---

Individualized Risk Prediction of Hospital-Acquired Infections Following Median Sternotomy Cardiac Surgery: A Retrospective Cohort Study and Nomogram Development

Submitted by Dan Huang

Tags: blood cardiac care clinical diabetes heart failure infection patients perioperative pulmonary review risk risk factors study surgery thyroid training transfusion

Share Article:

---

Dan Huang¹, Moying Qu¹, Xiaoling Zhang¹, Dou Yuan², Linjia Yan³, Lin Wang^1*

¹Department of Cardiology, West China Hospital of Sichuan University, Chengdu, 610000, Sichuan, China.
²Department of Cardiology Surgery, Shangjin Nanfu Hospital, West China Hospital of Sichuan University, Chengdu, 610000, Sichuan, China.
³The Nethersole School of Nursing Faculty of Medicine, The Chinese University of Hong Kong, 999077, Hong Kong, China.

Abstract

Background: This study aims to develop and validate a predictive model for hospital-acquired infections (HAIs) in patients undergoing median sternotomy for cardiac surgery. The goal is to facilitate the early identification of high-risk individuals and inform targeted preventive measures.

Methods: This study conducted a retrospective analysis of clinical data from 1,840 patients (aged 12–84 years) who underwent median sternotomy for cardiac surgery at Chengdu Shangjin Nanfu Hospital between January 1, 2017, and December 31, 2024. The cohort included 793 males and 1,047 females. Patients were randomly divided into a training set (n = 1,288) and a validation set (n = 552) in a 7:3 ratio. Potential risk factors were initially screened using univariate logistic regression (P < 0.05) in the training set. Significant variables were subsequently incorporated into a multivariate logistic regression analysis to identify independent predictors of postoperative HAIs, based on which a nomogram prediction model was constructed.

Results: Among the 1,840 patients, 1,046 developed HAIs, resulting in an incidence rate of 56.85%. The independent risk factors identified for HAIs following median sternotomy included: type of surgery, age, weight, body mass index (BMI), smoking history, diabetes mellitus, hypertension, congestive heart failure, New York Heart Association (NYHA) functional classification, arrhythmia, preoperative serum creatinine level, peak postoperative serum creatinine level, preoperative thyroid-stimulating hormone (TSH) level, fibrinogen (FIB) level, lung disease, history of cerebrovascular accident, coronary artery disease, and total hospitalization cost. A nomogram was developed based on these predictors. Internal validation demonstrated a model-corrected C-index of 0.702. The calibration curve indicated good agreement between the predicted and observed probabilities of postoperative HAIs.

Conclusions: The developed nomogram provides an individualized tool with good discriminative ability, calibration, and clinical utility for predicting HAIs after median sternotomy cardiac surgery. It may assist in early risk stratification and support the implementation of timely, tailored interventions to reduce the incidence of HAIs in this patient population.

Keywords: cardiac surgery, median sternotomy, hospital-acquired infection, predictive model, nomogram, risk stratification

Introduction

Cardiac surgery is a critical therapeutic intervention for end-stage heart disease, with median sternotomy (e.g., for valve replacement) being one of the most commonly employed approaches since its introduction in 1957 (Song et al., 2023). However, due to its highly invasive nature, prolonged cardiopulmonary bypass time, and frequent need for postoperative intensive care unit admission, patients undergoing this surgery face a significantly elevated risk of hospital-acquired infections (HAIs). The occurrence of such infections can lead to extended hospitalization, increased healthcare costs, and compromised surgical outcomes, contributing to heightened morbidity and mortality (O'Brien et al., 2020). Thus, recognizing and mitigating these risks remains essential for optimizing postoperative recovery. Understanding the fundamentals of cardiac assessment provides important context for appreciating the complexity of these surgical patients.

Previous studies have reported variable rates of surgical site infections (SSIs) following median sternotomy. A recent meta-analysis of 81 studies encompassing 467,333 patients estimated the incidence of sternal wound infections (SWIs) at 1.62% (Chen et al., 2024), while earlier research indicated that associated mortality rates could reach up to 16.5% (Perezgrovas-Olaria et al., 2023). Despite the critical nature of cardiac surgery, patients remain at high risk for HAIs postoperatively. Currently, the clinical prevention and control of HAIs in this population primarily rely on generalized infection control measures, lacking individualized early warning systems for high-risk patients (Farber, 2021). While certain risk factors, such as age, diabetes, and prolonged operative time—have been identified, most conclusions stem from single-center investigations with limited sample sizes and reliance on univariate analysis (Squiccimarro et al., 2022).

The reported incidence of postoperative infections following cardiac surgery varies considerably across studies, ranging from 4.2% to 26.2% (Bateman et al., 2016; Massart et al., 2022). A large-scale retrospective study in China indicated that the prevalence of healthcare-associated infections among patients undergoing cardiac and vascular procedures can be as high as 8% (Jiang et al., 2018). Postoperative infections not only prolong hospital stays and increase healthcare costs but may also lead to surgical failure and pose significant threats to patient survival. Therefore, identifying and analyzing risk factors for infection after median sternotomy cardiac surgery is of considerable clinical importance.

Postoperative pneumonia (POP) is the most common infectious complication after cardiac surgery, with reported rates between 2% and 10%, particularly during the first postoperative week (Massart et al., 2022; Zardi et al., 2022). Established risk factors for POP include advanced age, smoking history, and prolonged mechanical ventilation. Other frequently cited risk factors encompass hypoalbuminemia, hypertension, diabetes mellitus, poor cardiac function (NYHA class III–IV), BMI ≥24 kg/m², prior cardiac surgery, cardiopulmonary bypass time exceeding 120 minutes, and perioperative blood transfusion (D.-S. Wang et al., 2021).

A comprehensive review of the literature from national and global sources revealed considerable variation in the incidence of infections after cardiac surgery, with rates ranging from 4.2% to 26.2% (de la Varga-Martínez et al., 2021; Massart et al., 2022). Additionally, a substantial retrospective study conducted in China revealed that the prevalence of healthcare-associated infections (HAIs) among patients undergoing cardiac and vascular surgery reached as high as 8% (Jiang et al., 2018).

A nomogram is a graphical scoring tool that integrates multiple predictive variables to estimate the individualized probability of a clinical outcome. It offers an intuitive interface and has demonstrated favorable predictive performance in various medical fields (Shariat et al., 2009). However, existing literature has yet to effectively synthesize multidimensional clinical indicators into a comprehensive predictive model for HAIs following median sternotomy.

Therefore, developing an early, accurate predictive model to identify high-risk patients for hospital-acquired infections after median sternotomy cardiac surgery holds urgent clinical significance. Such a tool would facilitate proactive infection prevention, optimize resource allocation, and ultimately improve postoperative outcomes and the overall efficacy of cardiovascular surgical care. The broader challenge of recognizing and managing sepsis underscores why early identification of infection risk is critical in surgical populations.

Methods

Study Design and Participants

This retrospective cohort study was approved by the Ethics Committee of Chengdu Shangjin Nanfu Hospital, with a waiver of written informed consent. The clinical data of patients aged 12–84 years who underwent median sternotomy for cardiac surgery at the hospital between January 1, 2017, and December 31, 2024 were reviewed.

Patients were included if they met all of the following criteria: (1) diagnosis of valvular heart disease confirmed by echocardiography; (2) underwent cardiac valve replacement surgery via median sternotomy; and (3) had a hospital stay exceeding 48 hours. Exclusion criteria included: (1) presence of a hospital-acquired infection prior to surgery; (2) diagnosis of a mental disorder; or (3) incomplete or missing key clinical data.

Based on the above criteria, a total of 1,840 patients were enrolled, comprising 793 males and 1,047 females. Patients who developed a hospital-acquired infection during hospitalization were categorized into the infection group, while those who did not were assigned to the non-infection group.

Data Collection

Clinical data were extracted using a structured, investigator-developed case report form designed for surveillance of hospital-acquired infections (HAIs) following median sternotomy cardiac surgery. Data collection was conducted through the hospital's electronic health record system and infection monitoring platform. The following variables were systematically collected:

Preoperative variables: age, sex, marital status, ethnicity, height, weight, body mass index (BMI), smoking history (documented in medical records), diabetes mellitus (defined by typical clinical symptoms plus fasting blood glucose ≥7.0 mmol/L or random blood glucose ≥11.1 mmol/L, or prior physician diagnosis), hypertension (systolic blood pressure > 140 mmHg or diastolic blood pressure > 90 mmHg, or prior diagnosis), hyperlipidemia, chronic kidney disease, lung disease, peripheral vascular disease, history of cerebrovascular accident, coronary artery disease, congestive heart failure, angina classification, cardiac function assessed by the New York Heart Association (NYHA) classification (class III or higher defined as significantly limited daily activity), arrhythmia, prior cardiac surgery, blood type, preoperative thyroid-stimulating hormone (TSH) level, and preoperative coagulation profile.

Intraoperative and postoperative variables: type of surgery, cardiopulmonary bypass duration, use of intra-aortic balloon counter pulsation (IABP), length of intensive care unit (ICU) stay, duration of mechanical ventilation (categorized as < 24 h or ≥24 h), and perioperative blood product transfusion (including red blood cells, fresh frozen plasma, and platelets).

Laboratory data: postoperative complete blood count and coagulation parameters.

Outcome definition: the primary outcome was the occurrence of an HAI between 48 hours after surgery and hospital discharge.

All personnel involved in data collection underwent standardized training prior to the study to ensure consistency and accuracy.

Diagnostic Criteria for Hospital-Acquired Infections

HAIs were diagnosed in accordance with the national Diagnostic Criteria for Hospital Infections (Kouchak & Askarian, 2012). This process involved a comprehensive assessment of imaging findings, laboratory results, microbiological evidence, and clinical signs and symptoms. Each potential case was jointly reviewed and confirmed by the attending clinician, a dedicated hospital infection control practitioner, and the research investigator to ensure diagnostic accuracy. In alignment with the diagnostic guidelines, a peripheral white blood cell count exceeding 10.0×10⁹/L was considered indicative of infection.

Diagnostic Criteria Based on General Clinical Features

Diagnosis of diabetes: Diabetes mellitus was defined according to established clinical guidelines (Committee, 2025), using one or more of the following criteria: (1) Fasting plasma glucose (FPG) ≥ 126 mg/dL (7.0 mmol/L), with fasting defined as no caloric intake for at least 8 hours. (2) Two-hour plasma glucose (2-h PG) ≥ 200 mg/dL (11.1 mmol/L) during a 75-g oral glucose tolerance test (OGTT), performed according to World Health Organization (WHO) standards. (3) Glycated hemoglobin (A1C) ≥ 6.5% (48 mmol/mol), measured in a laboratory using a method certified by the National Glycohemoglobin Standardization Program (NGSP) and standardized to the Diabetes Control and Complications Trial (DCCT) assay. (4) A random plasma glucose ≥ 200 mg/dL (11.1 mmol/L) in a patient presenting with classic symptoms of hyperglycemia or hyperglycemic crisis. A documented prior clinical diagnosis of diabetes by a physician was also accepted. Diabetes as a comorbidity is well recognized in nursing literature, and readers may find additional context in this comprehensive review of diabetes management.

Congestive heart failure (CHF): CHF is a chronic, progressive clinical syndrome characterized by impaired cardiac function, leading to inadequate perfusion to meet the body's metabolic demands. It is commonly associated with fluid accumulation in tissues and organs, particularly the lungs, liver, and lower extremities, resulting in symptoms such as dyspnea, orthopnea, and peripheral edema. If untreated, pulmonary congestion can progress to acute pulmonary edema, significantly compromising respiratory function. CHF often impairs renal sodium and water excretion, contributing to volume overload and systemic congestion, and is predominantly observed in older adults, representing a leading cause of hospitalization in individuals over 65 years of age. The condition may arise from structural or functional cardiac abnormalities, including hypertension, prior myocardial infarction, cardiomyopathy, valvular heart disease, or pericardial pathology, and can occasionally be precipitated by conditions inducing sustained high cardiac output, such as severe hyperthyroidism. Recognized risk factors for CHF include obesity, diabetes mellitus, tobacco use, and excessive consumption of alcohol or illicit substances like cocaine (Mohan et al., 2023). The integration of digital technology in heart failure management represents one approach to improving outcomes for these patients.

NYHA New York Heart Association (NYHA): The NYHA classification system is a widely recognized and standardized clinical tool for assessing the functional capacity and symptom severity of patients with heart failure. First introduced in 1921, it has evolved from a basic activity-based symptom scale to a cornerstone criterion in modern heart failure research and clinical trial design. Consequently, current evidence-based treatment guidelines are closely aligned with a patient's NYHA functional class (Rohde et al., 2023).

The classification comprises four categories, determined by the treating physician based on the degree of physical limitation imposed by cardiac symptoms:

• Class I: No limitation of physical activity; ordinary activity does not cause undue fatigue, palpitation, or dyspnea.
• Class II: Slight limitation of physical activity; comfortable at rest, but ordinary physical activity results in symptoms.
• Class III: Marked limitation of physical activity; comfortable at rest, but less than ordinary activity leads to symptoms.
• Class IV: Symptoms of heart failure are present at rest; any physical activity increases discomfort.

Preoperative Thyroid Function (TSH): Particularly thyroid-stimulating hormone (TSH) levels, plays a crucial role in regulating cellular, tissue, and organ functions throughout the body (Sinha & Yen, 2024). Thyroid hormones (TH) are widely expressed in various tissues, with significant concentrations in the heart and blood vessels. Their influence on the cardiovascular system is complex and multifaceted, potentially affecting endothelial function, vascular tone, myocardial contractility, and lipid metabolism (Razvi et al., 2018).

There is substantial evidence linking thyroid dysfunction to cardiovascular disease (CVD), highlighting the importance of thyroid hormone status as a risk factor in perioperative contexts (Lang et al., 2022). As an essential modulator of systemic metabolism and cardiovascular homeostasis, thyroid hormone status has thus been increasingly recognized as a relevant risk factor in perioperative settings (Yang et al., 2022). Recent studies indicate that both overt and subclinical thyroid dysfunction are associated with elevated postoperative risks, including arrhythmias, myocardial infarction, hemodynamic instability, and prolonged recovery, all of which may extend the duration of mechanical ventilation and intensive care unit stay and consequently increasing HAI risk (Barasch et al., 2024; Kim et al., 2020). Consequently, preoperative assessment of thyroid function, particularly via thyroid-stimulating hormone (TSH) measurement, has been proposed as a valuable predictor of surgical outcomes and a potential target for perioperative risk stratification (Mascarella et al., 2020). Nursing considerations for thyroid emergencies such as thyroid storm further illustrate the clinical significance of thyroid function in acute care settings.

Postoperative pneumonia (POP): POP represents one of the most common infectious complications following cardiac surgery, and is strongly associated with increased morbidity, mortality, and healthcare costs (He et al., 2014). Chronic obstructive pulmonary disease (COPD) is a well-established preoperative risk factor for POP (Wang et al., 2022). A history of smoking has also been frequently reported as a significant predictor, although some studies have presented inconsistent conclusions regarding its impact (Shuker et al., 2016).

Statistical Methods

All statistical analyses were conducted using R software (version 4.5.1) and SPSS (version 26.0). Continuous variables were assessed for normality using appropriate tests. Normally distributed data are presented as mean ± standard deviation (Mean ± SD) and compared between groups using the independent samples t-test. Non-normally distributed data are summarized as median with interquartile range [M (Q1, Q3)] and compared using the Mann–Whitney U test. Categorical variables are expressed as number and percentage [n (%)] and compared using the chi-square test or Fisher's exact test, as appropriate. A two-sided p-value < 0.05 was considered statistically significant for all analyses.

Model Construction and Validation: Using computer-generated random numbers, all 1,840 patients were randomly allocated into a training set (n = 1,288) and a validation set (n = 552) at a 7:3 ratio. The training set was utilized for variable selection, model development, and internal training, while the validation set was reserved exclusively for evaluating the model's generalization ability.

Within the training set, univariate logistic regression analysis was first performed on all candidate predictors. Variables with a p-value < 0.05 in the univariate analysis were considered significantly associated with postoperative hospital-acquired infection and were included in subsequent multivariate analyses (White et al., 2022). All selected variables were entered into a multivariate logistic regression model. Stepwise variable selection was conducted using the forward likelihood ratio method to identify independent risk factors for infection following median sternotomy, with a significance threshold of P < 0.05 for final model inclusion. Based on these factors, an individualized prediction model was constructed.

A nomogram was developed to visualize the model for predicting the risk of postoperative hospital-acquired infection. The discriminative ability of the model was evaluated using receiver operating characteristic (ROC) curves and the area under the curve (AUC), with an AUC between 0.7 and 0.9 indicating good discriminative performance. Calibration curves were used to assess the agreement between predicted probabilities and observed outcomes.

Results

Current Status of Hospital-Acquired Infections

Among the 1,840 patients included in the final analysis, 1,288 (70%) were randomly assigned to the training set and 552 (30%) to the validation set. The incidence of hospital-acquired infections was 57.76% (744/1,288) in the training set and 54.71% (302/552) in the validation set. Detailed baseline and clinical characteristics of the included patients are presented in Table 1.

Table 1. Baseline Characteristics of the Training and Validation Sets

Univariate Logistic Regression

Following univariate screening, more than twenty variables—including sex, type of surgery, and age—were entered into the multivariable logistic regression analysis. Among patients who developed infection, 508 underwent valvular surgery and 236 underwent non-valvular surgery; 338 were male and 406 were female. The specific variables included in the analysis are presented in Table 2.

Table 2. Univariate Analysis of Risk Factors for HAIs (Training Set)

Multivariate Logistic Regression Analysis

As shown in Table 3, the predictive factors that ultimately entered the model included type of surgery, age, smoking, diabetes, lung disease, cerebrovascular accident, congestive heart failure, NYHA functional classification, arrhythmia, preoperative thyroid function and TSH, among others.

Table 3. Multivariate Logistic Regression Analysis of Risk Factors for HAIs

Development of a Nomogram for Individualized Risk Prediction

Based on the final multivariable model, a nomogram was developed to facilitate individualized prediction of the risk of hospital-acquired infection. For each variable included in the nomogram, a corresponding score is assigned based on the patient's clinical characteristics. These scores are summed to obtain a total point value, which is then projected onto the risk axis to estimate the individual probability of postoperative infection. The nomogram is presented in Figure 1.

[Figure 1: Nomogram model for individualized prediction of hospital infection risk after median sternotomy.]

Model Validation Using Receiver Operating Characteristic (ROC) Curves

The discriminative performance of the nomogram was evaluated using ROC analysis. In the training set, the area under the ROC curve (AUC) for predicting hospital-acquired infection was 0.714 (95% CI: 0.686–0.742) (Figure 2). In the validation set, the AUC was 0.702 (95% CI: 0.658–0.745) (Figure 3). These results indicate that the model demonstrates stable and consistent discriminative ability in both cohorts.

[Figure 2: ROC curve of the model in the training set.]

[Figure 3: ROC curve of the model in the validation set.]

Validation of the Nomogram Model

Calibration curves were constructed for the training and validation sets using the bootstrap method with 1,000 resampling iterations. The curves are presented in Figures 4 and 5, respectively.

[Figure 4: Calibration curve of the nomogram in the training set.]

[Figure 5: Calibration curve of the nomogram in the validation set.]

Clinical Application

Decision curve analysis demonstrated favorable clinical utility for the nomogram in both the training and validation sets (Figure 6, Figure 7). Specifically, employing the nomogram with a threshold probability greater than 0.6% for HAIs prediction yielded higher clinical net benefit than the alternative extreme strategies of "diagnosing all patients" or "diagnosing none."

[Figure 6: Decision curve analyses for the HAIs nomogram in the training set.]

[Figure 7: Decision curve analyses for the HAIs nomogram in the validation set.]

Discussion

This single-center retrospective cohort study developed and internally validated a nomogram for predicting hospital-acquired infection (HAI) following median sternotomy cardiac surgery. The model demonstrated good discriminative ability, with an area under the receiver operating characteristic curve (AUC) of 0.714 in the training set and 0.702 in the validation set, and exhibited satisfactory calibration, indicating stable performance across both datasets. Independent predictors identified included surgical type, age, smoking history, diabetes, lung disease, cerebrovascular accident, congestive heart failure (CHF), New York Heart Association (NYHA) functional class, arrhythmia, and preoperative thyroid-stimulating hormone (TSH) level. These findings align with and extend the existing literature by integrating multidimensional clinical indicators into an individualized risk prediction tool applicable to a broad age range and contemporary surgical practice.

The risk model developed in this study aligns with prior studies in cardiac surgery. Factors such as smoking, cardiopulmonary comorbidities, diabetes, and impaired cardiac function have been widely recognized as determinants of infectious complications, particularly POP, which represents the most prevalent HAI subtype following cardiac procedures (Massart et al., 2022; Shuker et al., 2016). Previous predictive models (Wang et al., 2022) in valve surgery and broader cardiac surgery cohorts have similarly identified age, smoking, diabetes, prolonged mechanical ventilation, and blood transfusion as significant risk factors, with reported AUC values around 0.70, which is comparable to the discriminative performance of the present model. Notably, a study (Liu et al., 2024) focusing on elderly cardiac surgery patients also identified smoking, prior myocardial infarction, cardiopulmonary bypass (CPB) use, transfusion, preoperative hospital stay, and prolonged mechanical ventilation as independent risk factors, achieving a C-index of 0.706 with good calibration and favorable decision-curve utility, results that are consistent with the internal validation performed in this analysis. Mechanistically, these factors contribute to immune dysregulation, impaired mucociliary clearance, increased colonization pressure, and reduced host resistance, thereby elevating the risk of HAI.

Several distinct aspects of these findings warrant further discussion. First, CHF and a higher NYHA functional class were identified as strong predictors within this cohort, demonstrating a stepwise increase in infection odds from NYHA class II to IV. This observation aligns with the established pathophysiological link between venous congestion, pulmonary edema, and compromised pulmonary defense mechanisms, which predispose patients to POP and secondary infections (Narayan et al., 2022). Large-scale heart failure registries have indicated that a worse NYHA classification correlates with adverse clinical outcomes, likely reflecting underlying frailty and diminished physiological reserve, thereby amplifying infection risk (Rohde et al., 2023). While previous HAI prediction models in cardiac surgery have noted cardiac dysfunction in univariate analyses (Massart et al., 2022), the present analysis highlights the independent and graded contribution of NYHA class within a mixed population of valve and non-valve surgical patients.

Second, preoperative TSH level emerged as an independent predictor in the final model. Accumulating evidence links thyroid dysfunction, both overt and subclinical, with adverse perioperative outcomes, including arrhythmias, myocardial ischemia, hemodynamic instability, and prolonged recovery, all of which may extend the duration of mechanical ventilation and intensive care unit stay and consequently increasing HAI risk (Barasch et al., 2024; Kim et al., 2020). The reference study did not incorporate thyroid function parameters; the inclusion of TSH in the present model suggests that thyroid function may represent a clinically relevant and modifiable preoperative infection risk stratification.

Conversely, established intraoperative and perioperative risk factors, including CPB time and blood transfusion, were not retained in the final multivariable model, despite previous associations with HAI reported in earlier studies and in the reference study (Massart et al., 2022; Wang et al., 2022). Several explanations may account for this discrepancy. First, surgical type (valve versus non-valve) may have partially served as a surrogate for CPB exposure and procedural complexity within the dataset, thereby attenuating the independent effect of CPB time once procedure category was included in the modeling process (Meeting, 2019). Second, evolving transfusion thresholds, blood conservation strategies, and institution-specific enhanced recovery after surgery (ERAS) protocols during the study period may have reduced the collinearity-adjusted effect size of transfusion (Anastasiadis et al., 2013). Third, the outcome definition encompassed all HAIs occurring from 48 hours after surgery until hospital discharge across a wide age range, which could dilute risk signals primarily driven by early ventilator-associated pneumonia, where the effects of CPB and transfusion are most pronounced (Varughese, 2019). Finally, residual confounding and sample characteristics, such as a lower prevalence of massive transfusion, may influence variable retention in stepwise regression modeling (Roubinian et al., 2017).

The overall incidence of HAI in this cohort (56.85%) was notably higher than the rates typically reported in the literature, which generally range between 4% and 26% for overall HAI and 2% to 10% for POP (Jiang et al., 2018; Wang et al., 2022). Several factors may explain this discrepancy. First, the study specifically focused on median sternotomy cases within a real-world clinical setting employing comprehensive, active infection surveillance, which may increase the detection of clinically suspected infections compared to passive reporting systems. Second, the broad inclusion of infection syndromes and the application of stringent surveillance definitions at the study institution likely captured events that might be overlooked in administrative databases or datasets relying exclusively on microbiological confirmation. Third, the extended surveillance window—from 48 hours postoperatively until discharge, coupled with variable lengths of hospital stay—likely increases cumulative risk exposure, as longer hospitalization elevates colonization pressure and opportunities for nosocomial transmission. Fourth, the cohort encompassed a wide age distribution and included a substantial proportion of high-acuity valve surgery patients with significant comorbidities (e.g., advanced NYHA class, preexisting lung disease, cerebrovascular disease), thereby enriching the population for infection risk. Differences in antimicrobial stewardship, intensive care unit practices, and device utilization across institutions and over time may further contribute to the variability in reported incidence compared to meta-analyses and Western cohorts (Ya et al., 2023).

From a clinical perspective, the nomogram provides a practical tool for bedside estimation of individualized HAI risk using routinely available clinical variables at or shortly after admission (D. Wang et al., 2021). This facilitates early and targeted preventive measures, such as smoking cessation support, glycemic control in diabetic patients, optimization of heart failure and pulmonary status, screening and management of thyroid dysfunction, meticulous perioperative airway care, and tailored postoperative mobilization and weaning protocols (Anter & Fazal, 2021). Resource allocation, including intensified oral care bundles, selective decolonization strategies, and dedicated respiratory therapy, can be prioritized for patients identified as high risk. Given that the discrimination and calibration of the present model are comparable to those reported in the reference study, a similar level of decision-analytic utility is likely achievable in the corresponding clinical setting. This study has several limitations. The retrospective, single-center design introduces potential selection bias and may limit generalizability to other populations and institutions. Although multivariable adjustment was performed, unmeasured confounding, such as colonization status, specific ventilator settings, antibiotic exposure, and nurse-to-patient ratio, cannot be excluded. Infection adjudication, though structured and based on integrated clinical, laboratory, and imaging criteria, may differ from purely microbiology-confirmed endpoints and could contribute to an apparently higher incidence rate (Magill et al., 2014). Furthermore, the model has not been externally validated; future prospective, multicenter studies are warranted to confirm its generalizability. Subsequent research should also consider subtype-specific modeling (e.g., POP versus urinary tract infection versus surgical site infection), inclusion of more granular intraoperative variables (e.g., CPB time, cross-clamp time, transfusion volumes), and incorporation of postoperative care processes to enhance predictive accuracy and clinical applicability.

In summary, the developed nomogram demonstrates consistent performance and identifies several modifiable risk factors, including smoking, diabetes, heart failure status, lung disease management, and thyroid function, that can inform targeted preoperative optimization and postoperative prevention strategies. Future work should focus on external validation of the model, integration of dynamic perioperative variables, application of decision-curve analysis to quantify net clinical benefit, and implementation within digital clinical decision support systems to reduce the burden of HAIs following median sternotomy cardiac surgery.

Conclusion

This single-center retrospective cohort study developed and internally validated a nomogram to predict hospital-acquired infections (HAIs) after median sternotomy cardiac surgery. Using readily available clinical variables, the model showed consistent discrimination (AUC 0.714 training; 0.702 validation) and good calibration. Independent predictors included surgical type, age, BMI/weight, smoking, diabetes, hypertension, congestive heart failure, NYHA class, arrhythmia, lung disease, cerebrovascular disease, coronary artery disease, pre- and postoperative creatinine, preoperative TSH, fibrinogen, and total hospitalization cost. These results align with prior evidence and identify modifiable targets—cardiometabolic control, pulmonary optimization, heart failure management, and thyroid assessment—for preventive strategies and perioperative optimization. The higher HAI incidence likely reflects rigorous surveillance and broad definitions in a high-acuity cohort. Limitations include retrospective design, single-center setting, and absence of external validation. Future work should perform multicenter prospective validation, add granular intra/postoperative variables, conduct decision-curve analysis, and integrate the tool into clinical decision support to reduce HAI burden.

Abbreviations

• HAIs: Hospital-Acquired Infections
• BMI: Body Mass Index
• NYHA: New York Heart Association
• TSH: Thyroid-Stimulating Hormone
• FIB: Fibrinogen
• SSIs: Surgical Site Infections
• SWIs: Sternal Wound Infections
• POP: Postoperative Pneumonia
• IABP: Intra-Aortic Balloon Counter Pulsation
• ICU: Intensive Care Unit
• WHO: World Health Organization
• NGSP: National Glycohemoglobin Standardization Program
• DCCT: Diabetes Control and Complications Trial
• CHF: Congestive Heart Failure
• TH: Thyroid Hormones
• CVD: Cardiovascular Disease
• COPD: Chronic Obstructive Pulmonary Disease
• ROC: Receiver Operating Characteristic
• AUC: Area Under the Curve
• CPB: Cardiopulmonary Bypass

Ethics and Data Availability

Human Ethics and Consent to Participate declarations: Not applicable.

Clinical trial number: Not applicable.

Institutional Review Board: The Ethics Committee of Chengdu Shangjin Nanfu Hospital, Shangjin Nanfu Hospital, West China Hospital of Sichuan University.

Data availability: The data analyzed in this study were obtained from the internal medical records/database of the Shangjin Nanfu Hospital, West China Hospital, Sichuan University. The acquisition and use of these data have been approved by the Ethics Review Board of Shangjin Nanfu Hospital, West China Hospital, Sichuan University. As the data contain patient privacy information and are subject to the hospital's data security management regulations and ethical protocols, the raw data are not publicly available. Reasonable requests for data access may be granted upon approval by the Ethics Committee and data management department of Shangjin Nanfu Hospital, West China Hospital, Sichuan University, in compliance with ethical guidelines.

References

Anastasiadis, K., Antonitsis, P., & Argiriadou, H. (2013). Principles of miniaturized extracorporeal circulation. Springer.

Anter, A., & Fazal, Z. (2021). Medically Optimizing the Cardiac Surgical Patient. In Cardiac Anesthesia: The Basics of Evaluation and Management (pp. 125-134). Springer.

Barasch, E., Gottdiener, J., Buzkova, P., Cappola, A., Shah, S., DeFilippi, C., Gardin, J., & Kizer, J. R. (2024). Association of thyroid dysfunction in individuals ≥ 65 years of age with subclinical cardiac abnormalities. The Journal of Clinical Endocrinology & Metabolism, 109(10), e1847-e1856.

Bateman, R. M., Sharpe, M. D., Jagger, J. E., et al. (2016). 36th International Symposium on Intensive Care and Emergency Medicine: Brussels, Belgium. Critical Care, 20(Suppl 2), 94. https://doi.org/10.1186/s13054-016-1208-6

Chen, D., Zhang, J., Wang, Y., Jiang, W., Xu, Y., Xiong, C., Feng, Z., Han, Y., & Chen, Y. (2024). Risk factors for sternal wound infection after open-heart operations: a systematic review and meta-analysis. International Wound Journal, 21(3), e14457.

Committee, A. D. A. P. P. (2025). 2. Diagnosis and Classification of Diabetes: Standards of Care in Diabetes—2026. Diabetes Care, 49(Suppl 1), S27.

de la Varga-Martínez, O., Gómez-Sánchez, E., Muñoz, M. F., Lorenzo, M., Gómez-Pesquera, E., Poves-Álvarez, R., Tamayo, E., & Heredia-Rodríguez, M. (2021). Impact of nosocomial infections on patient mortality following cardiac surgery. Journal of Clinical Anesthesia, 69, 110104.

Farber, D. (2021). New Approaches to Prevention of Hospital Acquired Infections in ICU, Surgery and Other High Risk Facilities. Lithuanian University of Health Sciences.

He, S., Chen, B., Li, W., Yan, J., Chen, L., Wang, X., & Xiao, Y. (2014). Ventilator-associated pneumonia after cardiac surgery: a meta-analysis and systematic review. The Journal of Thoracic and Cardiovascular Surgery, 148(6), 3148-3155.

Jiang, W.-l., Hu, X.-p., Hu, Z.-p., Tang, Z., Wu, H.-b., Chen, L.-h., Wang, Z.-w., & Jiang, Y.-a. (2018). Morbidity and mortality of nosocomial infection after cardiovascular surgery: a report of 1606 cases. Current Medical Science, 38(2), 329-335.

Kim, H., Kong, S. H., Moon, J. H., Kim, S. Y., Park, K.-H., Kim, J. S., Choh, J. H., Park, Y. J., & Lim, C. (2020). Subclinical hypothyroidism affects the long-term outcomes of patients who undergo coronary artery bypass grafting surgery but not heart valve surgery. Endocrinology and Metabolism, 35(2), 308-318.

Kouchak, F., & Askarian, M. (2012). Nosocomial infections: the definition criteria. Iranian Journal of Medical Sciences, 37(2), 72.

Lang, X., Li, Y., Zhang, D., Zhang, Y., Wu, N., & Zhang, Y. (2022). FT3/FT4 ratio is correlated with all-cause mortality, cardiovascular mortality, and cardiovascular disease risk: NHANES 2007-2012. Frontiers in Endocrinology, 13, 964822.

Liu, J., Li, X., Xie, W., Wang, Y., Xu, Z., Bai, Y.-X., Zhou, Q., & Wu, Q. (2024). Risk factors and short-term outcomes of postoperative pulmonary complications in elderly patients after cardiopulmonary bypass. Clinical Interventions in Aging, 31-39.

Magill, S. S., Edwards, J. R., Bamberg, W., Beldavs, Z. G., Dumyati, G., Kainer, M. A., Lynfield, R., Maloney, M., McAllister-Hollod, L., & Nadle, J. (2014). Multistate point-prevalence survey of health care-associated infections. New England Journal of Medicine, 370(13), 1198-1208.

Mascarella, M. A., Milad, D., Richardson, K., Mlynarek, A., Payne, R. J., Forest, V.-I., Hier, M., Sadeghi, N., & Mayo, N. (2020). Preoperative risk index among patients undergoing thyroid or parathyroid surgery. JAMA Otolaryngology–Head & Neck Surgery, 146(1), 7-12.

Massart, N., Mansour, A., Ross, J. T., Piau, C., Verhoye, J.-P., Tattevin, P., & Nesseler, N. (2022). Mortality due to hospital-acquired infection after cardiac surgery. The Journal of Thoracic and Cardiovascular Surgery, 163(6), 2131-2140.

Meeting, I. A. S. (2019). Cardiac Track Abstracts. Innovations, 14(2_suppl), 11S-46S.

Mohan, I. K., Baba, K. S., Iyyapu, R., Thirumalasetty, S., & Satish, O. S. (2023). Advances in congestive heart failure biomarkers. Advances in Clinical Chemistry, 112, 205-248.

Narayan, M., Meredyth, N., Aveson, V., & Barie, P. S. (2022). Pulmonary Infections. In Management of Chest Trauma: A Practical Guide (pp. 321-344). Springer.

O'Brien, W. J., Gupta, K., & Itani, K. M. (2020). Association of postoperative infection with risk of long-term infection and mortality. JAMA Surgery, 155(1), 61-68.

Perezgrovas-Olaria, R., Audisio, K., Cancelli, G., Rahouma, M., Ibrahim, M., Soletti, G. J., Chadow, D., Demetres, M., Girardi, L. N., & Gaudino, M. (2023). Deep sternal wound infection and mortality in cardiac surgery: a meta-analysis. The Annals of Thoracic Surgery, 115(1), 272-280.

Razvi, S., Jabbar, A., Pingitore, A., Danzi, S., Biondi, B., Klein, I., Peeters, R., Zaman, A., & Iervasi, G. (2018). Thyroid hormones and cardiovascular function and diseases. Journal of the American College of Cardiology, 71(16), 1781-1796.

Rohde, L. E., Zimerman, A., Vaduganathan, M., Claggett, B. L., Packer, M., Desai, A. S., Zile, M., Rouleau, J., Swedberg, K., & Lefkowitz, M. (2023). Associations between New York Heart Association classification, objective measures, and long-term prognosis in mild heart failure: a secondary analysis of the PARADIGM-HF trial. JAMA Cardiology, 8(2), 150-158.

Roubinian, N., Brambilla, D., & Murphy, E. L. (2017). Statistical caution in big data approaches to transfusion medicine research. JAMA Internal Medicine, 177(6), 860-861.

Shariat, S. F., Capitanio, U., Jeldres, C., & Karakiewicz, P. I. (2009). Can nomograms be superior to other prediction tools? BJU International, 103(4), 492-497.

Shuker, N., Bouamar, R., van Schaik, R. H., et al. (2016). A randomized controlled trial comparing the efficacy of CYP3A5 genotype-based with body-weight-based tacrolimus dosing after living donor kidney transplantation. American Journal of Transplantation, 16(7), 2085-2096. https://doi.org/10.1111/ajt.13691

Sinha, R. A., & Yen, P. M. (2024). Metabolic messengers: thyroid hormones. Nature Metabolism, 6(4), 639-650.

Song, Y., Chu, W., Sun, J., Liu, X., Zhu, H., Yu, H., & Shen, C. a. (2023). Review on risk factors, classification, and treatment of sternal wound infection. Journal of Cardiothoracic Surgery, 18(1), 184.

Squiccimarro, E., Stasi, A., Lorusso, R., & Paparella, D. (2022). Narrative review of the systemic inflammatory reaction to cardiac surgery and cardiopulmonary bypass. Artificial Organs, 46(4), 568-577.

Vacante, M., Biondi, A., Basile, F., Ciuni, R., Luca, S., Di Saverio, S., Buscemi, C., Vicari, E. S. D., & Borzì, A. M. (2019). Hypothyroidism as a predictor of surgical outcomes in the elderly. Frontiers in Endocrinology, 10, 258.

Varughese, A. (2019). Evaluation of Perioperative Period in Adult Patients with Excessive Bleeding After Cardiac Surgery on Cardiopulmonary Bypass. Lithuanian University of Health Sciences.

Wang, D., Lu, Y., Sun, M., Huang, X., Du, X., Jiao, Z., Sun, F., & Xie, F. (2022). Pneumonia after cardiovascular surgery: incidence, risk factors and interventions. Frontiers in Cardiovascular Medicine, 9, 911878.

Wang, D.-S., Huang, X.-F., Wang, H.-F., Le, S., & Du, X.-L. (2021). Clinical risk score for postoperative pneumonia following heart valve surgery. Chinese Medical Journal, 134(20), 2447-2456.

White, N. M., Balasubramaniam, T., Nayak, R., & Barnett, A. G. (2022). An observational analysis of the trope "A p-value of < 0.05 was considered statistically significant" and other cut-and-paste statistical methods. PLoS ONE, 17(3), e0264360.

Ya, K. Z., Win, P. T. N., Bielicki, J., Lambiris, M., & Fink, G. (2023). Association between antimicrobial stewardship programs and antibiotic use globally: a systematic review and meta-analysis. JAMA Network Open, 6(2), e2253806.

Yang, L., Sun, X., Zhao, Y., & Tao, H. (2022). Effects of antihypertensive drugs on thyroid function in type 2 diabetes patients with euthyroidism. Frontiers in Pharmacology, 13, 802159.

Zardi, E. M., Chello, M., Zardi, D. M., Barbato, R., Giacinto, O., Mastroianni, C., & Lusini, M. (2022). Nosocomial extracardiac infections after cardiac surgery. Current Infectious Disease Reports, 24(11), 159-171.