Mortality and morbidity in community-acquired sepsis in European pediatric intensive care units: a prospective cohort study from the European Childhood Life-threatening Infectious Disease Study (EUCLIDS)

The EUCLIDS is a Seventh Framework Programme (FP7) project in the context of the European Union’s Research and Innovation funding program for 2007–2013. This large-scale prospective, multicenter, cohort study aimed to identify genes, and biological pathways that determine susceptibility and severity in life-threatening bacterial infections of childhood. The EUCLIDS clinical network includes predominantly academic pediatric hospitals that host a total of 52 PICUs from 7 European countries; Austria (9), Germany (7), Lithuania (1), The Netherlands (5), Spain (9), Switzerland (8), and the UK (13).

From July 2012 to January 2016, patients aged 29 days to 18 years admitted with community-acquired sepsis to PICUs in participating centers were prospectively enrolled in the study. The 2005 pediatric consensus criteria for sepsis were used, dividing patients into those with sepsis, severe sepsis, or septic shock [21]. Healthcare-associated infections [22], patients undergoing bone marrow transplant, and patients already recruited who were readmitted within the same illness episode were excluded. Children with a central venous catheter at admission were not excluded. Although the consortium was specifically interested in patients with invasive meningococcal, pneumococcal, staphylococcal, salmonella, and group A streptococcal infections, representing the most common causes of community-acquired sepsis in children, patients with illness due to other organisms were included as well. Patients were recruited as early as possible in the illness within a time window from presentation to the time when culture results became available.

This study was conducted in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines. The study protocol was approved by at least one ethical review board in every country (Coordinating Center Research Ethics Committee reference: 11/LO/1982) [23]. Written informed consent was obtained from parents or legal guardians. In the Swiss study [24, 25], consent was obtained for collection of blood for research, but waiver of consent for collection of anonymized epidemiological data was approved.

Data on clinical presentation, underlying disease, illness severity, management, microbiological results, and outcome were collected prospectively. Children were split into four age categories; infants (29 days to < 1 year), toddlers (≥ 1 year to < 5 years), school-aged children (≥ 5 years to < 12 years), and adolescents (≥ 12 years to < 18 years). Underlying conditions at admission to the PICU were classified following the pediatric complex chronic conditions classification system [26]. Illness severity was measured by the Pediatric risk of mortality score (PRISM) [27] and Pediatric Index of Mortality (PIM2) [28]. We studied lactate values obtained on admission, concomitant with PIM2 data collection. Invasive bacterial infections were defined as isolation by culture or PCR of a bacterial organism from a normally sterile site. We considered blood, cerebrospinal fluid, urine, bronchoalveolar lavage, joint aspirate, abscess aspirate, intraoperative swabs, and pleural aspirate as sterile sites. Urine positive for pneumococcal antigen was also considered as an invasive bacterial infection if patients met sepsis criteria. Positive cultures from sites such as endotracheal tube aspirate, nasopharyngeal aspirate, throat/nasal swabs, and wounds were not considered as sterile sites. We defined potentially vaccine-preventable infections as infections caused by pathogens that are included in currently available national immunization programs, with a focus on Haemophilus influenzae type B (HiB), meningococcus serogroups ACWY (MenACWY), meningococcus serogroup B (MenB), meningococcus serogroup C (MenC), pneumococcal conjugate vaccine 7 (PCV7, Prevnar, serotypes 4, 6B, 9V, 14, 18C, 19F and 23F), pneumococcal conjugate vaccine 10 (PCV10, Synflorix, additional serotypes 1, 5, 7F), pneumococcal conjugate vaccine 13 (PCV13, Prevnar 13, additional serotypes 3, 6A, 19A) and pneumococcal polysaccharide vaccine 23 (PPSV23, additional serotypes 2, 8, 9N, 10A, 11A, 12F, 15B, 17F, 20, 22F, 33F). Data on routine immunization schedules and uptake in the countries involved are presented in Additional file 1: Table S1. We classified patients as primary bloodstream infection and sepsis without a known source (grouped as no focus) versus patients with a clinical focus of infection. Patients admitted with systemic inflammatory response syndrome (SIRS) in the presence of suspected infection (i.e. sepsis), in whom a bacterial, viral, or fungal infection eventually could not be confirmed, were categorized as clinical presentation other.

The primary outcome measure was death in hospital, recorded as alive or death status at the time of hospital discharge. Secondary outcomes were assessed at time of hospital discharge and included disability, PICU-free days censored at day 28 (days alive and free from the need for intensive care), and hospital length of stay. Disability was defined as a Pediatric Overall Performance Category (POPC) scale > 1 [29], need for skin graft, amputation, or hearing loss. The POPC scale was determined either by direct observation or by chart review and ranges from 1 to 6: (1) good overall performance, (2) mild overall disability, (3) moderate overall disability, (4) severe overall disability, (5) coma or vegetative state, and (6) brain death. A description of these categories is presented in Additional file 1: Table S2 [29]. PICU-free days in patients who died were considered zero. All data were collected in web-based case report forms. Monthly telephone conferences, biannual meetings, clinical protocols including case definitions, data audits, and monitoring, ensured uniform procedures among study sites.

Categorical variables are presented as counts (percentages). We used the chi-Square test or Fisher’s exact test to compare frequency distributions between two categorical variables. Post-hoc Bonferroni correction for multiple testing was applied when we compared age groups with features on clinical presentation or pathogens. Continuous variables are presented either as mean (± standard deviation (SD)) for data with a parametric distribution or as median (interquartile range (IQR)) for non-parametric data. We tested differences between groups with analysis of variance (ANOVA) or Kruskal-Wallis and Student’s t test, or Mann-Whitney U test, as appropriate. Logistic regression (of binary outcome measures) and linear regression (of continuous outcome measures) were used to identify independent predictors. Variables with a P value <0.20 in the univariable analysis were included in the multivariable analysis. In the multivariable analysis, we only included one parameter of illness severity (PIM2), because of multicollinearity of the illness severity parameters. Area under the receiver operating characteristic (AUROC) curve analysis was applied to determine the Youden index. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (PLR), and negative likelihood ratio (NLR) were calculated for the optimal cutoff value of lactate. Statistical analyses were performed with SPSS version 21 (Armonk, USA). Graphs were created in GraphPad Prism 5.00. A P value <0.05 was considered statistically significant.

Primary bloodstream infection and sepsis without a known source among patients with community-acquired sepsis accounted for 278 (35%) admissions to the PICU. The other most common clinical illnesses were meningitis/encephalitis (n = 182, 23%) and pneumonia (n = 149, 19%) (Additional file 1: Figure S1). Clinical presentation were similar across age groups, apart from osteomyelitis/septic arthritis, which was diagnosed more frequently in school-aged children than in infants (7.3% versus 0.4%, P value = 0.002).

Bacterial etiology was confirmed in 428 patients (54%), including 334 patients (42%) with a positive blood culture, and pathogen distribution was associated with age (Fig. 2). Neisseria meningitidis was the most commonly identified pathogen (n = 131, 31%), of which serogroup B was most prevalent (n = 89, 68%), followed by Streptococcus pneumoniae (n = 78, 18%, of which serotypes 3 (n = 7, 14%) and 10A (n = 6, 12%) were most commonly identified in those with serotyping information available (n = 51)) (Additional file 1: Table S3).

Of the 466 patients with septic shock, an invasive bacterial infection was confirmed in 255 patients (55%). N. meningitidis (n = 91, 36%) and group A streptococcus (n = 49, 19%) were the most commonly identified pathogens, followed by Streptococcus pneumoniae (n = 33, 13%).

Invasive ventilation was used in 519 patients (69%) (median length of invasive respiratory support 5 days, IQR 3–8, n = 43 with missing data) and vasoactive agents for 418 patients (57%) (median 3 days, IQR 2–5 days, n = 56 with missing data). Infants needed invasive ventilation more frequently than adolescents (73% versus 58%, P value = 0.03).

Of the 795 children admitted to PICU with community-acquired sepsis, 51 patients (6%) died. Mortality increased to 10% (n = 45) in patients with septic shock. Univariable analysis showed that the presence of bacteremia (odds ratio (OR) 4.4, 95% confidence interval (CI) 2.3–8.4, P < 0.001) and infections caused by S. pneumoniae (OR 2.5, 95% CI 1.2–5.1, P = 0.01) were associated with mortality in patients with sepsis (Table 2). In addition, illness severity, as measured by PRISM score, PIM2 score, invasive ventilation, the need for inotropes, and higher lactate at PICU admission, were also associated with sepsis mortality. The AUROC for lactate as a predictor of mortality was 0.723 (95% CI 0.624–0.822), with an optimal cutoff value of 2.2 mmol/L (sensitivity 0.78, specificity 0.60) (Additional file 1: Figure S2).

Infection caused by S. pneumoniae (OR 4.1, 95% CI 1.1–16.0, P = 0.04) and illness severity (PIM2 score OR 2.8, 95% CI 1.3–6.1, P < 0.01) remained independently significantly associated with mortality in multivariable analysis. A trend towards higher mortality was observed for bacteremia (OR 7.4, 95% CI 1.0–56.6, P = 0.06). PICU mortality did not differ significantly across age categories or countries. Also, the presence of an underlying condition at admission to the PICU was not associated with mortality.

The median PICU-free days to day 28 were 23 days (IQR 18–25) and the median hospital length of stay was 12 days (IQR 8–21). PIM2 score (B = − 0.202, P < 0.001), invasive S. pneumoniae infections (B = − 0.161, P = 0.02), and invasive Staphylococcus aureus infections (B = − 0.163, P = 0.01) were independent predictors of PICU-free days. PIM2 score (B = 0.270, P < 0.001), pneumonia (B = 0.145, P = 0.04), and invasive Staphylococcus aureus infections (B = 0.234, P = 0.001) were independent predictors of hospital length of stay (Additional file 1: Table S4).

Among survivors who did not have an underlying condition at admission to PICU, i.e. previously healthy children, 24% (83/349 patients whose data were available) were discharged with some disability. Disability data were available on 339/421 (81%) survivors of septic shock. In these, 120/339 (35%) patients had disability at PICU discharge, including 45 (13%) with mild overall disability, 29 (9%) with moderate overall disability, 35 (10%) with severe overall disability, 4 (1.2%) who had undergone amputation, and 7 (2.0%) who had undergone skin graft. Outcome as measured by mortality and POPC score was worst in patients admitted with pneumonia (Additional file 1: Figure S3) and in patients with invasive bacterial infections caused by Streptococcus pneumoniae (Additional file 1: Figure S4). When comparing patients discharged with and without disability, by univariable and multivariable analysis, the PIM2 score (OR 3.4, 95% CI 1.8–6.4, P < 0.001) and infections caused by Streptococcus pneumoniae (OR 5.4, 95% CI 1.8–15.8, P < 0.01) were independent predictors of disability (Table 3).