The study patients were diagnosed with IE in our center between 1st January 2005 and 31st July 2011. All met the criteria of definitive or probable endocarditis according to the Duke classification.9 In all cases a history was compiled, including disease antecedents, microbiological study, echocardiographic variables, disease complications, and severity assessment based on the APACHE II (Acute Physiology and Chronic Health Evaluation II),10 SAPS II (Simplified Acute Physiology Score II)11 and SOFA (Sepsis-related Organ Failure Assessment) scores.12

Dependent variables were taken to be admission to the ICU (regarded as any stay in the Unit, based on the need to treat at least the failure of one organ, and excluding habitual stay after heart surgery) and in-hospital mortality (including deaths before leaving hospital regardless of whether the patient had been subjected to surgery or not).

Acute IE was defined as episodes with an evolutive course of under two weeks from appearance of the first signs and symptoms. The development of acute renal failure was considered based on the RIFLE criteria referred to renal damage or failure. We only considered New York Heart Association (NYHA) functional classes III and IV in reference to the development of heart failure. Cerebral or peripheral embolism had to be evidenced by imaging techniques (CAT, MRI or ultrasound) in order to be diagnosed as such. Emergency surgery in turn referred to all surgery not delayed more than 24h after establishing the indication.

In all patients at least three blood cultures were performed, with serological testing for Brucella, Legionella, Mycoplasma and Coxiella burnetii in the case of negative blood culture findings and/or under the criterion of the supervising physician.

In the transthoracic and transesophageal echocardiographic study, we registered the location of the vegetation (where present), the existence of new onset valve insufficiency ≥2/4, the appearance of a new prosthetic dehiscence, and the development of periannular complications (abscesses, pseudoaneurysms or fistulas).

Unless precluded by the seriousness of the patient condition, ultrasound or an abdominal scan was performed, together with a cranial imaging study in search of embolic phenomena.

The data are presented as the percentage or mean±standard deviation (SD) or median (interquartile range), as applicable. Normality of the distribution of the continuous variables was evaluated using the Kolmogorov–Smirnoff test. Categorical variables in turn are presented as the absolute frequency (percentage), and were compared using the chi-squared test or Fisher exact test, as required. Continuous variables were compared with the Student t-test for non-paired data or the Mann–Whitney U-test as appropriate. The Wilcoxon test was used to assess the evolutive differences in the severity scores. Forward stepwise logistic regression analysis (likelihood ratio test) was carried out in application to those variables showing p<0.05, in order to identify the risk factors for admission to the ICU and for in-hospital mortality. Goodness of fit was assessed using the Hosmer–Lemeshow test. The SPSS (version 16.0) statistical package was used for the analysis of the results.

During the study period a total of 102 patients were diagnosed with IE in our center. Admission to the ICU was considered necessary in 38 of these patients. Among these subjects, 25 (65.8%) came directly from the emergency area, 7 (18.4%) from the Department of Cardiology, 3 (7.9%) from Internal Medicine, and the remaining three from other Departments.

The causes indicating intensive care were predominantly severe sepsis or septic shock (14 patients, 36.8%) and severe heart failure or cardiogenic shock (14 cases, 36.8%). In three patients (7.9%) the cause of admission was neurological worsening. In the remaining cases (7 patients, 18.4%) the reasons were diverse but related to the severity of the patient condition, and included resuscitated cardiopulmonary arrest and pneumonia. Nineteen (50%) of the patients required mechanical ventilation.

The most relevant demographic and clinical differences between the patients admitted to Intensive Care and those not admitted to the ICU are shown in Table 1. The patients admitted to Intensive Care more frequently presented acute endocarditis and mitral valve disease, and it was more common for the disorder to be complicated by heart failure and embolic phenomena. The mortality predictor scores (SAPS II, APACHE II, SOFA) upon admission to hospital were also higher in the case of admission to the ICU.

While there were no differences in the global indication of surgery, a total of 11 patients were emergency referred for surgery from the ICU (45.8% of those operated upon), versus only two patients (5.9%) from other medical Departments (p<0.001). The causes leading to surgery were mainly heart failure (36.2%), valve destruction with severe valve insufficiency (27.6%), and persistent fever (15.5%). Other causes included annular alterations and the presence of large vegetations.

Regarding the etiological agent, in the patients admitted to the ICU there was a clear predominance of Staphylococcus aureus, which proved significantly more common than in the patients not admitted to Intensive Care (39.5 vs 20.3%, p=0.036). Other causal organisms were streptococci (18.4%), enterococci (10.5%) and other agents (7.9%). Coagulase-negative staphylococci were less frequent compared with the patients not admitted to the ICU (5.3 vs 20.3%, p=0.038). Lastly, patients with negative blood culture findings represented 18.4% of the cases in the ICU.

Table 2 reports the echocardiographic characteristics. No significant differences were observed in the proportion of cases of valve prosthesis endocarditis, in the frequency of annular alterations or in the presence of moderate or severe valve insufficiency. However, the percentage of patients with mitral valve disease was greater in the ICU, and no cases of tricuspid valve endocarditis were registered during the period studied.

As predictors of admission, the multivariate analysis identified cerebral embolism (OR=3.89; 95%CI, 1.06–14.3; p=0.041), the SOFA score upon admission to hospital (OR=1.60; 95%CI, 1.29–2.00; p<0.001), and mitral valve disease (OR=7.13; 95%CI, 2.12–24; p=0.002). The Hosmer–Lemeshow test yielded p=0.87.

The in-hospital mortality rate reached 42.1% in the ICU versus 18.8% among the patients not admitted to Intensive Care (p=0.011). The predominant cause of death was septic shock (51.8%), followed by severe heart failure or cardiogenic shock (25.9%) and neurological complications (14.8%).

According to the univariate analysis, the patients who died during admission were older, with a greater prevalence of diabetes mellitus, more frequent infection due to S. aureus, and with complications most often in the form of heart failure, cerebral embolism and renal failure. The scores of the severity scales upon admission were also significantly greater. Although the percentage of operated patients was higher among the survivors, statistical significance was not reached. Table 3 shows the main differences between the patients who survived the hospital stage and those who did not.

As independent predictors of mortality, the multivariate analysis (Table 4) identified infection due to S. aureus (OR=3.49; 95%CI, 1.02–11.93; p=0.046), the development of heart failure (OR=4.18; 95%CI, 1.17–14.94; p=0.028), cerebral embolism (OR=8.45; 95%CI, 1.89–37.74; p=0.005) and the SAPS II score (OR=1.09; 95%CI, 1.04–1.15; p<0.001). The Hosmer–Lemeshow test yielded p=0.81.