Abstract
Background
The Sepsis-3 guidelines diagnose sepsis based on organ dysfunction in patients with either proven or suspected infection. The objective of this study was to assess the incidence and outcomes of sepsis diagnosed using these guidelines in patients in a cardiac intensive care unit (CICU) after cardiac surgery.
Methods
Daily sequential organ failure assessment (SOFA) scores were calculated for 2230 consecutive adult cardiac surgery patients between January 2013 and May 2015. Patients with an increase in SOFA score of ≥2 and suspected or proven infection were identified. The length of CICU stay, 30-day mortality and 2-yr survival were compared between groups. Multivariable linear regression, multivariable logistic regression, and Cox proportional hazards regression were used to adjust for possible confounders.
Results
Sepsis with suspected or proven infection was diagnosed in 104 (4.7%) and 107 (4.8%) patients, respectively. After adjustment for confounding variables, sepsis with suspected infection was associated with an increased length of CICU stay of 134.1h (95% confidence interval (CI) 99.0–168.2, P<0.01) and increased 30-day mortality risk (odds ratio 3.7, 95% CI 1.1–10.2, P=0.02). Sepsis with proven infection was associated with an increased length of CICU stay of 266.1h (95% CI 231.6–300.7, P<0.01) and increased 30-day mortality risk (odds ratio 6.6, 95% CI 2.6–15.7, P<0.01).
Conclusions
Approximately half of sepsis diagnoses were based on proven infection and half on suspected infection. Patients diagnosed with sepsis using the Sepsis-3 guidelines have significantly worse outcomes after cardiac surgery. The Sepsis-3 guidelines are a potentially useful tool in the management of sepsis following cardiac surgery.
Keywords: cardiac surgical procedures, sepsis, thoracic surgery
Editor's key points.
-
•
The systemic inflammatory response syndrome (SIRS) after major cardiac surgery can complicate the diagnosis of sepsis
-
•
There are no data on the incidence and consequences of sepsis after cardiac surgery using the Sepsis-3 criteria
-
•
In this single-centre study, the incidence of sepsis diagnosed using the 2016 Sepsis-3 criteria was higher compared with previous reports using SIRS criteria
-
•
This study found that sepsis was associated with significantly increased morbidity and mortality, particularly with documented infection
-
•
Multicentre studies are needed to evaluate the use of the Sepsis-3 criteria after cardiac surgery further
The Sepsis-3 guidelines were introduced in 2016 and define sepsis as organ dysfunction in the presence of proven or suspected infection.1 In the critical care setting, organ dysfunction identified using the Sequential Organ Failure Assessment (SOFA) score (Table 1)2 replaces the systemic inflammatory response syndrome (SIRS) as the means by which the adverse physiological effects of infection are identified.3 According to the new guidelines, suspected or proven infection with organ dysfunction (defined as an increase of ≥2 in SOFA score) results in the diagnosis of sepsis. In previous definitions, suspected infection could only result in a diagnosis of suspected sepsis until infection was proven on microbiological culture.3, 4
Table 1.
The daily sequential organ failure assessment (SOFA) score.2 PaO2, arterial partial pressure of oxygen; FiO2, fraction of inspired oxygen; MAP, mean arterial pressure
| SOFA score | 1 | 2 | 3 | 4 |
|---|---|---|---|---|
| Respiration | ||||
| PaO2/FiO2 (kPa) | <53.3 | <40.0 | <26.7 and mechanical ventilation | <13.3 and mechanical ventilation |
| Coagulation | ||||
| Platelets×103 μl−1 | <150 | <100 | <50 | <20 |
| Liver | ||||
| Bilirubin (μmol litre−1) | 20–32 | 33–101 | 102–204 | >204 |
| Cardiovascular | ||||
| Hypotension | MAP<70 mm Hg | Dopamine ≤5 μg kg−1min−1 or dobutamine (any dose) | Dopamine >5 μg kg−1 min−1 or adrenaline ≤0.1 μg kg−1 min−1 or noradrenaline ≤0.1 μg kg−1 min−1 | Dopamine >15 μg kg−1min−1 or adrenaline >0.1 μg kg−1 min−1 or noradrenaline >0.1 μg kg−1 min−1 |
| Central nervous system | ||||
| Glasgow Coma Score | 13–14 | 10–12 | 6–9 | <6 |
| Renal | ||||
| Creatinine (μmol litre−1) or urine output (ml day−1) |
110–170 | 171–299 | 300–440 <500 |
>440 200 |
There is limited published data on the frequency of sepsis following cardiac surgery. Previous studies often limited their investigations to patients with positive microbiological cultures from specific sites such as the wound or the respiratory tract. Such studies used previous definitions of sepsis and identified sepsis in 0.5–2% of cardiac surgery patients.5, 6, 7 In these studies, sepsis was associated with mortality rates in the range of 17–79%. As organ dysfunction is frequent following cardiac surgery due to the inflammatory response to surgery and cardiopulmonary bypass (CPB), Sepsis-3 criteria could potentially diagnose sepsis in patients with transient organ dysfunction due to surgery and coincidental minor infection.
The objective of this study was to ensure that adoption of the Sepsis-3 guidelines is appropriate for patients undergoing cardiac surgery. To achieve this objective, we have assessed the incidence of sepsis as defined by the new guidelines and also investigated whether diagnosis with proven or suspected infection influences short- and mid-term clinical outcomes.
Methods
Patients and data collection
All relevant clinical and monitoring data were collected, prospectively, from consecutive adult patients admitted to the Cardiac Intensive Care Unit (CICU) after cardiac surgery at University Hospital of South Manchester between January 2013 and May 2015. Patients undergoing cardiac transplantation were excluded from the study.
Data were collected for the duration of the patients' first CICU admission following cardiac surgery from three sources. First, patient characteristics, preoperative morbidity, and outcome data were collected from the hospital's clinical governance database. Secondly, physiological variables, medication data, and case note data regarding the suspicion or diagnosis of infection were collected from the electronic patient record (EPR). Thirdly, haematology and biochemistry results, together with all microbiology reports were collected from the hospital's pathology database. Hourly recordings of physiological variables, medication administrated, and all available biochemical and haematological results were cleaned using cleaning algorithms in R Studio (R Foundation for statistical computing).8
Daily SOFA scores (Table 1) were calculated for each patient using the most abnormal value recorded for each variable on each day.2 For all patients who experienced a SOFA score increase of ≥2, the clinical notes were examined to identify suspected or proven infection. Proven infection was that confirmed by microbiological cultures (excluding isolated Candida albicans-positive sputum cultures, mixed growth urine samples, or screening swabs which were considered to indicate colonization). Infection was classified as suspected if antibiotics other than those given as standard prophylaxis were administered or suspicion of infection was documented in the clinical notes section of the EPR. All indicators of suspected or proven infection recorded within 24 h of the day of an increase in SOFA score of ≥2 increase were included to ensure that no suspected or proven infection were missed.
Missing data
Where blood analyses necessary for calculation of daily SOFA scores were missing, the last known appropriate result recorded for that patient was substituted. Bilirubin concentrations were not routinely measured for low risk patients, so there were 340 occasions (5.2% of all SOFA score calculations) when there was no bilirubin level available for the SOFA score calculation. In 257 of these cases, a bilirubin level subsequently measured for that patient was used. In the remaining 83 calculations (total of 39 patients), SOFA was calculated incorporating the median bilirubin concentration for all patients. In the one patient with no available creatinine level and two with no available platelet count, the median for the missing variable was used. All other data were complete.
Statistical analysis
Normally distributed data were described using the mean and standard deviation; data with non-parametric distributions were described using the median and inter-quartile range (IQR). Outcome measures were length of CICU stay (hours), 30-day mortality (defined as death due to any cause within the first 30 days after cardiac surgery), and 2-yr survival. The relationship between sepsis and length of CICU stay was analysed using the Kruskal–Wallis test as length of stay was not normally distributed. Univariate analyses of the relationship between sepsis and 30-day mortality were conducted using Fisher's exact test due to the low observed mortality rate. Two-year survival rates were compared using the log-rank test.
Multivariable linear regression analyses were performed to adjust for the effects of logistic European System for Cardiac Operative Risk Evaluation (EuroSCORE) and CPB time on length of CICU stay. Multivariable logistic regression analyses were performed to adjust for the effect of logistic EuroSCORE on 30-day mortality. The low number of deaths prevented the inclusion of additional confounders in these analyses.9 Finally, Cox proportional hazards analyses were performed to adjust for the influence of logistic EuroSCORE and CPB time on 2-yr survival. The Logistic EuroSCORE10 is an extensively validated preoperative risk prediction model for perioperative mortality that includes patient co-morbidities, variables reflecting cardiac function, and operative risk factors. It demonstrates good discriminative ability for UK cardiac surgery.11
Data collection was approved by the National Research Ethics Service–Haydock as part of the Vascular Governance Northwest Project (09/H 1000 10/2+5) and all analyses were performed using R.8
Results
During the study period, 2230 patients were admitted to CICU after cardiac surgery. The mean (range) age was 66.1 (18–93) yr and the majority of patients were men (1615, 72.4%). Full patient characteristics for the study population are shown in Table 2. Median length of CICU stay (IQR) was 48.8 (40.1–93.0) h. Overall 30-day mortality was 1.5% and 2-yr survival was 93.0%. SOFA increases of ≥2 were identified on 710 occasions in 323 patients. A total of 573 patients were discharged from the CICU on the first postoperative day. In these patients, only one SOFA score was available preventing the calculation of a difference between the daily SOFA scores. As a result, these patients were classified as not suffering sepsis during the CICU admission.
Table 2.
Patient characteristics. CABG, coronary artery bypass graft; CPB, cardiopulmonary bypass
| Variable | All (n=2230) | Unable to calculate SOFA increase (n=573) | No SOFA increase (n=1334) | SOFA increase >2 but no infection (n=112) | Sepsis suspected Infection (n=104) | Sepsis proven Infection (n=107) |
|---|---|---|---|---|---|---|
| Age, mean (range), yr | 66.1 (18–93) | 63.2 (23–86) | 66.9 (19–93) | 69.6 (28–87) | 65.5 (29–89) | 68.6 (18–91) |
| Female gender, n (%) | 615 (27.6) | 99 (17.3) | 418 (31.3) | 36 (32.1) | 31 (29.8) | 31 (29.0) |
| Height, mean (SD), cm | 169.6 (9.2) | 171.5 (8.7) | 168.9 (9.3) | 169.1 (8.2) | 169.5 (8.8) | 168.0 (10.6) |
| Weight, mean (SD), kg | 81.0 (15.9) | 81.7 (14.1) | 81.7 (16.6) | 81.6 (14.4) | 81.6 (15.2) | 80.1 (17.8) |
| Type of surgery | ||||||
| Isolated CABG, n (%) | 1214 (54.4) | 479 (83.6) | 619 (46.4) | 46 (41.1) | 31 (29.8) | 39 (36.4) |
| Isolated valve, n (%) | 482 (21.6) | 37 (6.5) | 367 (27.5) | 26 (23.2) | 26 (25.0) | 26 (24.3) |
| Isolated aortic, n (%) | 23 (1.0) | 2 (0.3) | 13 (1.0) | 2 (1.8) | 4 (3.8) | 2 (1.9) |
| Combined cardiac procedures, n (%) | 404 (18.1) | 19 (3.3) | 296 (22.2) | 30 (26.7) | 32 (30.8) | 27 (25.2) |
| Other, n (%) | 107 (4.8) | 36 (6.3) | 39 (2.9) | 8 (7.1) | 11 (10.6) | 13 (12.1) |
| Urgency | ||||||
| Elective/scheduled, n (%) | 1324 (59.3) | 317 (55.3) | 823 (61.7) | 61 (54.5) | 62 (59.6) | 61 (57.0) |
| Urgent, n (%) | 842 (37.8) | 244 (42.5) | 488 (36.5) | 45 (40.1) | 29 (27.9) | 36 (33.6) |
| Emergency/salvage, n (%) | 64 (2.9) | 12 (2.1) | 23 (1.7) | 6 (5.4) | 13 (12.5) | 10 (9.3) |
| Duration of CPB, median (inter-quartile range), min | 100.0 (79.0–128.0) | 87.0 (69–105.0) | 104.0 (84.0–134.0) | 101.0 (84.0–135.5) | 118.0 (94.0–165.0) | 108.0 (84.0–147.0) |
| Logistic EuroSCORE median (inter-quartile range) | 4.0 (2.1–7.7) | 2.3 (1.4–3.5) | 4.6 (2.4–8.4) | 5.1 (3.3–11.8) | 7.5 (3.1–19.7) | 7.8 (3.6–15.1) |
| Requirement for renal replacement therapy, n (%) | 107 (4.8) | 7 (1.2) | 24 (1.8) | 10 (8.9) | 26 (25.0) | 40 (37.4) |
| Mechanical ventilation >72 h, n (%) | 147 (6.6) | – | 37 (2.8) | 7 (6.3) | 29 (27.9) | 50 (46.7) |
Sepsis
The Sepsis-3 criteria for sepsis were met by 211 (9.5%) of the 2230 patients. Sepsis with suspected infection occurred in 104 patients (4.7%) and sepsis with proven infection was demonstrated in 107 (4.8%). The respiratory tract was the most frequent source of both proven (72.1%) and suspected infection (55.4%). Other sources of infection are shown in Table 3.
Table 3.
Suspected or proven sources of infection in those diagnosed with sepsis. NA, not applicable. *Multiple sites were implicated in many patients
| Suspected source | Suspected infection* | Proven infection* |
|---|---|---|
| Not specified (antibiotics started) | 51 | NA |
| Unknown | 10 | NA |
| Respiratory | 148 | 96 |
| Abdominal/gastrointestinal | 6 | – |
| Wound | 6 | 8 |
| Genitourinary | 3 | 5 |
| Bacteraemia/catheters | 6 | 24 |
| Endocarditis/myocarditis | 9 | – |
| Dental | 1 | – |
The median length of CICU stay (IQR) was 145.2h (114.5–261.7) for those with sepsis due to suspected infection, 211.5h (117.2–478.1) for those with sepsis due to proven infection, and 47.0h (28.8–72.6) for those without sepsis (P<0.01 for both). After controlling for the logistic EuroSCORE and CPB time using linear regression modelling, patients with sepsis had significantly longer CICU stays than those without. The increase in length of CICU stay (95% CI) was 134.1h (99.0–168.2) for those with suspected infection and 266.1h (231.6–300.7) for those with proven infection (P<0.01 for both). The linear regression model is detailed in Appendix I.
To ensure the length of stay analysis was not skewed by the 573 patients discharged on the first postoperative day, a sensitivity analysis using the same linear regression model on data taken exclusively from patients with two or more daily SOFA scores (n=1657) was performed. The increase in CICU length of stay (95% CI) attributed to a diagnosis of sepsis with suspected infection in this subgroup remained significant at 135.7h (99.1–172.3); the increase related to proven infection was 265.8 (229.7–301.9) h (P<0.01 for both).
The 30-day mortality was 6.6% for those who suffered sepsis compared with 1.0% for those who did not (P<0.01). The mortality rates for sepsis with suspected infection (5.8%) and sepsis with proven infection (7.5%) were both significantly higher than the rate of 1.0% for those without sepsis (P<0.01 for both). After adjusting for preoperative and intraoperative confounders using the logistic EuroSCORE (full model detailed Appendix I), the odds ratio associated with sepsis was 3.7 (95% CI 1.1–10.2, P=0.02) for suspected infection and 6.6 (95% CI 2.6–15.7, P<0.01) for proven infection.
Among those who suffered from sepsis, the main differences between survivors and those who died were that those who died had a higher median logistic EuroSCORE (16.7 vs 6.7, P<0.01) and were less likely to have undergone isolated coronary artery bypass graft or valve surgery (21.4% vs 60.4%, P=0.01). Rates of renal replacement therapy (71.4% vs 28.4%, P<0.01) and prolonged mechanical ventilation (92.9% vs 33.5%, P<0.01) were higher in non-survivors than survivors.
Among patients with sepsis, the 2-yr survival was 87.5% for those with suspected infection and 73.8% for those with proven infection compared with 94.3% for those without sepsis (P<0.01 for both). As seen in Fig. 1, the greatest difference in mortality rates was seen in the first 12 postoperative months. A second log-rank analysis which included only those patients alive 1 yr post surgery showed a smaller difference in the rates of survival to 2 yr between those who had suffered sepsis (98.1%) and those who had not (96.0%) which was no longer statistically significant (P=0.06).
Fig 1.
Two-year survival according to sepsis status.
The confounding effects of preoperative logistic EuroSCORE and CPB time on 2-yr survival were adjusted for using Cox proportional hazards regression. For those with sepsis due to suspected infection, the hazard ratio was 1.1 and the effect on survival was not statistically significant (95% CI 0.5–2.4, P=0.76). However, for sepsis due to proven infection compared with those without sepsis, the hazard ratio was 3.6 (95% CI 2.2–5.9, P<0.01). The model is detailed in Appendix I.
SOFA increase ≥2 in the absence of sepsis
A total of 112 patients developed a SOFA increase ≥2 in the absence of proven or suspected infection. The median length of CICU stay (IQR) for these patients was 83.2h (48.5–124.9). This was significantly shorter than the median CICU stays of 211.5h for those with sepsis due to proven infection and 145.2h for those with suspected infection (P<0.01 for both). On multivariable analysis (full model detailed in Appendix I), a SOFA increase ≥2 without sepsis was associated with a non-statistically significant difference in length of stay of 6.9h (95% CI -28.2–41.9, P=0.70).
The 30-day mortality rate for those with a SOFA increase ≥2 in the absence of sepsis was 2.7%. This was higher than that for patients with lesser increases in SOFA scores and lower than that for those with sepsis and suspected or proven infection, but none of these differences was statistically significant (P=0.10, P=0.32, and P=0.13, respectively). After adjusting for preoperative and intraoperative confounders using the logistic EuroSCORE (full model detailed in Appendix I), a SOFA increase ≥2 in the absence of sepsis was not significantly associated with 30-day mortality (odds ratio 2.1, 95% CI 0.5–6.2, P=0.23). The 2-yr survival rate for patients who suffered a SOFA increase ≥2 without sepsis was 91.1%. This was not significantly different to the rate of 94.3% for patients with stable or small increases (<2) in the SOFA score (P=0.13), neither was it significantly different from the 87.5% in those with suspected infection (P=0.38). It was however significantly higher than the 73.8% for those with proven infection and sepsis (P<0.01).
Septic shock
Of the 211 patients diagnosed with sepsis, 159 patients (75.4%) met criteria relating to serum lactate concentration and use of vasopressors compatible with a diagnosis of septic shock. For this subgroup, median (IQR) length of CICU stay was 193.2 (139.5–364.0) h, 30-day mortality was 8.8%, and the 2-yr survival rate was 76.7%. All of these results were significantly worse than for patients with sepsis who did not suffer septic shock (P<0.02 for all) (Table 4).
Table 4.
Patient outcomes. ICU, intensive care unit. *P<0.05 when compared with the frequency of the outcome in the No Sepsis group during univariate analyses
| Sepsis status | Subgroup | n (%) | 30-day mortality, n (%) | Median ICU stay, h (inter-quartile range) | 2-yr survival, n (%) |
|---|---|---|---|---|---|
| No sepsis | All patients | 2019 (90.5) | 20 (1.0) | 47.0 (28.8–72.6) | 1904 (94.3) |
| - SOFA increase not calculable | 573 (25.7) | 4 (0.7) | 22.6 (20.0–25.3) | 553 (96.5) | |
| - SOFA increase <2 | 1334 (59.8) | 13 (1.0) | 52.6 (45.4–86.2) | 1248 (93.6) | |
| - SOFA increase ≥2 without infection | 112 (5.0) | 3 (2.7) | 83.2 (48.5–124.9) | 102 (91.1) | |
| Sepsis | All patients | 211 (9.5) | 14 (6.6)* | 176.0 (115.7–404.6)* | 170 (80.5)* |
| - Suspected infection | 104 (4.7) | 6 (5.8)* | 145.2 (114.5–261.7)* | 91 (87.5)* | |
| - Proven infection | 107 (4.8) | 8 (7.5)* | 211.5 (117.2–478.1)* | 79 (73.8)* |
Discussion
This is the first study to validate the Sepsis-3 guidelines in a cohort of patients after cardiac surgery. The new guidelines allow patients with only suspected infection to be diagnosed with sepsis. This may be one reason why the incidence of sepsis following cardiac surgery in this study (9.5%) is higher than that reported in studies which only included those with infection proven by microbiological culture.5, 6, 7, 12 The frequency of sepsis with positive cultures (4.8%) was also higher than most previous studies of sepsis after cardiac surgery. However, the majority of the previous studies only included specific sources of infection such as the wound or respiratory tract.5, 6, 12, 13 Despite the higher incidence of sepsis observed in our cohort, the 30-day mortality of 6.6% for patients with sepsis was lower than that found in previous studies in cardiac surgery.5, 7, 14, 15
Sepsis based on the Sepsis-3 guidelines was a significant risk factor for adverse outcomes in our cohort. The 30-day mortality risk increased markedly in patients who met the Sepsis-3 criteria. Patients who suffered sepsis also had significantly longer CICU stays compared with patients who did not. Overall 2-yr survival rates were lower for patients with sepsis, although our secondary analysis, including only those who survived to 1 yr, illustrates that most of the impact of sepsis on mortality risk appears to be observed in the first 12 months. This relatively short-term effect on risk is different to that reported in patients from general ICUs which detected impact on survival in the longer term.16 This difference may be due to the cohort of patients included in our study. In the cardiac surgery patients studied, the organ dysfunction that triggered the diagnosis of sepsis often progressed to organ failure in patients who had already been physiologically stressed by their surgery. Some 31% of those with sepsis required renal replacement therapy and 37% required prolonged mechanical ventilation. Patients either recovered from these critical complications or died as a result of them within a relatively short period.
Studies in general ICU populations found that when Sepsis-3 criteria and SIRS-related sepsis criteria were applied to the same patients with suspected infection, the Sepsis-3 criteria identified fewer patients than SIRS-related criteria. They also showed that patients identified by the Sepsis-3 criteria were likely to suffer worse outcomes.17, 18 However, the variables included in the SIRS criteria are influenced by the inflammatory response to major surgery as well as treatments such as mechanical ventilation, patient warming, and perioperative beta-blockade which are frequently employed following cardiac surgery. Importantly, unlike the SOFA-related definitions, the SIRS criteria cannot recognise the effects of interventions on the absolute values of these parameters. Moreover, in our cohort, 88% of patients fulfilled the criteria for SIRS postoperatively. Therefore, using SIRS-related criteria to diagnose sepsis would have led to the vast majority of suspected infections resulting in a diagnosis of sepsis, even where inflammation did not progress above the postoperative baseline. Consequently, the Sepsis-3 definitions seem to provide the most appropriate means for detecting sepsis after cardiac surgery. This may be also true for patient groups who require critical care treatments after undergoing other types of major surgery or suffering from conditions which result in non-infective, inflammatory responses such as pancreatitis or severe burns.
A significant proportion of the patients diagnosed with sepsis in this study went on to meet the criteria for septic shock. The length of CICU stay, 30-day mortality, and 2-yr survival rate associated with septic shock were significantly worse than those for patients who suffered sepsis without septic shock. Further work to identify patients at the highest risk of developing septic shock would therefore be of clinical importance.
Approximately half of the patients diagnosed with sepsis in our study had proven infection and half had suspected infection; a proportion similar to that documented in the general ICU population.19 The new Sepsis-3 guidelines include suspicion of infection to ensure all patients with sepsis were identified and treated early.1 This analysis demonstrates that patients with suspected infection suffered outcomes more similar to those with proven infection than to those with no sepsis. While outcomes for patients with proven infection were worse than for those with suspected infection, it is clearly appropriate to adopt Sepsis-3, as this improves the early identification of patients at a high risk of adverse short-term outcomes.
Limitations
The new Sepsis-3 criteria identify worsening organ function by the change in the daily SOFA score rather than absolute values. The guidelines state that the baseline SOFA should be assumed to be 0 unless a patient has ‘pre-existing organ dysfunction (acute or chronic)’. Following cardiac surgery, the mean Day 1 SOFA score was 5.4 and over 90% of patients had a Day 1 SOFA score >2. Consequently, assuming a baseline SOFA score of 0 for patients undergoing this major surgery would be inappropriate. As a result, we required two daily postoperative scores in order to calculate the change in SOFA score and were therefore unable to diagnose sepsis before the second postoperative day. Some 26% of our patients were discharged on the first postoperative day and therefore could not be classified as having sepsis. The median CICU stay in these early discharge patients was 22.6 h and their 30-day mortality was only 0.7%. Although sepsis would be unusual on the first day following cardiac surgery we have performed a sensitivity analysis excluding these patients and the conclusions from the analysis are unchanged.
The small 30-day mortality rate in the study (34 deaths) prevented the inclusion of additional confounders into the logistic regression analyses for 30-day mortality.9 Logistic EuroSCORE was chosen to be the sole confounder included in the model as it was considered the most clinically relevant variable.
As with any observational study, there were missing data. The incidence of missing data was, however, very low and SOFA scores were calculated using imputed bilirubin values for less than 2% of patients. In three of these patients, creatinine concentration or platelet counts were also imputed. A further potential limitation of this study is that it was conducted at a single centre. Although our centre performs almost all aspects of adult cardiac surgery, a larger, multicentre study would be the optimal method to validate these findings further.
Conclusion
This is the first study exploring how Sepsis-3 criteria influence the diagnosis of sepsis in cardiac surgery patients. Patients with sepsis due to proven or suspected infection suffered prolonged CICU stays and increased 30-day mortality, justifying the adoption of Sepsis-3 guidelines in cardiac surgery.
Authors' contributions
Study design: S.H.H., I.M., C.N.M., S.W.G.
Data collection and cleaning: S.H.H., M.H.
Data analysis: all authors.
First draft of manuscript: S.H.H.
Revision of manuscript: all authors.
Declaration of interest
None declared.
Funding
This work was supported by funding from the British Heart Foundation (grant number PG/16/80/32411).
Editorial decision: October 21, 2017
Handling editor: J.P. Thompson
Appendix I.
Models described in the manuscript
Table A1.
Linear regression model for length of cardiac intensive care unit (CICU) stay accounting for effects of confounders
| Variable | Beta-coefficient (h) | 95% CI | P-value | |
|---|---|---|---|---|
| Intercept | 2.5 | −14.0 | 19.1 | 0.76 |
| Sepsis–proven infection | 266.1 | 231.6 | 300.7 | <0.01 |
| Sepsis–suspected infection | 134.1 | 99.0 | 168.2 | <0.01 |
| Logistic EuroSCORE (per point) | 2.4 | 1.5 | 3.3 | <0.01 |
| Cardiopulmonary bypass time (per min) | 0.5 | 0.4 | 0.6 | <0.01 |
Table A2.
Linear regression model for length of cardiac intensive care unit (CICU) stay accounting for effects of confounders in those who stayed long enough for two or more sequential organ failure assessment (SOFA) scores to be calculated
| Variable | Beta-coefficient (h) | 95% CI | P-value | |
|---|---|---|---|---|
| Intercept | 28.3 | 7.9 | 48.6 | <0.01 |
| Sepsis–proven infection | 265.8 | 229.7 | 301.9 | <0.01 |
| Sepsis–suspected infection | 135.7 | 99.1 | 172.3 | <0.01 |
| Logistic EuroSCORE (per point) | 2.1 | 1.1 | 3.1 | <0.01 |
| Cardiopulmonary bypass time (per min) | 0.3 | 0.2 | 0.5 | <0.01 |
Table A3.
Logistic regression model for 30-day mortality
| Variable | Beta coefficient | Odds ratio | 95% CI | P-value | |
|---|---|---|---|---|---|
| Intercept | −5.2 | <0.01 | |||
| Sepsis–proven infection | 1.9 | 6.6 | 2.6 | 15.7 | <0.01 |
| Sepsis–suspected infection | 1.3 | 3.7 | 1.1 | 10.2 | 0.02 |
| Logistic EuroSCORE (per point) | 0.0 | 1.1 | 1.0 | 1.1 | <0.01 |
Table A4.
Cox proportional hazards ratio model for 2 yr non-survival
| Variable | Hazard ratio | 95% CI | P-value | |
|---|---|---|---|---|
| Sepsis–proven infection | 3.6 | 2.2 | 5.9 | <0.01 |
| Sepsis–suspected infection | 1.1 | 0.5 | 2.4 | 0.76 |
| Logistic EuroSCORE (per point) | 1.0 | 1.0 | 1.0 | 0.02 |
| Cardiopulmonary bypass time (per min) | 1.0 | 1.0 | 1.0 | <0.01 |
Table A5.
Linear regression model for length of cardiac intensive care unit (CICU) stay investigating significance of a sequential organ failure assessment (SOFA) rise ≥2 in the absence of sepsis
| Variable | Beta-coefficient (h) | 95% CI | P-value | |
|---|---|---|---|---|
| Intercept | 6.3 | −11.4 | 24.1 | 0.48 |
| SOFA rise ≥2 without sepsis | 6.9 | −28.2 | 41.9 | 0.70 |
| Logistic EuroSCORE (per point) | 3.6 | 2.7 | 4.6 | <0.01 |
| Cardiopulmonary bypass time (per min) | 0.6 | 0.4 | 0.7 | <0.01 |
Table A6.
Logistic Regression Model for 30-day mortality investigating significance of a sequential organ failure assessment (SOFA) rise ≥2 in the absence of sepsis
| Variable | Beta-coefficient | Odds ratio | 95% CI | P-value | |
|---|---|---|---|---|---|
| Intercept | −4.9 | <0.01 | |||
| SOFA rise ≥2 without sepsis | 0.7 | 2.1 | 0.5 | 6.2 | 0.23 |
| Logistic EuroSCORE (per point) | 0.1 | 1.1 | 1.0 | 1.1 | <0.01 |
References
- 1.Singer M., Deutschman C.S., Seymour C. The third international consensus definitions for sepsis and septic shock (Sepsis-3) JAMA. 2016;315:801–810. doi: 10.1001/jama.2016.0287. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Vincent J.L., Moreno R., Takala J. The SOFA (Sepsis-related organ failure assessment) score to describe organ dysfunction/failure. On behalf of the working group on sepsis-related problems of the European society of intensive care medicine. Intensive Care Med. 1996;22:707–710. doi: 10.1007/BF01709751. [DOI] [PubMed] [Google Scholar]
- 3.Bone R.C., Balk R.A., Cerra F.B. Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. The ACCP/SCCM Consensus Conference Committee. American College of Chest Physicians/Society of Critical Care Medicine. Chest. 1992;101:1644–1655. doi: 10.1378/chest.101.6.1644. [DOI] [PubMed] [Google Scholar]
- 4.Levy M.M., Fink M.P., Marshall J.C. 2001 SCCM/ESICM/ACCP/ATS/SIS international Sepsis definitions conference. Intensive Care Med. 2003;29:530–538. doi: 10.1007/s00134-003-1662-x. [DOI] [PubMed] [Google Scholar]
- 5.Michalopoulos A., Stavridis G., Geroulanos S. Severe sepsis in cardiac surgical patients. Eur J Surg. 1998;164:217–222. doi: 10.1080/110241598750004670. [DOI] [PubMed] [Google Scholar]
- 6.Fowler V.G., O’Brien S.M., Muhlbaier L.H., Corey G.R., Ferguson T.B., Peterson E.D. Clinical predictors of major infections after cardiac surgery. Circulation. 2005;112 doi: 10.1161/CIRCULATIONAHA.104.525790. I-358–I-365. [DOI] [PubMed] [Google Scholar]
- 7.Oliveira D.C., Oliveira Filho J.B., Silva R.F. Sepsis in the postoperative period of cardiac surgery: problem description. Arq Bras Cardiol. 2010;94:332–336. doi: 10.1590/s0066-782x2010000300012. [DOI] [PubMed] [Google Scholar]
- 8.R Core Team . R Foundation for Statistical Computing; Vienna, Austria: 2015. R: A language and environment for statistical computing. [Google Scholar]
- 9.Peduzzi P., Concato J., Kemper E., Holford T.R., Feinstein A.R. A simulation study of the number of events per variable in logistic regression analysis. J Clin Epidemiol. 1996;49:1373–1379. doi: 10.1016/s0895-4356(96)00236-3. [DOI] [PubMed] [Google Scholar]
- 10.Roques F., Michel P., Goldstone A.R., Nashef S.A. The logistic EuroSCORE. Eur Heart J. 2003;24:881–882. doi: 10.1016/s0195-668x(02)00799-6. [DOI] [PubMed] [Google Scholar]
- 11.Hickey G.L., Grant S.W., Murphy G.J. Dynamic trends in cardiac surgery: why the logistic EuroSCORE is no longer suitable for contemporary cardiac surgery and implications for future risk models. Eur J Cardiothorac Surg. 2013;43:1146–1152. doi: 10.1093/ejcts/ezs584. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Chen L.F., Arduino J.M., Sheng S. Epidemiology and outcome of major postoperative infections following cardiac surgery: risk factors and impact of pathogen type. Am J Infect Control. 2012;40:963–968. doi: 10.1016/j.ajic.2012.01.012. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Sahu M., Siddharth B., Choudhury A. Incidence, microbiological profile of nosocomial infections, and their antibiotic resistance patterns in a high volume cardiac surgical intensive care unit. Ann Card Anaesth. 2016;19:281–287. doi: 10.4103/0971-9784.179625. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Michalopoulos A., Geroulanos S., Rosmarakis E.S., Falagas M.E. Frequency, characteristics, and predictors of microbiologically documented nosocomial infections after cardiac surgery. Eur J Cardiothorac Surg. 2006;29:456–460. doi: 10.1016/j.ejcts.2005.12.035. [DOI] [PubMed] [Google Scholar]
- 15.Falagas M.E., Rosmarakis E.S., Rellos K., Michalopoulos A., Samonis G., Prapas S.N. Microbiologically documented nosocomial infections after coronary artery bypass surgery without cardiopulmonary bypass. J Thorac Cardiovasc Surg. 2006;132:481–490. doi: 10.1016/j.jtcvs.2006.05.019. [DOI] [PubMed] [Google Scholar]
- 16.Winters B.D., Eberlein M., Leung J., Needham D.M., Pronovost P.J., Sevransky J.E. Long-term mortality and quality of life in sepsis: a systematic review. Crit Care Med. 2010;38:1276–1283. doi: 10.1097/CCM.0b013e3181d8cc1d. [DOI] [PubMed] [Google Scholar]
- 17.Donnelly J.P., Safford M.M., Shapiro N.I., Baddley J.W., Wang H.E. Application of the third international consensus definitions for Sepsis (Sepsis-3) classification: a retrospective population-based cohort study. Lancet Infect Dis. 2017;17:661–670. doi: 10.1016/S1473-3099(17)30117-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Sterling S.A., Puskarich M.A., Glass A.F., Guirgis F., Jones A.E. The impact of the Sepsis-3 septic shock definition on previously defined septic shock patients. Crit Care Med. 2017;45:1436–1442. doi: 10.1097/CCM.0000000000002512. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Vincent J.L., Sakr Y., Sprung C.L. Sepsis in European intensive care units: results of the SOAP study. Crit Care Med. 2006;34:344–353. doi: 10.1097/01.ccm.0000194725.48928.3a. [DOI] [PubMed] [Google Scholar]