Doi H, Matsumura T, et al. Population-Based Study of the Incidence and Mortality Rate of Acute Aortic Dissection. Circulation Journal. 2023;88:297–306. HIRST AE, JOHNS VJ, KIME SW. DISSECTING ANEURYSM OF THE AORTA. Medicine. 1958;37:217. Zhu Y, Lingala B, Baiocchi M, Tao JJ, Toro Arana V, Khoo JW, et al. Type A Aortic Dissection—Experience Over 5 Decades: JACC Historical Breakthroughs in Perspective. Journal of the American College of Cardiology [Internet]. 2020 [cited 2022 Oct 21];76:1703–13. Available from: https://www.sciencedirect.com/science/article/pii/S0735109720362409 Jiang F, Wang X, Carmichael M, Chen Y, Huang R, Xiao Y, et al. Contemporary comparative surgical outcomes of type A aortic dissection in US and China: an analysis of the national inpatient sample database and a Chinese multi-institutional registry. Journal of Cardiothoracic Surgery. 2024;19. Nazir S, Ariss RW, Minhas AMK, Issa R, Michos ED, Birnbaum Y, et al. Demographic and Regional Trends of Mortality in Patients With Aortic Dissection in the United States, 1999 to 2019. Journal of the American Heart Association. 2022;11. Liu H, Zhang S, Zhang C, Gao Q, Liu Y, Liao F, et al. Risk factors for prolonged postoperative ICU stay in the patients with Stanford type A aortic dissection. Journal of Cardiothoracic Surgery [Internet]. 2024 [cited 2024 Apr 6];19. Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/pmid/38310273/ Tang Z, Liu H, Shao Y. Efficacy of CRP in combination with D-dimer in predicting adverse postoperative outcomes of patients with acute Stanford type A aortic dissection. Journal of Cardiothoracic Surgery. 2022;17. Li Z, Zhang H, Sulaiman Baraghtha, Mu J, Yusanjan Matniyaz, Jiang X, et al. Short- and Mid-Term Survival Prediction in Patients with Acute Type A Aortic Dissection Undergoing Surgical Repair: Based on the Systemic Immune-Inflammation Index. Journal of Inflammation Research. 2022;Volume 15:5785–99. Xu H, Li Y, Wang H, Yuan Y, Chen D, Sun Y, et al. Systemic immune‐inflammation index predicted short‐term outcomes in ATAD patients undergoing surgery. Journal of Cardiac Surgery. 2022;37:969–75. Silva GFS, Fagundes TP, Teixeira BC, Chiavegatto Filho ADP. Machine Learning for Hypertension Prediction: a Systematic Review. Current Hypertension Reports. 2022;24:523–33. Black J, Kueper JK, Williamson T. An introduction to machine learning for classification and prediction. Family Practice [Internet]. 2022 [cited 2023 Nov 4];40:200–4. Available from: https://academic.oup.com/fampra/article/40/1/200/6742730?login=false Azmi J, Arif M, Nafis MT, Alam MA, Tanweer S, Wang G. A systematic review on machine learning approaches for cardiovascular disease prediction using medical big data. Medical Engineering & Physics. 2022;105:103825. Parashar G, Chaudhary A, Pandey D. Machine Learning for Prediction of Cardiovascular Disease and Respiratory Disease: A Review. SN Computer Science. 2024;5. Ou H, Yao Y, He Y. Missing Data Imputation Method Combining Random Forest and Generative Adversarial Imputation Network. Sensors. 2024;24:1112–2. Dablain D, Krawczyk B, Chawla NV. DeepSMOTE: Fusing Deep Learning and SMOTE for Imbalanced Data. IEEE Transactions on Neural Networks and Learning Systems. 2022;1–15. Liu K, Chen K, Jia K. Convolutional Fine-Grained Classification With Self-Supervised Target Relation Regularization. IEEE Transactions on Image Processing. 2022;31:5570–84. Cao Y, Bass GA, Ahl R, Pourlotfi A, Geijer H, Montgomery S, et al. The statistical importance of P-POSSUM scores for predicting mortality after emergency laparotomy in geriatric patients. BMC Medical Informatics and Decision Making. 2020;20. Goodswen SJ, Barratt JLN, Kennedy PJ, Kaufer A, Calarco L, Ellis JT. Machine learning and applications in microbiology. FEMS microbiology reviews [Internet]. 2021 [cited 2022 Jan 14];45:fuab015. Available from: https://pubmed.ncbi.nlm.nih.gov/33724378/ Sylvester S, Sagehorn M, Gruber T, Atzmueller M, Schöne B. SHAP value-based ERP analysis (SHERPA): Increasing the sensitivity of EEG signals with explainable AI methods. Behavior Research Methods. 2024; Meng M, Zhao C. Application of Support Vector Machines to a Small-Sample Prediction. Advances in petroleum exploration and development. 2015;10:72–5. Chen J, Yin Y, Han L, Zhao F. Optimization Approaches for Parameters of SVM. Lecture notes in electrical engineering. 2019;575–83. Shehab M, Abualigah L, Shambour Q, Abu-Hashem MA, Shambour MKY, Alsalibi AI, et al. Machine learning in medical applications: A review of state-of-the-art methods. Computers in Biology and Medicine. 2022;145:105458. Guyon I, Weston J, Barnhill S, Vapnik V. Gene Selection for Cancer Classification Using Support Vector Machines. Machine Learning. 2002;46:389–422. Uchikov P, Khalid U, Harris D-SG, Ghale D, Rajadurai H, Kraeva M, et al. Artificial Intelligence in Breast Cancer Diagnosis and Treatment: Advances in Imaging, Pathology, and Personalized Care. Life [Internet]. 2024;14:1451. Available from: https://www.mdpi.com/2075-1729/14/11/1451 Zhang R, Chen S, Zhang H, Wang W, Xing J, Wang Y, et al. Biomarkers Investigation for In-Hospital Death in Patients With Stanford Type A Acute Aortic Dissection. International Heart Journal. 2016;57:622–6. Liu J, Sun L-L, Wang J, Ji G. The relationship between fibrinogen and in-hospital mortality in patients with type A acute aortic dissection. The American Journal of Emergency Medicine. 2017;36:741–4. Oz K, Taner Iyigun, Karaman Z, Ömer Çelik, Ertan Akbay, Okan Akınc, et al. Prognostic Value of Neutrophil to Lymphocyte Ratio and Risk Factors for Mortality in Patients with Stanford Type A Aortic Dissection. The heart surgery forum/The Heart surgery forum. 2017;20:119–9. Liu J, Sun L-L, Wang J, Ji G. Blood urea nitrogen in the prediction of in-hospital mortality of patients with acute aortic dissection. Cardiology Journal. 2018;25:371–6. Bedel C, Fatih Selvi. Association of Platelet to Lymphocyte and Neutrophil to Lymphocyte Ratios with In-Hospital Mortality in Patients with Type A Acute Aortic Dissection. Brazilian Journal of Cardiovascular Surgery. 2019;34. Guo T, Zhou X, Zhu A, Peng W, Zhong Y, Chai X. The Role of Serum Tenascin-C in Predicting In-Hospital Death in Acute Aortic Dissection. International Heart Journal. 2019;60:919–23. Zhang Y, Xu X, Lu Y, Guo L, Ma L. Preoperative uric acid predicts in-hospital death in patients with acute type a aortic dissection. Journal of Cardiothoracic Surgery. 2020;15. Maalouf M. Logistic regression in data analysis: an overview. International Journal of Data Analysis Techniques and Strategies. 2011;3:281. Poullis M. 080 * LIMITATIONS OF COX REGRESSION FOR SURVIVAL ANALYSIS IN THORACIC SURGERY. Interactive CardioVascular and Thoracic Surgery. 2013;17:S88–8. Guo T, Fang Z, Yang G, Zhou Y, Ding N, Peng W, et al. Machine Learning Models for Predicting In-Hospital Mortality in Acute Aortic Dissection Patients. Frontiers in Cardiovascular Medicine. 2021;8. Wu Z, Li Y, Xu Z, Liu H, Liu K, Qiu P, et al. Prediction of preoperative in-hospital mortality rate in patients with acute aortic dissection by machine learning: a two-centre, retrospective cohort study. BMJ open. 2023;13:e066782–2. Xie L, Xie Y, Wu Q, He J, Lin X, Qiu Z, et al. A predictive model for postoperative adverse outcomes following surgical treatment of acute type A aortic dissection based on machine learning. Journal of Clinical Hypertension. 2024;26:251–61. Zhang J, Xiong W, Yang J, Sang Y, Zhen H, Tan C, et al. Enhanced machine learning models for predicting one-year mortality in individuals suffering from type A aortic dissection. Journal of Thoracic and Cardiovascular Surgery. 2024; Liu H, Zhang S, Zhang C, Gao Q, Liu Y, Liao F, et al. Risk factors for prolonged postoperative ICU stay in the patients with Stanford type A aortic dissection. Journal of Cardiothoracic Surgery [Internet]. 2024;19. Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/pmid/38310273/ Zhu X, Wang J, Chong H, Jiang Y, Fan F, Pan J, et al. Analysis of risk factors for postoperative mortality in acute type A aortic dissection patients under different critical levels. Scientific Reports [Internet]. 2023;13:8107. Available from: https://www.nature.com/articles/s41598-023-35351-w Yao R, Yan D, Fu X, Deng Y, Xie X, Li N. The effects of plasma to red blood cells transfusion ratio on in-hospital mortality in patients with acute type A aortic dissection. Frontiers in Cardiovascular Medicine. 2023;10. Magouliotis DE, Rad AA, Viviano A, Oo AY, Xanthopoulos A, Serge Sicouri, et al. Hemostatic Properties of Aortic Root Preservation versus Root Replacement for Acute Type A Aortic Dissection: A Pooled Analysis. Life. 2024;14:1255–5. Mazzolai L, Teixido-Tura G, Lanzi S, Boc V, Bossone E, Brodmann M, et al. 2024 ESC Guidelines for the management of peripheral arterial and aortic diseases. European Heart Journal. 2024; Salerno S, Li Y. High-Dimensional Survival Analysis: Methods and Applications. Annual Review of Statistics and Its Application. 2022;10. Spooner A, Chen E, Sowmya A, Sachdev P, Kochan NA, Trollor J, et al. A comparison of machine learning methods for survival analysis of high-dimensional clinical data for dementia prediction. Scientific Reports. 2020;10. Additional Declarations No competing interests reported. Supplementary Files supplementarymaterial.docx Cite Share Download PDF Status: Published Journal Publication published 15 Apr, 2025 Read the published version in European Journal of Medical Research → Version 1 posted Editorial decision: Revision requested 29 Jan, 2025 Reviews received at journal 20 Jan, 2025 Reviews received at journal 18 Jan, 2025 Reviewers agreed at journal 15 Jan, 2025 Reviews received at journal 15 Jan, 2025 Reviewers agreed at journal 15 Jan, 2025 Reviewers agreed at journal 15 Jan, 2025 Reviewers invited by journal 15 Jan, 2025 Editor assigned by journal 09 Jan, 2025 Submission checks completed at journal 09 Jan, 2025 First submitted to journal 08 Jan, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5786813","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":400284776,"identity":"12090e5e-ce52-4267-a518-8b3eb1e77f31","order_by":0,"name":"Hao Cai","email":"","orcid":"","institution":"The First Affiliated Hospital of Chongqing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Hao","middleName":"","lastName":"Cai","suffix":""},{"id":400284777,"identity":"a6de44b6-c650-4577-a18b-6923796e587c","order_by":1,"name":"Yue Shao","email":"","orcid":"","institution":"The First Affiliated Hospital of Chongqing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Yue","middleName":"","lastName":"Shao","suffix":""},{"id":400284778,"identity":"b39d789b-a17d-4ce4-814e-86604ec50ce3","order_by":2,"name":"Xuan-yu Liu","email":"","orcid":"","institution":"The First Affiliated Hospital of Chongqing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xuan-yu","middleName":"","lastName":"Liu","suffix":""},{"id":400284779,"identity":"f694b9e3-f7e0-41dd-a7b3-7e25613a19a5","order_by":3,"name":"Chang-ying Li","email":"","orcid":"","institution":"The First Affiliated Hospital of Chongqing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Chang-ying","middleName":"","lastName":"Li","suffix":""},{"id":400284780,"identity":"05f9046f-aceb-4af4-afdd-e015e1060a1a","order_by":4,"name":"Hao-yu Ran","email":"","orcid":"","institution":"The First Affiliated Hospital of Chongqing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Hao-yu","middleName":"","lastName":"Ran","suffix":""},{"id":400284781,"identity":"956e41f3-41c2-4026-acd9-309622a319e2","order_by":5,"name":"Hao-ming Shi","email":"","orcid":"","institution":"The First Affiliated Hospital of Chongqing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Hao-ming","middleName":"","lastName":"Shi","suffix":""},{"id":400284783,"identity":"915df6bd-7036-4f85-84b6-61af2d3d2cb6","order_by":6,"name":"Cheng Zhang","email":"","orcid":"","institution":"The First Affiliated Hospital of Chongqing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Cheng","middleName":"","lastName":"Zhang","suffix":""},{"id":400284787,"identity":"27bdb16a-add3-4028-a211-b42487fb05de","order_by":7,"name":"Qing-chen Wu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA50lEQVRIiWNgGAWjYJCCA0CcAGXb8PCzN5CmJU1GsucAcTbBtBy2MbjhgF+pwY3kjYcLftnl8Us3P7x1s+08D8MNBsYPH3PwaUkrODyzL7lYcs4xY+vctts8jLMbmCVnbsOtxexGjsFh3p4DiRtu5LBJg7QwyxxgY+YlRst+iJZzPGwSCURo4fkBtEUCrOUADw8hLfZnnhUc5m1ITpxxI83YOudcMo8Ez8FmvH6RbE/e/Jnnj11i/4zkh7dzyuzs7Y83H/zwEY8WIDBgYGyDsCQgFGMDXvVgLQx/ULSMglEwCkbBKEAFAJ67Vwdy3EWBAAAAAElFTkSuQmCC","orcid":"","institution":"The First Affiliated Hospital of Chongqing Medical University","correspondingAuthor":true,"prefix":"","firstName":"Qing-chen","middleName":"","lastName":"Wu","suffix":""}],"badges":[],"createdAt":"2025-01-08 07:53:10","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5786813/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5786813/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s40001-025-02510-w","type":"published","date":"2025-04-15T15:57:37+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":73784267,"identity":"6f025bb0-bf6b-4a55-ad7b-1703260c6fba","added_by":"auto","created_at":"2025-01-14 15:57:52","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":73855,"visible":true,"origin":"","legend":"\u003cp\u003eWorkflow for Patient Inclusion (A), Feature Variable Selection (B).\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-5786813/v1/4e936544c74345ffd62cb6ee.png"},{"id":73784262,"identity":"b87a9815-6fb9-438c-8eba-8c7271781743","added_by":"auto","created_at":"2025-01-14 15:57:51","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":84129,"visible":true,"origin":"","legend":"\u003cp\u003eFeature variable selection based on the LASSO COX regression analysis. (A) Tuning parameter selection cross-validation error curve. (LASSO, least absolute shrinkage and selection operator); (B) Plot of the LASSO coefficient profiles.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-5786813/v1/63582e42b3d81922189ffa84.png"},{"id":73784899,"identity":"08af5510-bcef-4f7c-bfca-834afc25d232","added_by":"auto","created_at":"2025-01-14 16:05:52","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":77612,"visible":true,"origin":"","legend":"\u003cp\u003eCorrelation Heatmap of Continuous Variables in the Model.\u003c/p\u003e\n\u003cp\u003eAbbreviations: POHS Postoperative Hospital Stay, ICUHS Intensive Care Unit Hospital Stay, WBC White Blood Cell count.\u003c/p\u003e\n\u003cp\u003eNote: ‘*’ denotes p\u0026lt;0.05.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-5786813/v1/60d6a37f80d17b812d251b2f.png"},{"id":73784273,"identity":"d6684fe7-77e2-4ee6-bcd6-5148dfe9b989","added_by":"auto","created_at":"2025-01-14 15:57:52","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":325920,"visible":true,"origin":"","legend":"\u003cp\u003eComprehensive Performance Evaluation of the SVM Model.\u003c/p\u003e\n\u003cp\u003e(A) ROC curve with AUC values for the training and test sets; (B) PR curve with AP values for the training and test sets; (C) Accuracy curve illustrating the performance across different thresholds for the training and test sets; (D) Precision curve showing the precision at various thresholds for the training and test sets; (E) DCA curve comparing net benefit across thresholds for the training and test sets.\u003c/p\u003e\n\u003cp\u003eAbbreviations: SVM Support Vector Machine, ROC Receiver Operating Characteristic.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-5786813/v1/b64a7a73ef8beeac686c206f.png"},{"id":73784275,"identity":"bd1098c7-8aea-4cf3-9bba-75c085728145","added_by":"auto","created_at":"2025-01-14 15:57:52","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":42339,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP analysis of SVM model. (A) Ranking feature importance based on the absolute mean values of SHAp values; (B) The scatter plot of feature distributions using the SHAP analysis.\u003c/p\u003e\n\u003cp\u003eAbbreviations: POHS Postoperative Hospital Stay, ICUHS Intensive Care Unit Hospital Stay, WBC White Blood Cell count, AP Abdominal Pain, SII Systemic Immune-Inflammation Index, ToPlasma Transfusion of Plasma.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-5786813/v1/67ae44a81f163d3263a0acb0.png"},{"id":73784278,"identity":"7f64dd85-e1bd-48f5-8676-acdedd66fdbb","added_by":"auto","created_at":"2025-01-14 15:57:52","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":35400,"visible":true,"origin":"","legend":"\u003cp\u003eThe SVM model SHAP decision plot. (A) Individual decision-making processes in the survivor group; (B) Individual decision-making processes in the outcome group.\u003c/p\u003e\n\u003cp\u003eAbbreviations: POHS Postoperative Hospital Stay, ICUHS Intensive Care Unit Hospital Stay, WBC White Blood Cell count, AP Abdominal Pain, SII Systemic Immune-Inflammation Index, ToPlasma Transfusion of Plasma.\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-5786813/v1/16eb6cabf7228081b41d5e73.png"},{"id":81050984,"identity":"a0d2f823-cd82-43dd-95cc-5d77036fa592","added_by":"auto","created_at":"2025-04-21 16:09:21","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1542465,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5786813/v1/6331147f-5d3f-465f-b7f3-76911b92ebff.pdf"},{"id":73784264,"identity":"b2f4af32-537b-4dd9-9f20-291ef9d69ce1","added_by":"auto","created_at":"2025-01-14 15:57:52","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":226831,"visible":true,"origin":"","legend":"","description":"","filename":"supplementarymaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-5786813/v1/f2014f994ed76d49772f1667.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Interpretable Prognostic Modeling for Long-Term Survival of Type A Aortic Dissection Patients Using Support Vector Machine Algorithm","fulltext":[{"header":"Introduction","content":"\u003cp\u003eAortic dissection is a life-threatening condition with an increasing incidence worldwide [1]. According to the Stanford classification system, established in the last century, Type A aortic dissection (TAAD) involves the ascending aorta and may extend to the aortic arch and descending aorta. Unlike Type B aortic dissection, TAAD has well-defined emergency indications for surgical repair [3]. Early studies indicate that the mortality rate for TAAD increases by 1\u0026ndash;2% per hour during the first 48 hours [2]. Despite advancements in surgical techniques and life support systems, the prognosis for TAAD patients remains poor due to the time-critical nature of the condition and a high risk of complications in the early stages. Recent data reveal that in-hospital and 30-day mortality rates for TAAD surgery exceed 20%, underscoring the urgent need for improved prognostic strategies [4,5]. Identifying high-risk patients is therefore crucial for optimizing prognosis management and guiding clinical decisions.\u003c/p\u003e \u003cp\u003ePerioperative conditions, such as ICU stay duration, and specific biomarkers, including C-reactive protein (CRP) levels, have been identified as significant prognostic factors. For instance, Liu et al. found that TAAD patients with ICU stays longer than 7 days face increased risks of adverse outcomes, including severe organ damage and respiratory complications [6]. Similarly, Tang et al. demonstrated that elevated preoperative CRP levels are independent predictors of in-hospital mortality, renal dysfunction, and stroke in TAAD patients [7]. Moreover, the Systemic Immune-Inflammation Index (SII), which combines neutrophil, lymphocyte, and platelet counts, has shown promise in predicting postoperative complications and three-year survival [8,9]. Despite their statistical significance, these indicators are not widely implemented in clinical practice, primarily due to their limited ability to account for the complexity of TAAD prognosis in real-world settings.\u003c/p\u003e \u003cp\u003eThe limited clinical application of these indicators may stem from their narrow focus on either short-term outcomes or long-term survival, which fails to address the broader, multi-organ impact of TAAD. Moreover, an exclusive focus on significantly abnormal markers may overlook the interrelationships between variables, leading to missing key information and underestimating the prognostic value of certain factors. Most of the prognostic indicators identified in previous studies have predominantly been used for preoperative risk stratification to guide surgical timing. Unlike other acute or malignant conditions, early surgical intervention is critical to mitigate risks of aortic rupture and mortality[42]. Thus, relying solely on preoperative assessments is insufficient for comprehensive prognostic management. Addressing this gap necessitates the development of models that integrate multiple perioperative, biomarker, and clinical factors to improve risk stratification accuracy and guide both pre- and postoperative decisions.\u003c/p\u003e \u003cp\u003eTraditional regression methods, such as logistic regression and Cox regression, provide valuable insights but struggle to manage high-dimensional data and complex interactions, potentially overlooking critical prognostic factors [43,44]. In contrast, machine learning (ML) techniques excel in identifying complex patterns within datasets, offering advantages in model optimization and predictive accuracy [10]. ML has demonstrated significant potential in healthcare, particularly in predicting health outcomes and disease progression in cardiovascular conditions [10\u0026ndash;13]. However, the application of ML to TAAD prognosis research, especially for long-term survival prediction, remains underexplored.\u003c/p\u003e \u003cp\u003eThis study aims to develop and validate a practical ML model utilizing SHapley Additive exPlanations (SHAP)-based visualization to identify key prognostic factors and stratify TAAD patients at high risk of poor outcomes. By enhancing the interpretability and accuracy of prognostic assessments, this model seeks to support clinicians in devising effective treatment strategies and improving patient outcomes.\u003c/p\u003e"},{"header":"Material and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003ePatients\u003c/h2\u003e \u003cp\u003eIn this retrospective cohort study, we analyzed the clinical data of patients with Stanford Type A Aortic Dissection (TAAD) who consecutively underwent surgical repair at the Department of Thoracic and Cardiovascular Surgery, First Affiliated Hospital of Chongqing Medical University, from September 2017 to December 2020. The inclusion criteria for this study were as follows: (1) patients diagnosed with Stanford TAAD following aortic computed tomography angiography (CTA); (2) patients who underwent surgical repair; (3) patients aged 18 years or older. The exclusion criteria included: (1) patients with more than 20% absence of clinical or laboratory data; (2) patients with preoperative comorbidities such as malignant tumors, hematological disorders, infections, systemic inflammatory diseases, or undergoing treatments that could influence biomarker levels and survival; (3) patients who died directly or indirectly from causes other than TAAD. Ultimately, 175 patients were included in the study (\u003cb\u003eFig.\u0026nbsp;1A\u003c/b\u003e). The study was approved by the Independent Ethics Committee of the First Affiliated Hospital of Chongqing Medical University (approval number: 2024-583-01), and the study was conducted in compliance with the ethical standards of the World Medical Association Declaration of Helsinki.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eData Collection\u003c/h3\u003e\n\u003cp\u003eDemographic characteristics, clinical symptoms, and hemodynamic profiles at admission\u0026mdash;including gender, age, smoking history, hypertension (HTN), diabetes mellitus (DM), cardiovascular disease (CVD), blood pressure, and heart rate (HR)\u0026mdash;and intraoperative details such as blood loss, operation time, cardiopulmonary bypass time (CPB), aortic cross-clamp time (ACC), red blood cell transfusion volume, and plasma transfusion volume were retrospectively reviewed and extracted from electronic medical records. Blood indicators were assessed within 24 hours of hospital admission and prior to surgery. The biomarkers measured included red blood cell (RBC) count, creatinine, absolute neutrophil count (ANC), white blood cell (WBC) count, hemoglobin (Hb), platelet count, absolute lymphocyte count (ALC), monocyte count, serum albumin, uric acid, urea nitrogen, alanine aminotransferase (ALT), aspartate aminotransferase (AST), cardiac troponin T (cTnT), myoglobin, fibrinogen, and D-dimer, etc. A composite metric, the Systemic Immune-Inflammation Index (SII), was calculated using the following formula: platelet count \u0026times; neutrophil count / lymphocyte count.​\u003c/p\u003e\n\u003ch3\u003eFollow-up and Treatments\u003c/h3\u003e\n\u003cp\u003eAll patients included in this study underwent surgical repair. The primary outcome was patient mortality. Overall survival (OS) was defined as the time from surgery to either death or the last follow-up. Follow-up started three months after discharge and concluded on August 20, 2024. Telephone follow-ups, conducted by trained interviewers, were performed at 3, 6, and 12 months after discharge, and subsequently every six months to support post-discharge care. The maximum OS observed was 2,445 days, with a median OS of 1,190 days.\u003c/p\u003e\n\u003ch3\u003eData Preprocessing and Feature Variable Selection\u003c/h3\u003e\n\u003cp\u003eFor features with a missing data rate below 20%, random forest imputation was employed [14]. The dataset was then randomly divided into an 80% training set and a 20% test set. To reduce the risk of overfitting, feature selection was conducted using LASSO Cox regression analysis. Features were excluded based on regression results and clinical relevance, yielding an initial model with 16 features, including 2 categorical and 14 continuous variables. Subsequently, univariate analysis and correlation analysis were performed on the preliminarily selected variables. Features were finalized for inclusion in the machine learning model based on statistical significance, clinical relevance, correlation coefficients and LASSO Cox regression coefficients (Fig.\u0026nbsp;1B). To address data imbalance, data normalization was performed first, followed by the application of the Synthetic Minority Over-sampling Technique (SMOTE) to the training set [15,16].\u003c/p\u003e\n \u003c/div\u003e\n\u003ch3\u003eModel Selection and Performance Evaluation\u003c/h3\u003e\n\u003cp\u003eThe Support Vector Machine (SVM) stands out among machine learning models for its suitability in small-sample datasets, offering significant theoretical and algorithmic advantages. Based on statistical learning theory, SVM utilizes the Structural Risk Minimization (SRM) principle, which prioritizes minimizing generalization error over merely optimizing training error. This approach significantly enhances the model's ability to generalize, especially when working with small sample sizes. Unlike traditional models, such as neural networks, which depend on large datasets to approximate the target function, SVM can effectively predict unseen data by maximizing the margin between classes, even with limited sample sizes [20].\u003c/p\u003e \u003cp\u003eSVM is particularly effective in handling small sample, high-dimensional datasets, offering a robust solution to the \"curse of dimensionality.\" By employing kernel functions, SVM maps the original feature space into a higher-dimensional space, enabling the identification of an optimal, linearly separable hyperplane without the need to explicitly compute the high-dimensional mappings [21]. Additionally, when combined with LASSO Cox regression analysis, SVM incorporates regularization techniques to control model complexity and reduce the risk of overfitting. These features make SVM a widely adopted approach in various fields, including disease classification (e.g., early cancer diagnosis), survival analysis, and treatment recommendation [22\u0026ndash;24].\u003c/p\u003e \u003cp\u003eGiven the distinct advantages of Support Vector Machines (SVM), their widespread application in medicine and bioinformatics, and the characteristics of the dataset in this study, the SVM model was chosen for subsequent analysis. A 10-fold cross-validation strategy was employed to ensure a comprehensive evaluation of model performance. The training data was divided into 10 subsets, with each iteration using 9 subsets for model training and the remaining subset for validation. This process was repeated 10 times, allowing each subset to serve as the validation set once. Key performance metrics used to evaluate the model's predictive and generalization ability included the area under the receiver operating characteristic curve (AUC), accuracy, precision, recall, F1 score, Brier score, and the area under the precision-recall curve (AP) [18].\u003c/p\u003e \u003cp\u003eTo enhance model interpretability, this study incorporated the SHAP (SHapley Additive exPlanations) method. SHAP quantifies the marginal contribution of each feature to the model\u0026rsquo;s predicted outcomes, offering both global and local insights into the decision-making process. By assigning precise attribution values to each variable, SHAP effectively identifies the most influential features for model predictions [19].\u003c/p\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003eStatistical Analysis\u003c/h2\u003e \u003cp\u003eThis study included the demographic characteristics, risk factors, admission status, preoperative blood markers, and intraoperative conditions of TAAD patients. Data analysis was performed using SPSS 27.0 and Python 3.1. Continuous variables are presented as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation (SD), with comparisons made using Student's t-test or Mann-Whitney U test. Categorical variables are presented as frequencies and percentages (%) and analyzed using chi-square or Fisher's exact test. Pearson\u0026rsquo;s chi-squared test was used to evaluate correlations among continuous variables, visualized through a heatmap. The point-biserial correlation coefficient assessed correlation between categorical and continuous variables, with a bubble plot highlighting their strength and significance. The model was constructed using the SVM algorithm, and performance was evaluated using AUC, accuracy, precision, recall, F1 score, Brier score, and AP. Finally, the SHAP method was used for model interpretation. A two-tailed p-value of \u0026lt;\u0026thinsp;0.05 was considered statistically significant.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003ePatient characteristics\u003c/h2\u003e \u003cp\u003eA total of 175 TAAD patients were included in the study (\u003cb\u003eTable \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e\u003c/b\u003e), with 54 deaths observed during the follow-up period. The average age of the cohort was 48.87\u0026thinsp;\u0026plusmn;\u0026thinsp;9.57 years, and the majority were male (139 cases, 79.43%). While only a small proportion of patients (6.29%, 11 cases) experienced concomitant shock, this group had a markedly higher mortality rate compared to those without shock (63.64% vs. 28.66%).\u003c/p\u003e \u003cp\u003eAt admission, abdominal pain was reported in 9.71% of patients (17 cases) and was more prevalent in the outcome group, with 18.52% (10 cases) reporting this symptom. This observation suggests a potential association between abdominal pain and the severity of TAAD. Additionally, patients with endpoint events had significantly shorter postoperative hospital stays compared to survivors (10.77\u0026thinsp;\u0026plusmn;\u0026thinsp;10.29 days vs. 23.13\u0026thinsp;\u0026plusmn;\u0026thinsp;14.42 days). This likely reflects the fact that critically ill patients are more prone to early mortality and fail to benefit from postoperative care.\u003c/p\u003e \u003cp\u003e175 TAAD patients were randomly assigned to the training (140 patients, 80%) and test (35 patients, 20%) groups. No significant differences were found between the two groups in terms of demographic data, risk factors, admission conditions, hospital stays, preoperative laboratory results, and intraoperative conditions (\u003cb\u003eTable S2\u003c/b\u003e).\u003c/p\u003e \u003c/div\u003e \n\u003cp\u003eTable 1. Comparison of Perioperative Conditions: Variables Selected through LASSO Cox Regression for Initial Screening in the Training Set\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"577\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003eVariables\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 172px;\"\u003e\n \u003cp\u003eNo. of outcomes (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eNo. of survivors (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67px;\"\u003e\n \u003cp\u003eP value\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003eShock\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 172px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.007\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 172px;\"\u003e\n \u003cp\u003e33 (84.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003e97 (96.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 172px;\"\u003e\n \u003cp\u003e6 (15.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003e4 (4.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003eAbdominal pain\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 172px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.043\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 172px;\"\u003e\n \u003cp\u003e31 (79.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003e94 (93.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 172px;\"\u003e\n \u003cp\u003e8 (20.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003e7 (6.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003ePOHS (days)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 172px;\"\u003e\n \u003cp\u003e12.41 \u0026plusmn; 10.53\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 149px;\"\u003e\n \u003cp\u003e22.96 \u0026plusmn; 14.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003eICUHS (days)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 172px;\"\u003e\n \u003cp\u003e11.51 \u0026plusmn; 10.42\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 149px;\"\u003e\n \u003cp\u003e12.66 \u0026plusmn; 8.63\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.542\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003eWBC (\u0026times;10\u0026circ;9/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 172px;\"\u003e\n \u003cp\u003e13.99 \u0026plusmn; 5.15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 149px;\"\u003e\n \u003cp\u003e11.55 \u0026plusmn; 4.08\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.01\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003eAST (IU/L)\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 172px;\"\u003e\n \u003cp\u003e109.44 \u0026plusmn; 322.92\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 149px;\"\u003e\n \u003cp\u003e39.35 \u0026plusmn; 68.05\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.187\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003eCreatinine (\u0026mu;mol/L)\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 172px;\"\u003e\n \u003cp\u003e113.41 \u0026plusmn; 52.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 149px;\"\u003e\n \u003cp\u003e85.63 \u0026plusmn; 34.32\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.004\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003eFibrinogen (g/L)\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 172px;\"\u003e\n \u003cp\u003e3.14 \u0026plusmn; 2.06\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 149px;\"\u003e\n \u003cp\u003e3.05 \u0026plusmn; 1.66\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.801\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003eBNP (pg/mL)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 172px;\"\u003e\n \u003cp\u003e971.81 \u0026plusmn; 1186.24\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 149px;\"\u003e\n \u003cp\u003e953.77 \u0026plusmn; 1535.51\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.941\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003eSII\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 172px;\"\u003e\n \u003cp\u003e3746.52 \u0026plusmn; 6199.33\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 149px;\"\u003e\n \u003cp\u003e2340.46 \u0026plusmn; 2270.78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.075\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003eOperation time (min)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 172px;\"\u003e\n \u003cp\u003e624.69 \u0026plusmn; 175.27\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 149px;\"\u003e\n \u003cp\u003e523.07 \u0026plusmn; 112.79\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003eIntraoperative bleeding (ml)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 172px;\"\u003e\n \u003cp\u003e930.98\u0026nbsp;\u0026plusmn;\u0026nbsp;1148.21\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 149px;\"\u003e\n \u003cp\u003e582.0\u0026nbsp;\u0026plusmn;\u0026nbsp;467.63\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.073\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003eTransfusion of SRBC (U)\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 172px;\"\u003e\n \u003cp\u003e4.13\u0026nbsp;\u0026plusmn;\u0026nbsp;3.46\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 149px;\"\u003e\n \u003cp\u003e3.32\u0026nbsp;\u0026plusmn;\u0026nbsp;2.95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.202\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 189px;\"\u003e\n \u003cp\u003eTransfusion of Plasma (ml) \u0026nbsp; \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 172px;\"\u003e\n \u003cp\u003e944.65\u0026nbsp;\u0026plusmn;\u0026nbsp;419.55\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 149px;\"\u003e\n \u003cp\u003e750.3\u0026nbsp;\u0026plusmn;\u0026nbsp;371.83\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.014\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003eCPB time (min)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 172px;\"\u003e\n \u003cp\u003e305.9\u0026nbsp;\u0026plusmn;\u0026nbsp;86.56\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 149px;\"\u003e\n \u003cp\u003e264.15\u0026nbsp;\u0026plusmn;\u0026nbsp;63.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.008\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eAbbreviations: POHS Postoperative Hospital Stay, ICUHS Intensive Care Unit Hospital Stay, WBC White Blood Cell count, AST Aspartate Aminotransferase, BNP B-type Natriuretic Peptide, SII Systemic Immune-Inflammation Index, SRBC: Suspended Red Blood Cells, CPB Cardiopulmonary Bypass.\u0026nbsp;\u003c/p\u003e\n\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eFeature Variable Selection\u003c/h2\u003e \u003cp\u003eLASSO (Least Absolute Shrinkage and Selection Operator) Cox regression was used to preliminarily screen variables associated with long-term survival, yielding 16 perioperative characteristic variables. The optimal lambda value, determined through cross-validation, was 25.9502 (\u003cb\u003eFig.\u0026nbsp;2A, B\u003c/b\u003e). Based on these findings, univariate analysis was performed on perioperative characteristics between survivors and patients who experienced endpoint events in the training set (\u003cb\u003eTable\u0026nbsp;1\u003c/b\u003e). To avoid information leakage from the test set, this analysis was limited to the training set. Variables with p-values\u0026thinsp;\u0026lt;\u0026thinsp;0.05 were considered statistically significant. Among them, CBP and shock had small coefficients in the LASSO Cox regression, suggesting its limited contribution to the model, and was thus excluded.\u003c/p\u003e \u003cp\u003eIn the training group, the difference in preoperative SII values between survivor and outcome group did not reach statistical significance (p\u0026thinsp;=\u0026thinsp;0.075, \u003cb\u003eTable\u0026nbsp;1\u003c/b\u003e), likely due to large fluctuations in the SII data. However, given its relatively high regression coefficient in the LASSO Cox regression model, as well as its role as an important marker of inflammation and immune function, SII may still have significant prognostic value for TAAD patients. Therefore, we decided to include SII in further analysis.\u003c/p\u003e \u003c/div\u003e \n\u003cp\u003e\u003cstrong\u003eCorrelation Among Selected Variables Included in the Model\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAfter further screening, eight variables were selected for model training based on the results of LASSO Cox regression analysis, univariate analysis, and clinical relevance. These variables included seven continuous variables and one categorical variable: ICU hospital stay, postoperative hospital stay, plasma transfusion, creatinine, operation time, SII, WBC, and abdominal pain.\u003c/p\u003e\n\u003cp\u003eThe heatmap presented Pearson\u0026rsquo;s correlation coefficients among continuous variables, providing a visual depiction of the strength and direction of these correlations. The color gradient, ranging from dark blue to dark red, represented the spectrum from weak to strong correlations. Each cell in the heatmap was labeled with a Pearson\u0026rsquo;s correlation coefficient, quantifying the linear relationship between two variables. A coefficient of 1 indicated a perfect positive correlation, -1 signified a perfect negative correlation, and 0 reflects no correlation. Similarly, the bubble plot illustrated the Point-Biserial correlation coefficients between the categorical variable (abdominal pain) and the continuous variables, with the size and position of the bubbles representing the strength and direction of the correlations, respectively. Asterisks were used to denote statistical significance, with a single asterisk (*) representing a p-value of less than 0.05.\u003c/p\u003e\n\u003cp\u003ePearson\u0026apos;s chi-squared test demonstrated significant correlations between WBC and creatinine, operation time, and SII (p\u0026lt;0.001, p\u0026lt;0.05, p\u0026lt;0.001, respectively), with Pearson correlation coefficients of 0.29, 0.19, and 0.44, indicating weak to moderate relationships (\u003cstrong\u003eFig.3\u003c/strong\u003e). Conversely, no significant correlations were identified between WBC and ICU hospital stay, postoperative hospital stay, or plasma transfusion (p\u0026gt;0.05). Notably, a strong correlation was observed between ICU hospital stay and postoperative hospital stay (Pearson correlation coefficient: 0.74, p\u0026lt;0.001), suggesting that prolonged ICU stays are associated with extended postoperative hospitalizations. This strong relationship likely reflects the complexity of postoperative recovery and the severity of the patient\u0026apos;s condition. Hospital stay, as a metric, may also serve as an indirect indicator of the effectiveness of comprehensive management during hospitalization and the patient\u0026rsquo;s overall risk profile. Consequently, ICU stay and postoperative hospital stay were included for further analysis. For the remaining variables, the correlation strengths were generally weak or lacked statistical significance (\u003cstrong\u003eFig. 3 and Fig.S1\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eThe Performance of the SVM Model\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAfter selecting the feature variables, the SMOTE method was applied to the training set to address data imbalance. Subsequently, the SVM algorithm was implemented, and model generalization was enhanced using 10-fold cross-validation, with final validation performed on the test set. The model demonstrated excellent performance, achieving AUC values exceeding 0.85 for both the training and test sets, indicating strong discriminatory ability (\u003cstrong\u003eTable 2, Fig. 4A\u003c/strong\u003e). Specifically, the training set achieved an AUC of 0.9247 (95% CI: 0.9200-0.9279), while the test set achieved an AUC of 0.8800 (95% CI: 0.8492-0.9396). The slightly lower AUC for the test set suggests some performance decline, but the difference remains within a reasonable range, with no significant signs of overfitting.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe model demonstrated consistent accuracy across both the training and test sets, with the test set achieving slightly higher accuracy (training set: 0.8663, test set: 0.8857). Notably, the test set achieved 100% precision, underscoring the model\u0026rsquo;s reliability in predicting TAAD mortality (\u003cstrong\u003eTable 2, Fig. 4C and D\u003c/strong\u003e). Precision-Recall curves showed similar shapes and AP values above 0.9 for both sets, further confirming these results (\u003cstrong\u003eFig. 4B\u003c/strong\u003e). The Brier Scores for the training and test sets were notably low, at 0.1068 and 0.1070, respectively, indicating well-calibrated probabilistic predictions. The C-index values were 0.8901 and 0.8700 for the training and test sets, respectively, both close to 0.9, indicating robust predictive performance and strong generalizability. These results align with prior analyses, further supporting the reliability of the model.\u003c/p\u003e\n\u003cp\u003eIn the Decision Curve Analysis (DCA), the model effectively distinguished risk levels and provided high net benefits across a wide range of thresholds. Although the test set curve was slightly lower than the training set, the overall trends were consistent, reflecting strong generalization ability (\u003cstrong\u003eFig. 4E\u003c/strong\u003e). Performance metrics, including accuracy, precision, AUC, AP, Brier Score, F1 Score, and C-index, showed comparable results between the training and test sets, demonstrating the model\u0026rsquo;s robustness, reliability, and practical value in clinical applications (\u003cstrong\u003eFig. S2\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eTable 2. Performance of the SVM Model between Training Set and Test Set.\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"84%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003eMetrics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35px;\"\u003e\n \u003cp\u003eSVM (training)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 38px;\"\u003e\n \u003cp\u003eSVM (test)\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003eAUC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35px;\"\u003e\n \u003cp\u003e0.9247\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 38px;\"\u003e\n \u003cp\u003e0.8800\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e95%CI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35px;\"\u003e\n \u003cp\u003e(0.9200-0.9279)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 38px;\"\u003e\n \u003cp\u003e( 0.8492-0.9396)\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003eAccuracy\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35px;\"\u003e\n \u003cp\u003e0.8663\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 38px;\"\u003e\n \u003cp\u003e0.8857\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003ePrecision\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35px;\"\u003e\n \u003cp\u003e0.8627\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 38px;\"\u003e\n \u003cp\u003e1.0000\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003eRecall\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35px;\"\u003e\n \u003cp\u003e0.8713\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 38px;\"\u003e\n \u003cp\u003e0.7333\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003eF1 score\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35px;\"\u003e\n \u003cp\u003e0.8670\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 38px;\"\u003e\n \u003cp\u003e0.8462\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003eBrier score\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35px;\"\u003e\n \u003cp\u003e0.1068\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 38px;\"\u003e\n \u003cp\u003e0.1070\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003eAP\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35px;\"\u003e\n \u003cp\u003e0.9266\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 38px;\"\u003e\n \u003cp\u003e0.9086\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003eC-index\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35px;\"\u003e\n \u003cp\u003e0.8901\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 38px;\"\u003e\n \u003cp\u003e0.8700\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eAbbreviations: AUC Area Under the Curve, SVM Support Vector Machine, CI Confidence Interval, AP Average Precision, C-index Concordance Index.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSHAP Interpretation and Feature Importance Visualization\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe SHAP method was employed to interpret the SVM model\u0026apos;s predictions and evaluate its clinical relevance. By quantifying the contribution of each feature, SHAP values provided insights into their impact on the model\u0026apos;s outputs. Feature importance analysis identified POHS as the most critical predictor of long-term survival in TAAD patients, highlighting the potential benefits of comprehensive postoperative management for survivors (\u003cstrong\u003eFig. 5A\u003c/strong\u003e). Furthermore, the SHAP summary plot revealed that higher ICUHS, Plasma transfusion volume, creatinine, WBC, operation time, SII and abdominal pain positively influenced the model\u0026rsquo;s predictions, indicating their significant roles in survival outcomes (\u003cstrong\u003eFig. 5B\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eTo validate the model\u0026apos;s interpretability, decision curves were used to illustrate individualized predictions of long-term survival. The gray vertical line at 0 on the horizontal axis represented the model\u0026apos;s baseline. Fig. 6A visualizes the decision-making process for TAAD survivors, while Fig. 6B illustrates it for patients with endpoint events.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eTAAD is a life-threatening vascular emergency for which surgical repair remains the primary treatment. However, its management poses significant challenges, particularly in early diagnosis, perioperative complications, and mortality risk. The nonspecific TAAD symptoms may delay timely diagnosis, while timely imaging is constrained by both time and resource availability, allowing the condition to deteriorate rapidly. The severity of the disease, coupled with the risks associated with emergency surgery, contributes to substantial perioperative mortality and the potential for serious complications. As a result, prognostic management for TAAD patients has become a critical focus of ongoing research.\u003c/p\u003e\n\u003cp\u003eDespite advancements in prognostic assessment for TAAD, several challenges continue to limit the practical applicability of existing findings. For instance, Zhang et al. identified systolic blood pressure at admission, NT-proBNP, and white blood cell count as independent factors affecting in-hospital mortality among TAAD patients [25]. Similarly, numerous studies have highlighted preoperative indicators such as fibrinogen, BUN, NLR, PLR, D-dimer, UA, and CRP as prognostic markers for both short-term and long-term survival in TAAD patients [26-31]. However, relying solely on a limited set of preoperative indicators often fails to capture the full complexity of changes in a patient\u0026rsquo;s condition throughout the perioperative period, potentially overlooking critical factors. This limitation reduces the practical utility of these markers, even when statistically significant differences are observed.\u003c/p\u003e\n\u003cp\u003eIn addition, traditional Cox and logistic regression methods, while widely used, have inherent limitations. These methods struggle to manage complex, high-dimensional data and fail to fully capture intricate relationships between variables. Furthermore, their reliance on model assumptions can hinder their ability to generalize effectively to new datasets, thereby reducing their predictive performance and clinical applicability [32,33].\u003c/p\u003e\n\u003cp\u003eIn recent years, ML methods have gained increasing attention for prognostic assessment in TAAD. However, research in this area remains in its early stages. For example, Zhang et al. developed a Treebag model to predict one-year mortality in TAAD patients, using 51 clinical characteristics, including blood markers at admission [37]. Key survival predictors identified included surgical intervention, beta-blocker use during hospitalization, systolic blood pressure at admission, lymphocyte count, carbon dioxide combining capacity, eosinophil count, and white blood cell count. Nevertheless, relying exclusively on preoperative indicators and a single type of postoperative medication may fail to account for the dynamic fluctuations in TAAD patients\u0026rsquo; conditions over the perioperative period. This approach may also underestimate the potential benefits of a comprehensive perioperative management strategy.\u003c/p\u003e\n\u003cp\u003eMoreover, Most recent ML studies have primarily focused on short-term mortality and postoperative complications, with limited attention given to long-term survival prediction in TAAD patients [34-36]. Of note, studies relying solely on public database may exclude certain patient subgroups due to the database\u0026apos;s inclusion criteria. As a result, while statistically significant, these datasets may lack broad representativeness, limiting the model\u0026apos;s generalizability. Addressing these gap is crucial for enhancing the prognostic utility of ML approaches and improving their practical application in real-world clinical settings. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn this study, we used an interpretable machine learning approach to explore the relationship between preoperative vital signs, blood markers, clinical history, demographic factors, and long-term survival in TAAD patients. We specifically examined hospital stays to evaluate patients\u0026apos; overall clinical condition, hospitalization benefits, and perioperative risks. ICU stay reflected condition stability and critical care complexity, while postoperative hospital stay captured recovery progress and the effectiveness of comprehensive management. Together, these variables provided a multidimensional assessment of the balance between perioperative risks and long-term survival benefits, offering valuable insights into hospitalization outcomes in TAAD.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eBased on the data characteristics and the advantages of the SVM algorithm, we selected SVM for analysis. The SHAP method was applied to interpret feature importance and its relationship with long-term mortality. Our results showed strong performance of the SVM model in both training and test datasets, with postoperative hospital stay, ICU stay, and abdominal pain at admission identified as the most significant predictors. SHAP decision plots were used to further illustrate the model\u0026apos;s decision-making process.\u003c/p\u003e\n\u003cp\u003eFeature importance ranking and scatter plots identified postoperative hospital stay as the most critical factor, with a protective effect on long-term mortality in TAAD patients. This finding suggests that severely ill patients who are unable to tolerate prolonged hospital stays are more likely to experience early mortality, consistent with the observation that outcome group had shorter hospital stays. In contrast, patients with more stable conditions appeared to benefit from comprehensive postoperative management, resulting in improved long-term survival.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eICU hospital stay was identified as the second most significant feature. Scatter plots indicated that longer ICU stay were associated with higher long-term mortality, which aligns with previous studies [38]. Prolonged ICU stay did not improve outcomes but instead increased the risk of complications and adverse events. Additionally, while severe chest pain is commonly reported at admission, patients presenting with abdominal pain had a higher incidence of endpoint events compared to survivors. Scatter plots further highlighted abdominal pain at admission as a key predictor of long-term mortality, potentially signaling disease progression.\u003c/p\u003e\n\u003cp\u003eOther important features in the model included operation time, plasma transfusion volume, WBC, creatinine, and SII. Severe TAAD cases often require longer surgeries due to increased procedural complexity and coagulopathy, leading to higher plasma transfusion volumes\u0026mdash;observations consistent with the outcome group. Similar findings have been reported in prior studies, indicating that prolonged surgery and higher transfusion requirements are associated with poor clinical outcomes [39-42]. Furthermore, SII, a composite marker of inflammation and immune function, along with WBC and creatinine, reflects preoperative immune status, renal function, and inflammatory levels, all of which are linked to adverse prognoses in TAAD patients.\u003c/p\u003e\n\u003cp\u003eIn this study, we developed an SVM-based machine learning model that exhibited strong predictive performance on both training and test datasets. This model serves as a valuable clinical tool for assessing the long-term prognosis of TAAD patients, enabling more informed decision-making regarding critical aspects of care, such as surgical timing, ICU transfer, transition to general wards, and discharge planning. By providing a comprehensive evaluation of perioperative factors, the model supports physicians in optimizing short-term postoperative management while maximizing long-term survival outcomes, ultimately enhancing the quality of patient care.\u003c/p\u003e\n\u003cp\u003eDespite these contributions, this study has several limitations. First, although the SVM model showed good generalization with a small dataset, the limited sample size remains a concern. As a single-center retrospective study, the potential for selection bias may exist, underscoring the need for multi-center, large-scale studies to validate these findings. Second, while this study captured the impact of comprehensive perioperative management through hospital stay durations, it did not fully account for individual patient variability. Future research should aim to integrate fluctuations in perioperative blood markers and other individualized factors to refine and enhance the prognostic model.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study is the first to develop and validate a machine learning-based prognostic model for long-term survival in TAAD patients. The model demonstrated strong predictive performance in both training and testing groups, supporting its clinical potential. It offers clinicians a tool for dynamically assessing long-term outcomes in TAAD patients.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003e- TAAD: Type A Aortic Dissection\u003c/p\u003e\n\u003cp\u003e- ML: Machine Learning\u003c/p\u003e\n\u003cp\u003e- SVM: Support Vector Machine\u003c/p\u003e\n\u003cp\u003e- SHAP: SHapley Additive exPlanations\u003c/p\u003e\n\u003cp\u003e- LASSO: Least Absolute Shrinkage and Selection Operator\u003c/p\u003e\n\u003cp\u003e- Cox: Cox regression model\u003c/p\u003e\n\u003cp\u003e- CTA: Computed Tomography Angiography\u003c/p\u003e\n\u003cp\u003e- CPB: Cardiopulmonary Bypass\u003c/p\u003e\n\u003cp\u003e- ACC: Aortic Cross-Clamp\u003c/p\u003e\n\u003cp\u003e- RBC: Red Blood Cell\u003c/p\u003e\n\u003cp\u003e- WBC: White Blood Cell\u003c/p\u003e\n\u003cp\u003e- Hb: Hemoglobin\u003c/p\u003e\n\u003cp\u003e- Plt: Platelet\u003c/p\u003e\n\u003cp\u003e- ANC: Absolute Neutrophil Count\u003c/p\u003e\n\u003cp\u003e- ALC: Absolute Lymphocyte Count\u003c/p\u003e\n\u003cp\u003e- ALB: Albumin\u003c/p\u003e\n\u003cp\u003e- ALT: Alanine Aminotransferase\u003c/p\u003e\n\u003cp\u003e- AST: Aspartate Aminotransferase\u003c/p\u003e\n\u003cp\u003e- BUN: Blood Urea Nitrogen\u003c/p\u003e\n\u003cp\u003e- LDH: Lactate Dehydrogenase\u003c/p\u003e\n\u003cp\u003e- FDP: Fibrin/Fibrinogen Degradation Products\u003c/p\u003e\n\u003cp\u003e- CK-MB: Creatine Kinase-Muscle/Brain\u003c/p\u003e\n\u003cp\u003e- BNP: B-Type Natriuretic Peptide\u003c/p\u003e\n\u003cp\u003e- cTnT: Cardiac Troponin T\u003c/p\u003e\n\u003cp\u003e- SII: Systemic Immune-Inflammation Index\u003c/p\u003e\n\u003cp\u003e- SRBC: Suspended Red Blood Cells\u003c/p\u003e\n\u003cp\u003e- POHS: Postoperative Hospital Stay\u003c/p\u003e\n\u003cp\u003e- ICUHS: Intensive Care Unit Hospital Stay\u003c/p\u003e\n\u003cp\u003e- AUC: Area Under the Curve\u003c/p\u003e\n\u003cp\u003e- AP: Average Precision\u003c/p\u003e\n\u003cp\u003e- CI: Confidence Interval\u003c/p\u003e\n\u003cp\u003e- PR: Precision-Recall\u003c/p\u003e\n\u003cp\u003e- DCA: Decision Curve Analysis\u003c/p\u003e\n\u003cp\u003e- NT-proBNP: N-terminal pro-B-type Natriuretic Peptide\u003c/p\u003e\n\u003cp\u003e- NLR: Neutrophil to Lymphocyte Ratio\u003c/p\u003e\n\u003cp\u003e- PLR: Platelet to Lymphocyte Ratio\u003c/p\u003e\n\u003cp\u003e- ECG: Electrocardiogram\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthors’ contribution\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eHC was responsible for analyzing data, drafting manuscripts. YS was responsible for making critical revisions and data collection.XYL, CYL and HYR were responsible for patient follow-up and related clinicopathological data collection. HMS and CZ were responsible for technical support for data analysis, article grammar proofreading. QCW were responsible for the conception, design, and review of selected topics. This manuscript was read and approved by all credited authors. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by grants from National Natural Science Foundation of China (82270506), and Project of innovation team for Graduate Teaching (CYYY-YJSJXCX-202318).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgement\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNone.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets used in this study are available on request from the corresponding author.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was approved by the ethics committee of Chongqing Medical University (2024-583-01) and followed the ethical standards of the Helsinki Declaration. Informed consent was obtained from all the participants.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent to participate.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll patients provided written informed consent.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors have declared that no competing interest exists.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eMoriyama S, Hara M, Hirota T, Nakata K, Doi H, Matsumura T, et al. Population-Based Study of the Incidence and Mortality Rate of Acute Aortic Dissection. Circulation Journal. 2023;88:297\u0026ndash;306. \u003c/li\u003e\n\u003cli\u003eHIRST AE, JOHNS VJ, KIME SW. DISSECTING ANEURYSM OF THE AORTA. Medicine. 1958;37:217. \u003c/li\u003e\n\u003cli\u003eZhu Y, Lingala B, Baiocchi M, Tao JJ, Toro Arana V, Khoo JW, et al. Type A Aortic Dissection\u0026mdash;Experience Over 5 Decades: JACC Historical Breakthroughs in Perspective. Journal of the American College of Cardiology [Internet]. 2020 [cited 2022 Oct 21];76:1703\u0026ndash;13. Available from: https://www.sciencedirect.com/science/article/pii/S0735109720362409\u003c/li\u003e\n\u003cli\u003eJiang F, Wang X, Carmichael M, Chen Y, Huang R, Xiao Y, et al. Contemporary comparative surgical outcomes of type A aortic dissection in US and China: an analysis of the national inpatient sample database and a Chinese multi-institutional registry. Journal of Cardiothoracic Surgery. 2024;19. \u003c/li\u003e\n\u003cli\u003eNazir S, Ariss RW, Minhas AMK, Issa R, Michos ED, Birnbaum Y, et al. Demographic and Regional Trends of Mortality in Patients With Aortic Dissection in the United States, 1999 to 2019. Journal of the American Heart Association. 2022;11. \u003c/li\u003e\n\u003cli\u003eLiu H, Zhang S, Zhang C, Gao Q, Liu Y, Liao F, et al. Risk factors for prolonged postoperative ICU stay in the patients with Stanford type A aortic dissection. Journal of Cardiothoracic Surgery [Internet]. 2024 [cited 2024 Apr 6];19. Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/pmid/38310273/\u003c/li\u003e\n\u003cli\u003eTang Z, Liu H, Shao Y. Efficacy of CRP in combination with D-dimer in predicting adverse postoperative outcomes of patients with acute Stanford type A aortic dissection. Journal of Cardiothoracic Surgery. 2022;17. \u003c/li\u003e\n\u003cli\u003eLi Z, Zhang H, Sulaiman Baraghtha, Mu J, Yusanjan Matniyaz, Jiang X, et al. Short- and Mid-Term Survival Prediction in Patients with Acute Type A Aortic Dissection Undergoing Surgical Repair: Based on the Systemic Immune-Inflammation Index. Journal of Inflammation Research. 2022;Volume 15:5785\u0026ndash;99. \u003c/li\u003e\n\u003cli\u003eXu H, Li Y, Wang H, Yuan Y, Chen D, Sun Y, et al. Systemic immune‐inflammation index predicted short‐term outcomes in ATAD patients undergoing surgery. Journal of Cardiac Surgery. 2022;37:969\u0026ndash;75. \u003c/li\u003e\n\u003cli\u003eSilva GFS, Fagundes TP, Teixeira BC, Chiavegatto Filho ADP. Machine Learning for Hypertension Prediction: a Systematic Review. Current Hypertension Reports. 2022;24:523\u0026ndash;33. \u003c/li\u003e\n\u003cli\u003eBlack J, Kueper JK, Williamson T. An introduction to machine learning for classification and prediction. Family Practice [Internet]. 2022 [cited 2023 Nov 4];40:200\u0026ndash;4. Available from: https://academic.oup.com/fampra/article/40/1/200/6742730?login=false\u003c/li\u003e\n\u003cli\u003eAzmi J, Arif M, Nafis MT, Alam MA, Tanweer S, Wang G. A systematic review on machine learning approaches for cardiovascular disease prediction using medical big data. Medical Engineering \u0026amp; Physics. 2022;105:103825. \u003c/li\u003e\n\u003cli\u003eParashar G, Chaudhary A, Pandey D. Machine Learning for Prediction of Cardiovascular Disease and Respiratory Disease: A Review. SN Computer Science. 2024;5. \u003c/li\u003e\n\u003cli\u003eOu H, Yao Y, He Y. Missing Data Imputation Method Combining Random Forest and Generative Adversarial Imputation Network. Sensors. 2024;24:1112\u0026ndash;2. \u003c/li\u003e\n\u003cli\u003eDablain D, Krawczyk B, Chawla NV. DeepSMOTE: Fusing Deep Learning and SMOTE for Imbalanced Data. IEEE Transactions on Neural Networks and Learning Systems. 2022;1\u0026ndash;15. \u003c/li\u003e\n\u003cli\u003eLiu K, Chen K, Jia K. Convolutional Fine-Grained Classification With Self-Supervised Target Relation Regularization. IEEE Transactions on Image Processing. 2022;31:5570\u0026ndash;84. \u003c/li\u003e\n\u003cli\u003eCao Y, Bass GA, Ahl R, Pourlotfi A, Geijer H, Montgomery S, et al. The statistical importance of P-POSSUM scores for predicting mortality after emergency laparotomy in geriatric patients. BMC Medical Informatics and Decision Making. 2020;20. \u003c/li\u003e\n\u003cli\u003eGoodswen SJ, Barratt JLN, Kennedy PJ, Kaufer A, Calarco L, Ellis JT. Machine learning and applications in microbiology. FEMS microbiology reviews [Internet]. 2021 [cited 2022 Jan 14];45:fuab015. Available from: https://pubmed.ncbi.nlm.nih.gov/33724378/\u003c/li\u003e\n\u003cli\u003eSylvester S, Sagehorn M, Gruber T, Atzmueller M, Sch\u0026ouml;ne B. SHAP value-based ERP analysis (SHERPA): Increasing the sensitivity of EEG signals with explainable AI methods. Behavior Research Methods. 2024; \u003c/li\u003e\n\u003cli\u003eMeng M, Zhao C. Application of Support Vector Machines to a Small-Sample Prediction. Advances in petroleum exploration and development. 2015;10:72\u0026ndash;5. \u003c/li\u003e\n\u003cli\u003eChen J, Yin Y, Han L, Zhao F. Optimization Approaches for Parameters of SVM. Lecture notes in electrical engineering. 2019;575\u0026ndash;83. \u003c/li\u003e\n\u003cli\u003eShehab M, Abualigah L, Shambour Q, Abu-Hashem MA, Shambour MKY, Alsalibi AI, et al. Machine learning in medical applications: A review of state-of-the-art methods. Computers in Biology and Medicine. 2022;145:105458. \u003c/li\u003e\n\u003cli\u003eGuyon I, Weston J, Barnhill S, Vapnik V. Gene Selection for Cancer Classification Using Support Vector Machines. Machine Learning. 2002;46:389\u0026ndash;422. \u003c/li\u003e\n\u003cli\u003eUchikov P, Khalid U, Harris D-SG, Ghale D, Rajadurai H, Kraeva M, et al. Artificial Intelligence in Breast Cancer Diagnosis and Treatment: Advances in Imaging, Pathology, and Personalized Care. Life [Internet]. 2024;14:1451. Available from: https://www.mdpi.com/2075-1729/14/11/1451\u003c/li\u003e\n\u003cli\u003eZhang R, Chen S, Zhang H, Wang W, Xing J, Wang Y, et al. Biomarkers Investigation for In-Hospital Death in Patients With Stanford Type A Acute Aortic Dissection. International Heart Journal. 2016;57:622\u0026ndash;6. \u003c/li\u003e\n\u003cli\u003eLiu J, Sun L-L, Wang J, Ji G. The relationship between fibrinogen and in-hospital mortality in patients with type A acute aortic dissection. The American Journal of Emergency Medicine. 2017;36:741\u0026ndash;4. \u003c/li\u003e\n\u003cli\u003eOz K, Taner Iyigun, Karaman Z, \u0026Ouml;mer \u0026Ccedil;elik, Ertan Akbay, Okan Akınc, et al. Prognostic Value of Neutrophil to Lymphocyte Ratio and Risk Factors for Mortality in Patients with Stanford Type A Aortic Dissection. The heart surgery forum/The Heart surgery forum. 2017;20:119\u0026ndash;9. \u003c/li\u003e\n\u003cli\u003eLiu J, Sun L-L, Wang J, Ji G. Blood urea nitrogen in the prediction of in-hospital mortality of patients with acute aortic dissection. Cardiology Journal. 2018;25:371\u0026ndash;6. \u003c/li\u003e\n\u003cli\u003eBedel C, Fatih Selvi. Association of Platelet to Lymphocyte and Neutrophil to Lymphocyte Ratios with In-Hospital Mortality in Patients with Type A Acute Aortic Dissection. Brazilian Journal of Cardiovascular Surgery. 2019;34. \u003c/li\u003e\n\u003cli\u003eGuo T, Zhou X, Zhu A, Peng W, Zhong Y, Chai X. The Role of Serum Tenascin-C in Predicting In-Hospital Death in Acute Aortic Dissection. International Heart Journal. 2019;60:919\u0026ndash;23. \u003c/li\u003e\n\u003cli\u003eZhang Y, Xu X, Lu Y, Guo L, Ma L. Preoperative uric acid predicts in-hospital death in patients with acute type a aortic dissection. Journal of Cardiothoracic Surgery. 2020;15. \u003c/li\u003e\n\u003cli\u003eMaalouf M. Logistic regression in data analysis: an overview. International Journal of Data Analysis Techniques and Strategies. 2011;3:281. \u003c/li\u003e\n\u003cli\u003ePoullis M. 080 * LIMITATIONS OF COX REGRESSION FOR SURVIVAL ANALYSIS IN THORACIC SURGERY. Interactive CardioVascular and Thoracic Surgery. 2013;17:S88\u0026ndash;8. \u003c/li\u003e\n\u003cli\u003eGuo T, Fang Z, Yang G, Zhou Y, Ding N, Peng W, et al. Machine Learning Models for Predicting In-Hospital Mortality in Acute Aortic Dissection Patients. Frontiers in Cardiovascular Medicine. 2021;8. \u003c/li\u003e\n\u003cli\u003eWu Z, Li Y, Xu Z, Liu H, Liu K, Qiu P, et al. Prediction of preoperative in-hospital mortality rate in patients with acute aortic dissection by machine learning: a two-centre, retrospective cohort study. BMJ open. 2023;13:e066782\u0026ndash;2. \u003c/li\u003e\n\u003cli\u003eXie L, Xie Y, Wu Q, He J, Lin X, Qiu Z, et al. A predictive model for postoperative adverse outcomes following surgical treatment of acute type A aortic dissection based on machine learning. Journal of Clinical Hypertension. 2024;26:251\u0026ndash;61. \u003c/li\u003e\n\u003cli\u003eZhang J, Xiong W, Yang J, Sang Y, Zhen H, Tan C, et al. Enhanced machine learning models for predicting one-year mortality in individuals suffering from type A aortic dissection. Journal of Thoracic and Cardiovascular Surgery. 2024; \u003c/li\u003e\n\u003cli\u003eLiu H, Zhang S, Zhang C, Gao Q, Liu Y, Liao F, et al. Risk factors for prolonged postoperative ICU stay in the patients with Stanford type A aortic dissection. Journal of Cardiothoracic Surgery [Internet]. 2024;19. Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/pmid/38310273/\u003c/li\u003e\n\u003cli\u003eZhu X, Wang J, Chong H, Jiang Y, Fan F, Pan J, et al. Analysis of risk factors for postoperative mortality in acute type A aortic dissection patients under different critical levels. Scientific Reports [Internet]. 2023;13:8107. Available from: https://www.nature.com/articles/s41598-023-35351-w\u003c/li\u003e\n\u003cli\u003eYao R, Yan D, Fu X, Deng Y, Xie X, Li N. The effects of plasma to red blood cells transfusion ratio on in-hospital mortality in patients with acute type A aortic dissection. Frontiers in Cardiovascular Medicine. 2023;10. \u003c/li\u003e\n\u003cli\u003eMagouliotis DE, Rad AA, Viviano A, Oo AY, Xanthopoulos A, Serge Sicouri, et al. Hemostatic Properties of Aortic Root Preservation versus Root Replacement for Acute Type A Aortic Dissection: A Pooled Analysis. Life. 2024;14:1255\u0026ndash;5. \u003c/li\u003e\n\u003cli\u003eMazzolai L, Teixido-Tura G, Lanzi S, Boc V, Bossone E, Brodmann M, et al. 2024 ESC Guidelines for the management of peripheral arterial and aortic diseases. European Heart Journal. 2024; \u003c/li\u003e\n\u003cli\u003eSalerno S, Li Y. High-Dimensional Survival Analysis: Methods and Applications. Annual Review of Statistics and Its Application. 2022;10. \u003c/li\u003e\n\u003cli\u003eSpooner A, Chen E, Sowmya A, Sachdev P, Kochan NA, Trollor J, et al. A comparison of machine learning methods for survival analysis of high-dimensional clinical data for dementia prediction. Scientific Reports. 2020;10. \u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"european-journal-of-medical-research","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ejmr","sideBox":"Learn more about [European Journal of Medical Research](http://eurjmedres.biomedcentral.com)","snPcode":"40001","submissionUrl":"https://submission.nature.com/new-submission/40001/3","title":"European Journal of Medical Research","twitterHandle":"@BioMedCentral","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"type A aortic dissection, machine learning, long-term survival, predictive model, Support Vector Machine (SVM)","lastPublishedDoi":"10.21203/rs.3.rs-5786813/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5786813/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eObjective\u003c/h2\u003e \u003cp\u003eThis study aims to develop a reliable and interpretable predictive model for the risk of long-term survival in type A aortic dissection (TAAD) patients, utilizing machine learning (ML) algorithms.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003e We retrospectively reviewed the clinical data diagnosed with Type A Aortic Dissection (TAAD) who underwent open surgical repair at our institution between September 2017 and December 2020. Cases with less than 20% missing data were imputed using random forest algorithms. To identify significant prognostic factors, we performed LASSO (Least Absolute Shrinkage and Selection Operator) Cox regression analysis, including preoperative blood markers, previous medical history and perioperative condition. Based on the advantages of the model and the characteristics of the dataset, we subsequently developed a machine learning-based prognostic model using Support Vector Machine (SVM) and evaluated its performance across key metrics. To further explain the decision-making process of the SVM model, we employed SHapley Additive exPlanation (SHAP) values for model interpretation.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eA total of 175 patients with TAAD were included in the study. Through LASSO Cox regression, univariate analysis, and clinical relevance assessment, eight feature variables were selected for modeling. Performance evaluation revealed that the SVM model showed excellent performance in both the training and test sets, with no significant overfitting, indicating strong clinical applicability. In the training set, the model achieved an AUC of 0.9247 (95% CI: 0.9200\u0026ndash;0.9279), and in the testing set, 0.8800 (95% CI: 0.8492\u0026ndash;0.9396). The accuracy was 0.8663 and 0.8857, precision was 0.8627 and 1.0000, recall was 0.8713 and 0.7333, F1 score was 0.8670 and 0.8462, Brier score was 0.1068 and 0.1070, average precision (AP) was 0.9266 and 0.9086, and C-index was 0.8901 and 0.8700, respectively. SHAP analysis identified that longer ICU hospital stay, abdominal pain, plasma transfusion volume, creatinine, white blood cell count, operation time, and systemic immune-inflammation index (SII) had significant positive impact on the model's predictions.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eThis study developed an interpretable predictive model based on the SVM algorithm to assess long-term survival in TAAD patients. The model demonstrated accuracy, precision, and robustness in identifying high-risk patients, providing clinicians with reliable evidence for prognosis management.\u003c/p\u003e","manuscriptTitle":"Interpretable Prognostic Modeling for Long-Term Survival of Type A Aortic Dissection Patients Using Support Vector Machine Algorithm","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-01-14 15:57:47","doi":"10.21203/rs.3.rs-5786813/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-01-29T11:27:35+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-01-20T12:48:07+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-01-18T14:03:29+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"134031705783985001733632368951741753294","date":"2025-01-15T17:41:28+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-01-15T11:16:46+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"112934400972140796498900222184680555954","date":"2025-01-15T11:16:06+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"204405833057802332333288742342329436975","date":"2025-01-15T10:35:02+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-01-15T10:29:59+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-01-09T17:47:11+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-01-09T15:44:30+00:00","index":"","fulltext":""},{"type":"submitted","content":"European Journal of Medical Research","date":"2025-01-08T07:39:50+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"european-journal-of-medical-research","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ejmr","sideBox":"Learn more about [European Journal of Medical Research](http://eurjmedres.biomedcentral.com)","snPcode":"40001","submissionUrl":"https://submission.nature.com/new-submission/40001/3","title":"European Journal of Medical Research","twitterHandle":"@BioMedCentral","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"027b8be6-8b70-4f97-85b0-424bbbc7e27a","owner":[],"postedDate":"January 14th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-04-21T16:03:49+00:00","versionOfRecord":{"articleIdentity":"rs-5786813","link":"https://doi.org/10.1186/s40001-025-02510-w","journal":{"identity":"european-journal-of-medical-research","isVorOnly":false,"title":"European Journal of Medical Research"},"publishedOn":"2025-04-15 15:57:37","publishedOnDateReadable":"April 15th, 2025"},"versionCreatedAt":"2025-01-14 15:57:47","video":"","vorDoi":"10.1186/s40001-025-02510-w","vorDoiUrl":"https://doi.org/10.1186/s40001-025-02510-w","workflowStages":[]},"version":"v1","identity":"rs-5786813","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5786813","identity":"rs-5786813","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}