Aortic dissection is a critical condition, with Stanford type A aortic dissection (SAAD) representing 60–70% of cases and carrying a particularly high mortality risk. International data show that without surgery, mortality exceeds 50% within 48 hours and reaches 70% within two weeks of onset [1, 2]. Although surgical advances have substantially reduced postoperative mortality, early complications remain a major cause of in-hospital death [3, 4]. Among these, bloodstream infection (BSI) is a serious postoperative complication characterized by an insidious onset and a diagnosis largely dependent on time-consuming blood cultures, which complicates early detection [5]. Without timely intervention, BSI may progress to shock and multiorgan failure, resulting in increased morbidity and mortality.

Given these challenges, identifying risk factors for BSI after SAAD surgery is clinically important for guiding treatment and improving outcomes. This study systematically examined the incidence and risk factors for BSI following SAAD surgery and developed a predictive model to identify high-risk patients. We analyzed clinical data from 257 SAAD patients who underwent conventional open surgery at the Shandong Provincial Hospital Affiliated to Shandong First Medical University between January 2017 and July 2023. Based on comprehensive statistical analyses, a predictive model was established to assess the individual risk of infection and to facilitate targeted preventive strategies. The findings provide valuable scientific insights for postoperative management and prevention of complications in SAAD patients.

Of the 257 patients included in the final analysis, 27 (10.5%) developed postoperative bloodstream infection (BSI). A comprehensive analysis was performed to characterize postoperative BSI, evaluating baseline admission characteristics, preoperative status, and intraoperative and postoperative indicators. Key findings are summarized below.

3.5 Analysis of Risk Factors and Development of a Prediction Model

Variables with significant univariate associations were subsequently entered into a multivariate binary logistic regression analysis using the forward likelihood ratio method. This identified five independent predictors for postoperative bloodstream infection: preoperative C-reactive protein (OR = 1.010, 95% CI: 1.002–1.019, p = 0.020), tracheostomy (OR = 9.186, 95% CI: 2.463–34.266, p = 0.001), infectious pneumonia (OR = 32.872, 95% CI: 4.186–258.174, p = 0.001), circulatory arrest time (min) (OR = 1.048, 95% CI: 1.004–1.093, p = 0.033), and age (OR = 1.055, 95% CI: 1.010–1.103, p = 0.016). These factors were incorporated to establish the final prediction model.

Based on the multivariate logistic regression analysis, a prediction model for postoperative bloodstream infection was established using five significant independent factors: preoperative C-reactive protein, tracheostomy, infectious pneumonia, circulatory arrest time, and age. The Wald test confirmed the statistical significance of all included variables. The model’s goodness-of-fit was evaluated using the Hosmer-Lemeshow test, which showed good calibration with a chi-square of 9.873 (p = 0.274). The likelihood ratio test further supported model significance with p 0.001. Discrimination ability was assessed by ROC analysis, revealing an AUC of 0.897 (95% CI: 0.843–0.950, p 0.001). At the optimal cutoff corresponding to the maximum Youden’s index of 0.896, the model demonstrated sensitivity of 0.850 and specificity of 0.900, indicating strong predictive performance (Fig. 2, Table 5). The model’s discrimination ability was assessed by ROC analysis, revealing an AUC of 0.897 (95% CI: 0.843–0.950, p 0.001). Bootstrap internal validation (1000 repetitions) yielded an optimism-corrected AUC of 0.872 (95% CI: 0.815–0.929), indicating robust internal performance.

Fig. 2.

ROC curves for the model to predict the occurrence of bloodstream infection after admission in the SAAD. After plotting the ROC curve, the area under the (AUC) was 0.897 (95% CI: 0.843–0.950, p 0.001). ROC, receiver operating characteristic; AUC, area under the curve.

Table 5. Multivariate analysis of the included subjects.

	Odds ratio	95% confidence interval	p-value
Age	1.055	1.010, 1.013	0.016
Infectious Pneumonia	32.872	4.186, 258.174	0.001
Tracheostomy	9.186	2.463, 34.266	0.001
Preoperative C-Reactive Protein	1.010	1.004, 1.093	0.020
Circulatory Arrest Time	1.048	1.018, 1.095	0.003

Footnote:

Multivariate analysis showed that age, preoperative C-reactive protein (CRP), time of circulatory arrest, tracheotomy, and infectious pneumonia were independent predictors of BSI.

The above variables were then assigned appropriate scores based on their relative importance in the nomogram (Fig. 3). Finally, we constructed the decision curve analysis (DCA) curve (Fig. 4). These analyses demonstrated that the developed nomogram performed well across all threshold probabilities, confirming that it would provide the highest clinical net benefit across a wide range of clinical scenarios.

Fig. 3.

Prediction chart for all-cause in-hospital mortality of TAAAD. A bar plot was drawn to visualize the results of the prediction model. A total of five prediction variables were involved in the composition of the bar plot. According to the extent of influence of each prediction variable in the model, i.e., the magnitude of the regression coefficient, a score was assigned to each value level of the variable, resulting in five separate scores, and then these scores were added together to obtain the total score. TAAAD, Type A acute aortic dissection.

Fig. 4.

Curve A represents model construction, Curve B represents full intervention, and Curve C represents no intervention. The clinical decision curve for predicting all-cause in-hospital mortality in TAAAD. The DCA curve was plotted based on the constructed CPM. The horizontal axis indicates the risk threshold, and the vertical axis indicates the net benefit (NB) after considering benefits and harms. The DCA results show that the risk assessment model brings net benefit to patients over a risk threshold range of 0.01–0.7.

This retrospective study of 257 patients undergoing surgical repair for Stanford type A aortic dissection provides a comprehensive analysis of the incidence, risk factors, and clinical impact of postoperative bloodstream infection. Our key findings indicate that BSI is a frequent (10.5%) and devastating complication, associated with a significantly prolonged hospital course, a higher incidence of multi-organ failure, and a significant increase in in-hospital mortality (29.63% vs. 2.17%). Furthermore, we developed and internally validated a predictive model incorporating five independent risk factors—age, preoperative C-reactive protein, circulatory arrest time, tracheostomy, and infectious pneumonia—which demonstrated strong discriminatory power (AUC 0.897) for identifying high-risk patients.

The observed BSI incidence of 10.5% aligns with the higher end of rates reported for major cardiac surgery, a finding likely attributable to the extreme acuity and complexity of SAAD surgery [13]. The profound clinical consequences of BSI in our cohort underscore its role as a critical determinant for poor outcomes. Patients with BSI had a more invasive postoperative course, characterized by prolonged dependence on mechanical ventilation and invasive catheters, a greater need for renal replacement therapy, and higher rates of re-operation. The cascade of complications, including cerebral infarction, acute renal insufficiency, and liver failure, suggests that BSI acts as a potent trigger for exacerbation of the systemic inflammatory response syndrome and subsequent multi-organ dysfunction syndrome [14]. The dramatic 13-fold increase in mortality in the BSI group illustrates that BSI is not merely a comorbid condition but a pivotal event that can tip the precarious balance of a post-dissection patient towards a fatal outcome. Therefore, strategies aimed at preventing and aggressively managing BSI are paramount to improving overall survival in SAAD repair.

The five independent risk factors identified in our multivariate analysis provide crucial insights into the pathophysiology of BSI. Elevated preoperative CRP emerged as a significant predictor, highlighting the importance of the pre-existing inflammatory state [15]. SAAD itself is now recognized as a highly inflammatory condition [16], with recent studies elucidating the role of damage-associated molecular patterns released from the dissected aorta in triggering a fulminant innate immune response [17]. A high CRP level upon admission is a biomarker of this intense systemic inflammation, which is often coupled with a state of compensatory immunoparalysis [18]. This early immunodysfunctional state, characterized by impaired neutrophil function and monocyte suppression, renders the patient highly susceptible to secondary postoperative infections [19]. Thus, preoperative CRP serves as a quantifiable measure of a patient’s baseline vulnerability to infection even before the surgical insult [20].

The highly significant associations of tracheostomy and infectious pneumonia with BSI underscore the central role of the respiratory tract as a primary source for bloodstream invasion [21]. Tracheostomy, while sometimes necessary for prolonged ventilation, disrupts the anatomical barrier of the airway, facilitating colonization with hospital-acquired pathogens [22]. Infectious pneumonia, particularly ventilator-associated pneumonia, then establishes a persistent nidus of infection from which bacteria can readily translocate into the pulmonary circulation and cause BSI [23, 24]. The predominance of Gram-negative bacilli in our blood culture analysis is consistent with this pathway and strongly suggests that rigorous adherence to VAP prevention bundles is not only a pulmonary protective strategy but a critical defense against life-threatening BSI [25].

Prolonged circulatory arrest time, a marker of surgical complexity, implicates gut mucosal injury and bacterial translocation as another key pathophysiological pathway [26]. Deep hypothermic circulatory arrest induces splanchnic hypoperfusion and ischemia-reperfusion injury, which can compromise the integrity of the intestinal barrier [27]. This allows commensal Gram-negative bacteria and their endotoxins to leak into the systemic circulation. The role of gut-derived infection in critical illness is a topic of renewed interest, and our data provide strong clinical evidence supporting this pathway in SAAD surgery, linking the technical demands of the procedure directly to the risk of a subsequent septic complication. Furthermore, advanced age, a non-modifiable risk factor, contributes through the well-documented decline in immune competence known as immunosenescence, which reduces the patient’s ability to withstand the cumulative insults of dissection, surgery, and infection [28].

The predictive model developed from these factors offers a practical tool for early risk stratification. Its strength lies in the integration of readily available preoperative and immediate postoperative variables, allowing clinicians to identify high-risk patients at the conclusion of the operation or upon ICU admission. This risk stratification enables a proactive, targeted management approach. For high-risk patients, clinicians can institute intensified surveillance, such as enhanced microbiological monitoring, have a lower threshold for initiating prudent empiric antibiotic therapy guided by local antibiograms, aggressively pursue source control, and consider adjunct supportive measures such as immunonutrition. By focusing resources on the patients who need them most, this model facilitates a paradigm shift from reactive treatment to proactive prevention, potentially reducing the incidence and mortality of BSI.

Several limitations of our study must be acknowledged. Its single-center, retrospective design allows for the potential for selection and information bias. The number of BSI events, while substantial for this patient cohort, is relatively small for a multivariate model with five predictors, raising a potential risk of overfitting. Most critically, the model requires external validation in independent, multi-center cohorts to confirm its generalizability before widespread clinical adoption. Moreover, the number of BSI events, while substantial for this patient cohort, results in a low events-per-variable ratio in our multivariate model, which remains a limitation. Although bootstrap internal validation suggested acceptable optimism, external validation in independent, multi-center cohorts is crucial to confirm generalizability before any clinical application. Furthermore, the predictors identified in our model—advanced age, elevated preoperative inflammation, prolonged circulatory arrest, tracheostomy, and infectious pneumonia—are based on strong biological plausibility regarding immune function, procedural invasiveness, and infection pathways. As such, they are likely to remain relevant risk factors in other settings. However, the magnitude of their association (i.e., their specific odds ratios) may vary across institutions employing different surgical techniques (e.g., strategies to shorten circulatory arrest) or postoperative care bundles (e.g., rigorous ventilator-associated pneumonia prevention protocols that reduce the need for tracheostomy). Consequently, while the core predictors are expected to be generalizable, the model’s absolute risk estimates may require recalibration during external validation in centers with distinct clinical protocols before it can be applied for local risk stratification. Finally, the inclusion of “infectious pneumonia” as a predictor means the model is most accurately applied in the early postoperative period for dynamic risk assessment, rather than as a pure preoperative prediction tool. While the sample size provided sufficient statistical power to identify the strong independent predictors included in our final model, it may have been underpowered to detect weaker associations (e.g., odds ratios 2.0) or to definitively evaluate risk factors with very low baseline prevalence in this surgical population.

In conclusion, this study confirms the critical impact of postoperative BSI following type A aortic dissection surgery. We established a predictive model incorporating five independent risk factors: age, preoperative CRP, circulatory arrest time, tracheostomy, and infectious pneumonia. The model shows strong predictive ability, offering a practical tool for early risk stratification. This enables timely intervention for high-risk patients, potentially improving outcomes. External validation in future studies is recommended to generalize its application.

Data will be made available from the corresponding author upon reasonable request.

Conception and design of the study: ZFX, FYL, FHS, LYW, HZZ and XCM. Acquisition of data: FHS, ZFX, FYL, LYW, MQW, and XZL. Data analysis and interpretation: ZFX, FYL, FHS, LYW, and MQW. Drafting of manuscript: ZFX, FYL, FHS, LYW, MQW, and XZL. Critical revision: ZFX, FYL, FHS, LYW, MQW, XZL, HZZ, and XCM. Final approval and supervision: HZZ and XCM. All authors read and approved the final manuscript. All authors have participated sufficiently in the work and agreed to be accountable for all aspects of the work.

This retrospective study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of Shandong Provincial Hospital Affiliated to Shandong First Medical University (Number: NSFC2018-002). Informed consent was exempted due to the retrospective nature of the study.

Not applicable.

This study was supported by the fund with an additional project number. Add content as follows: This work was supported by the Young Experts of Taishan Scholar Program of Shandong Province (grant no. tsqn202408359), the Nature Science Foundation of Shandong Province (ZR2023MH124), and Jinan Science and Technology Plan Project (202225050).

The authors declare no conflict of interest.