Preoperative hospital length of stay as a modifiable risk factor for mediastinitis after cardiac surgery

The study population comprised of 6653 consecutive adult patients undergoing median sternotomy during the period between September 2000 and September 2009 at a single tertiary care hospital in Montreal, Canada. Heart transplant patients, adult congenital and patients receiving ventricular assist devices were excluded from this cohort.

Preoperative variables that were collected prospectively as part of a clinical registry maintained by the Division of Cardiac Surgery included the following: age, sex, obesity, ejection fraction, type of operation (coronary, valve, or combined procedures), smoking, diabetes mellitus, COPD, chronic renal failure (CRF), preoperative creatinine, preoperative estimated glomerular filtration rate (eGFR), congestive heart failure (CHF), hypertension, redo surgery, left main coronary disease, preoperative hospital length of stay, operative status (elective, urgent, or emergent), and Parsonnet score. Any missing data was obtained by reviewing individual patient charts at the time of the study. This study was approved by the McGill University Health Center Research Ethics Board.

Mediastinitis was defined according to the Center for Disease Control and Prevention criteria[14]; however, all suspected cases were also confirmed by CT scan imaging of the chest. Superficial wound infections sparing the sternum were not classified as mediastinitis. Estimated glomerular filtration rate was derived using the CKD-Epi study equation using the preoperative serum creatinine value closest to the date of surgery[15]. Preoperative hospital length of stay was calculated as the number of days from hospital admission (under any service) until the day of surgery. The calculation of Parsonnet score has been previously described[16]. Obesity was defined as BMI greater than 30 kg/m². Operative status was categorized as: emergent if the patient’s cardiac disease dictated that surgery be performed within hours outside of regular operating hours; urgent if the patient’s cardiac disease required the patient to remain in the hospital until the operation was performed; and elective if the patient’s clinical status allowed discharge from the hospital and readmission at a later date[17].

All patients receive a 2% chlorhexidine gluconate shower and mupirocin 2% ointment to both nostrils the night prior to surgery as well as the morning of. Antibiotic prophylaxis is given as Cefazolin 2 grams intravenously 30 minutes before skin incision and again after 4 hours. Additional doses are given every 4 hours, if required, until termination of the surgery. Patients undergoing valvular surgery, have a penicillin allergy, or who are methicillin-resistant Staphylococcus aureus (MRSA) positive receive Vancomycin 1 gram intravenously 30 minutes before skin incision. Removal of excess hair at the sternum, groin, and appropriate leg is performed with clippers in the operating suite. All patients are disinfected with Baxedine^© (Chlorhexidine 2% w/ 70% V/V alcohol solution) in the operating room prior to incision.

Cardiopulmonary bypass was performed using non-pulsatile flow. Cold blood cardioplegia was used to induce and maintain cardioplegic arrest. Patients are kept at mild hypothermia throughout cardiopulmonary bypass (lowest nasopharyngeal temperature 32°C). The left internal mammary artery, when required as per discretion of the surgeon, was harvested as a pedicle using electrocautery in all cases. Sternal closure was performed using seven or eight No.5 stainless steel wires. In the immediate 24 hr postoperative period, an insulin infusion titrated to a blood glucose defined by the intensive care physician was administered if blood glucose levels exceed 10 mmol/L. The insulin infusion was discontinued once the patient was transferred to the cardiac surgical intermediate care unit.

Patient characteristics are expressed as proportions, means and standard deviations, or medians and interquartile ranges, as appropriate. A natural log transformation was made to Parsonnet score, since its distribution was highly skewed.

Logistic regression was used to examine the association of various demographic, comorbidity, and preoperative laboratory variables with mediastinitis. Models were adjusted for covariates that were significant at the univariate level. The Hosmer Lameshow Goodness-of-fit test was used to assess calibration of the logistic regression model, by assessing the observed and predicted event rates in subgroups (deciles) of the model population. A p-value threshold of 0.05 was used to determine statistical significance. All analyses were carried out using SAS version 9.1 (SAS Institute Inc., Cary, NC).

During the 9-year period from September 2000 to September 2009, 6653 patients underwent cardiac surgery resulting in 108 (1.6%) cases of mediastinitis. Of these, 87 included coronary artery bypass grafting surgery (CABG) with use of a single internal mammary artery, 16 included CABG without any mammary artery, while 5 were isolated valve cases. The overall distribution of operations included isolated CABG at 64.9%, isolated valve in 15.4% of cases, and combined valve and CABG in 14.9% of cases.

Table 1 describes the baseline characteristics of the study population. The mean age of the patients was 66.1 years with the majority having a systolic ejection fraction of more than 50%. The mean Parsonnet score was 18.7 ± 13.3 leading to a mean predicted mortality risk of 7.6%.

Preoperative predictors for mediastinitis are shown in Table 2. Univariate risk factors were advanced age, Parsonnet score, CHF, COPD, diabetes, hypertension, left main disease, peripheral vascular disease, obesity, decreasing eGFR, operative status, and prolonged preoperative length of stay. After adjusting for all significant factors in univariate analyses only obesity, COPD, diabetes, preoperative length of stay, and decreasing eGFR were independent predictors of mediastinitis, with obesity being the most significant (OR 2.59, 95% CI 1.58-4.23). Standard eGFR cut-points of 30 and 60 ml/min/1.73 m² were used in Table 2; however, there was no eGFR cut-off value above which the risk of mediastinitis became significant, since the relationship between both elements remained linear across all eGFR values (data not shown). The area under the receiver operating characteristic curve for our fully adjusted model was 0.734. This implies that our model, which uses easy to collect variables, is acceptable to accurately discriminate those at risk of mediastinitis after cardiac surgery. The Hosmer Lameshow Goodness-of-Fit test also showed good calibration of our model (Chi-square 10.48, degrees of freedom=6, p-value=0.11).

Furthermore, a dose–response relationship was found with regards to preoperative hospital stay, whereby a stay beyond 7 days was associated with a 2.43-fold increased risk of mediastinitis compared to a preoperative stay of 4–7 days which increased this risk by 1.39 times (Table 3).

There were 381 in-hospital deaths. The in-hospital actuarial mortality rate for cardiac surgery cases complicated by mediastinitis was 13.9% compared to 5.6% in those patients without mediastinitis (OR 2.72, 95% CI 1.56 - 4.75).

Management of patients with mediastinitis was either by non-operative or operative means. Twenty-eight patients who had a stable clinical picture were treated by conservative management involving vacuum-assisted closure therapy with intravenous antibiotics. When the overall clinical picture displayed evidence of deterioration due to unremitting infection or lack of improvement secondary to an unstable sternum, operative revision was performed. Sternal debridement and re-wiring was utilized in 16 patients, while sternal debridement with pectoralis muscle flap advancement was performed in 64 cases of mediastinitis.

Of the 108 cases of postoperative mediastinitis, no bacteriology could be found in 4 of the cases after chart review. In the remaining 104 cases, the most common pathogen was coagulase negative Staphylococcus (82%) followed by Enterococcus (22%). Mixed flora was found in 49% of cases of mediastinitis (Table 4). There was no significant relationship between bacterial pathogen and preoperative hospital stay.

There are several limitations to our study. First, it is a retrospective study and as such, the results may be affected by confounding variables not controlled for in the study. In addition, there was no pre-specified follow-up of patients to assess wound infection following surgery, which could allow for patients with delayed mediastinitis to be lost to follow-up. However, as a tertiary care center, the majority of patients presenting at a peripheral hospital with a potential surgical complication are referred back to our institution. Currently, our database does not contain cardiopulmonary bypass and aortic cross clamp times for all patients, thus we could not evaluate its impact on mediastinitis risk. Another limitation is the failure of our database to capture the cause for the prolonged preoperative hospital stay. This is most often multifactorial. As our institution is a referral center, patients often come from afar, and therefore, remain in hospital until they can be operated on. The limited number of ICU beds in a public healthcare system also prevents prompt surgical intervention, thereby, further prolonging the hospital stay. For example, performing an emergency cardiac operation overnight often results in cancellations in the operating schedule the following day. Moreover, it is possible that patients remaining in hospital awaiting surgery did so because of an impaired medical condition (pneumonia, sepsis, CHF, etc.) that was not captured by our analysis. However, our study found no relationship between higher preoperative mortality risk as indicated by the Parsonnet score with the risk of mediastinitis. Lastly, this study was performed at a single institution, which may limit generalizability. However, this same circumstance also ensures that all patients were subjected to the same pre- and postoperative variables that could affect infection.