Introduction
Postoperative pneumonia (POP) represents a notably common type of pulmonary complications in patients undergoing CABG, Reference Tafelmeier, Luft and Zistler1 being associated with nearly a 9-fold increase in mortality and an additional 2-week stay in the hospital. Reference Ailawadi, Chang and O’Gara2 The current research landscape is dominated by studies assessing the clinical consequences of POP following cardiac surgeries, with significant disparities among them. Furthermore, epidemiological data on the microbiological etiology of POP are limited. Although certain studies have attempted to analyze risk factors and develop models for prediction, the risk factors for POP following cardiac surgeries may exhibit substantial regional variations. Reference Wang, Huang and Wang3–Reference Strobel, Liang and Zhang5 Additionally, the heterogeneity of the populations studied further illustrates the diversity. To enhance the understanding and management of POP following CABG, we initiated the subsequent study. This study focused on the epidemiology and perioperative contributing factors for POP post-isolated CABG, aiming to understand the baseline data within our hospital and construct a prediction model. The prediction model we have constructed is designed to facilitate the risk stratification management of POP in CABG patients, enabling the implementation of early preventive interventions for high-risk patients. This strategy might demonstrate significant advantages in optimizing the allocation of medical resources, controlling treatment costs, and improving patient outcomes.
Methods
Study design and data collection
The study employed a retrospective case-control methodology. Adult patients (aged ≥ 18 years) who underwent CABG at our hospital, including seven cardiovascular surgery wards across three campuses, between January 1, 2020 and November 30, 2023, were selected. The specific inclusion and exclusion criteria are detailed in Figure 1. Patients were categorized into two groups based on the occurrence of POP: the POP group and the non-POP group.
Data collection flow chart.
Data were systematically collated from the real-time infection surveillance network over the period from December 1, 2023, to July 31, 2024. The data encompassed an extensive range of demographic and clinical parameters, surgical profiles, and diagnostic data pertaining to POP. The selection of variables was based on an extensive literature review in relevant fields and the consensus of domestic experts, while also considering the data availability. Regarding the collection of etiological data, repeated culture results from the same infection site of the same patient were counted as one strain. For data with discrepancies, including missing values and outliers, a manual verification process was employed. Supplemental data were obtained through queries of the hospital’s information system and integrated manually by patient code. All patient data were anonymized during the collection and processing stages. This study protocol was approved by the Research and Clinical Ethics Committee of the First Affiliated Hospital of Zhengzhou University (Ethical review number:2023-KY-1086-001).
Definitions and diagnostic criteria
In the absence of international consensus on a clinical definition of POP, we defined “POP” using the consensus definitions of pneumonia developed by domestic specialists, Reference Zong6 which aligned with the majority of literature reports. Reference Hokari, Ohshima and Nakayama7–Reference Gupta, Gupta and Schuller9 Pneumonia cases that arose within the 30-day post-surgical timeframe were considered to be POP.
All cases of POP were required to meet the criteria for clinical diagnosis of pneumonia. For this, we referred to the WS 382-2012 standards, 10 which were adopted as the diagnostic criteria for POP in domestic consensus and remain consistent with the criteria of “clinically defined pneumonia” updated by the National Healthcare Safety Network (NHSN) of the United States in January 2025. 11 Although the WS 312-2012 standards have been officially superseded in China, the criteria have provided a unified framework for pneumonia diagnosis in China and continue to hold significant utility and reference value in clinical practice owing to their alignment with international guidelines and the absence of an updated domestic standard. This definition included a combination of pulmonary signs and symptoms, systemic findings (i.e., fever, leukocytosis or leukopenia, or altered mental status without other obvious cause in those ≥ 70 years of age), and an abnormal chest x-ray (Supplementary Table 1). Among those who met the criteria for clinically defined pneumonia, domestic hospital infection diagnosis standards 12 were used to assign a microbiological etiology to the case if one or more criteria for “pneumonia with specific laboratory findings” were met (Supplemental Table 1). Those criteria closely resemble the criteria of “Pneumonia with Common Bacterial and Specific Laboratory Findings” outlined in the U.S. NHSN surveillance standards; 11 however, the former provided more objective requirements for microbial colony counts
Dealing with missing values and outliers
Given that the missing data exhibited a pattern of random absence, we considered these missing values to be negligible and treated them as blank entries. The identified outliers were removed directly.
Statistical analysis
Data were analyzed via SPSS version 22.0. Normally distributed continuous data were expressed as mean ± standard deviation (
$\overline{\rm x}$
± s), and differences among groups were compared by independent samples t-test. Non-normally distributed continuous data were presented as median (P25–P75), and compared among groups using the Mann-Whitney U test. Categorical data were described as frequencies and percentages, with group comparisons performed using the χ2 test. Univariate analysis was conducted to filter potential factors that contributed to POP, and those with a P-value < 0.05 were subjected to multivariate logistic regression for identifying independent factors. The model’s calibration was assessed with the Hosmer-Lemeshow test. Use R software version 4.4.1 (R Foundation for Statistical Computing, Vienna, Austria) to perform RCS analysis with 5 nodes to explore the potential nonlinear relationships between continuous variables and POP. Receiver operating characteristic (ROC) curve was drawn to evaluate the prediction model. A P-value of < 0.05 was considered statistically significant.
Results
Incidence
A total of 5,360 patients were identified, with 791 patients excluded, resulting in 4,569 patients included in the study. Among them, 518 developed POP, including 381 with laboratory-confirmed pneumonia and 137 with clinically defined pneumonia, yielding an incidence of 11.34% (518/4569). The incidence of POP among males was 11.69% (379/3243), compared to 10.48% (139/1326) among females, with no statistically significant difference (P = 0.244). The incidences of POP in 2020-2021 and 2022-2023 were 9.13% (187/2048) and 13.13% (331/2521), respectively, showing a statistically significant difference (P < 0.001). The incidences of POP were 12.26% (163/1330) in the first quarter (Q1, January to March), 11.06% (139/1257) in Q2 (April to June), 9.70% (103/1062) in Q3 (July to September), and 12.28% (113/920) in Q4 (October to December), with no statistically significant differences(all P > 0.05).
Distribution of pathogens
A total of 416 pathogenic strains were isolated from 381 patients with positive pathogen cultures. Among these, 360 strains (86.5%) were Gram-negative bacteria and 56 strains (13.5%) were Gram-positive bacteria. The top five pathogens were Klebsiella pneumoniae (31.0%, 129), Pseudomonas aeruginosa (21.4%, 89), Acinetobacter baumannii (16.8%, 70), Staphylococcus aureus (9.6%, 40), and Stenotrophomonas maltophilia (5.3%, 22). Compared with the period of 2020-2021, the proportion of Enterobacteriaceae declined, while that of non-fermentative bacteria rose in 2022-2023. Notably, the increase in P.aeruginosa and S.maltophilia was significant (both P < 0.05) (Supplemental Table 2). From Q1 to Q4, there was a gradual decline in the proportion of G-positive bacteria, while that of G-negative bacteria and non-fermentative bacteria gradually increased. Compared with Q1 and Q2, the proportions of G-negative and G-positive bacteria in Q4 were significantly different (both P < 0.05). Staphylococcus aureus exhibited a significant decline in Q4 compared with Q1 (P < 0.05) (Supplemental Table 3).
Univariate and multivariate analysis
Thirty-four potentially clinically significant indicators were incorporated into the univariate analysis. The results indicated 20 variables associated with the occurrence of POP (Table 1). Due to poor data fit (indicated by the Hosmer-Lemeshow test, P < 0.1), 8 variables were excluded, including red blood cell (RBC) distribution width, blood glucose, serum albumin, brain natriuretic peptide, serum creatinine, D-dimer, pulmonary artery pressure and intraoperative blood loss. In the multivariate analysis, which included the remaining 12 variables, age, diabetes mellitus, smoking, operative duration, RBC transfusion, and duration of mechanical ventilation were shown to increase the risk of POP, whereas higher levels of hemoglobin and total cholesterol were linked to a reduced risk (all P < 0.05) (Table 2).
Univariate analysis of factors contributing to POP in patients undergoing CABG
Multivariate analysis of factors contributing to POP in patients undergoing CABG
RCS analysis
RCS was utilized to assess the nonlinear relationships of six continuous variables with the risk of POP, as identified through multivariate analysis. The analysis established a linear positive correlation between age, RBC transfusion, and duration of mechanical ventilation with POP (Figure 2.a, b, d). Conversely, hemoglobin level was negatively correlated with POP (Figure 2.e,f). Operative duration exhibited a nonlinear relationship with POP (Figure 2. c), whereas total cholesterol levels showed no significant association with the occurrence of POP.
Restricted cubic spline of the nonlinear relationships between age (a), red blood cell transfusion (b), operative duration (c), duration of mechanical ventilation (d), hemoglobin (e), and total cholesterol (f) with the risk of postoperative pneumonia. The red vertical lines and the blue shading respectively represent the reference points where HR = 1 and the 95% confidence intervals.
Construction and validation of the prediction model
A prediction model was formulated based on multivariate analysis. The Hosmer-Lemeshow test was applied to assess the model’s calibration, yielding a chi-squared value of 9.891 and a P-value of 0.273, which suggests an adequate fit. An ROC curve was plotted (Figure 3). The maximum cut-off point was identified as the threshold value. The analysis revealed a Youden index of 0.446, with a sensitivity of 70.6% and a specificity of 74.0%. The area under the curve was 0.774, with a 95% confidence interval (CI) of 0.735 to 0.813 (P < 0.001).
Receiver operating characteristic curve analysis for predicting postoperative pneumonia.
Discussion
This study revealed an overall incidence of 11.34% for POP in CABG patients, surpassing the rates documented in the majority of international research, Reference Zukowska, Kaczmarczyk, Listewnik and Zukowski13–Reference Barnett, Liesman and Strobel15 while being lower than those observed in domestic research. Reference Cheng, Chen and Sun16,Reference Wang, Xiong, Shi and Hou17 Discrepancies across different studies could be attributed to : a) varying inclusion criteria for study subjects; b) lack of uniform definitions and diagnostic criteria for POP; c) differences in demographic characteristics of patients; d) potential biases in information monitoring software. This study observed a noticeable increase in the detection of non-fermenting bacteria, particularly P.aeruginosa and S.maltophilia since 2022. The ratio of Gram-negative bacteria to Gram-positive bacteria exhibited a gradual increase across quarters. Our findings underscored the critical need for antibiotics selection tailored to local pathogen profiles to ensure the efficacy of early empirical treatment.
Kilic A et al. Reference Kilic, Ohkuma and Grimm4 found that patients over 65 years of age had a 1.4-fold increased risk of developing POP after cardiac surgery compared to younger patients. Ailawadi et al. Reference Ailawadi, Chang and O’Gara2 reported a positive correlation between age and POP risk following cardiac surgery. Our finding was aligned with these reports, showing an age-related increase in POP risk. Diabetes mellitus was frequently cited as a risk factor for surgical site infections, whereas its role in the onset of POP was less frequently studied. Reference Ma, Liu and Li18 This study confirmed that diabetes mellitus was an independent factor contributing to POP post-CABG. Previous studies revealed that smoking was associated with a significantly heightened risk of pneumonia following cardiac surgeries. Reference Wang, Lu and Sun19,Reference Zhang, Zhang and Li20 However, a small but notable group of studies suggested that this association may not be clinically significant. Reference Ailawadi, Chang and O’Gara2 Our findings indicated that smoking was a key contributing factor for POP post-isolated CABG. RBC transfusions have been identified as a significant risk factor for postoperative infections in CABG patients, with evidence suggesting that even a single unit can elevate infection risk. Reference Crawford, Magruder and Fraser21 Additionally, the volume of transfused RBCs was associated with an increased risk of infection. Reference Raphael, Chae and Feng14,Reference Ming, Liu and Zhang22 Our study corroborated these findings, affirming that any RBC transfusions posed a risk, and further revealed a linear positive correlation with POP. Prolonged operative duration was associated with a higher incidence of POP in cardiac surgery (HR, 1.38; 95% CI, 1.25-1.53; P < .001) Reference Ailawadi, Chang and O’Gara2 . The domestic consensus on the prevention and control of POP identified anesthesia time exceeding 180 minutes as a risk factor. However, threshold values in specialized surgeries have not yet been delineated. This study revealed a significant escalation in POP risk when CABG procedures surpassed 264 minutes. Extended mechanical ventilation postoperatively also elevates the risk of POP. Ailawadi G et al. Reference Ailawadi, Chang and O’Gara2 reported that POP risk was substantially elevated in patients who ventilated for 24-48 hours and more than 48 hours postoperatively, with respective increases of 2.83 times (95% CI, 1.72-4.66; P < .001) and 4.67 times (95% CI, 2.70-8.08; P < .001) compared to those extubated within 24 hours. The study confirmed that the duration of mechanical ventilation positively correlated with POP risk
The role of cholesterol metabolism in immune regulation has garnered attention in recent years, with cholesterol being believed to play a crucial role in the body’s defense against bacterial and viral infections. Reference Heisler, Johnson and Ma23,Reference Dai, Wang and Liao24 Previous literature found that low preoperative total cholesterol levels were associated with higher POP risk in other types of surgery. Reference Morimoto, Nakamura, Atsuko, Nagaie and Shirabe25,Reference Delgado-Rodríguez, Medina-Cuadros, Martínez-Gallego and Sillero-Arenas26 Our study indicated a tendency that individuals with higher levels of total cholesterol appeared to exhibit a reduced likelihood of developing POP. This could potentially offer new insights for the prevention of POP. Preoperative anemia is prevalent among cardiac surgery patients, ranging from 20% to 30%. Reference Corwin, Shander and Speiss27,Reference Bolliger, Erb and Buser28 Fowler AJ et al. Reference Fowler, Ahmad, Phull, Allard, Gillies and Pearse29 had documented that preoperative anemia was associated with increased infection rate (OR 1.93; 99% CI, 1.17-3.18; P = 0.01). Ailawadi G et al. Reference Ailawadi, Chang and O’Gara2 suggested that higher hemoglobin levels could confer a protective effect against POP in patients undergoing cardiac surgery. This study offered strong evidence suggesting that elevated hemoglobin levels were linked to a reduced risk of POP.
Based on these findings, we constructed a prediction model with a C-stastic of 0.774(95%CI: 0.735-0.813). Our model’s predictive accuracy was slightly lower than that of the model developed by Wang et al. Reference Wang, Huang and Wang3 for predicting POP in patients undergoing cardiac surgery (C-statistic:0.80, 95%CI: 0.79-0.82), but it outperformed the models developed by Strobel et al. Reference Strobel, Harrington and Hill30 for CABG patients (C-statistic: 0.74) and Kilic et al. Reference Kilic, Ohkuma and Grimm4 for cardiac surgery patients (C-statistic: 0.72). The model, which is based more on perioperative objective indicators and specifically designed for CABG patients, facilitates a more expedited and thorough risk assessment process for clinicians prior to surgical procedures. Based on this model, we can identify high-risk populations and develop personalized intervention plans targeting modifiable risk factors. For those factors that cannot be changed, we can indirectly reduce the risk of POP by optimizing perioperative care processes.
However, the retrospective nature of the analysis has several limitations. Firstly, the inability to include certain potential risk factors for POP (such as dysphagia, aspiration risk, and sedation) may introduce confounding bias. Secondly, the study was conducted during the COVID-19 pandemic, which may have affected the patient population’s demographics and surgical practices, potentially reducing the study’s generalizability. Thirdly, there is a lack of diagnostic criteria for POP that combines high sensitivity and specificity, which may not accurately reflect its true incidence. In future research, we will endeavor to develop more robust diagnostic criteria for POP and conduct prospective, multi-center studies. By collecting more comprehensive data, we aim to further validate and refine our findings.,as well as explore the effective use of risk prediction model in clinical decision-making, and ultimately reduce the occurence of POP.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/ice.2025.19
Acknowledgment
We acknowledge Zhicheng Huang, the engineer of the real-time infection surveillance network in our hospital, for his assistance in data access.
Financial support
This study was funding by Henan Province Medical Science and Technology Research Plan Joint Construction project (LHGJ20210315) and Henan Province Special Research Project on Traditional Chinese Medicine Science (2024ZY2137).
Competing interests
All authors declare that there are no conflicts of interest.
