Study population

This study was undertaken at Red Cross War Memorial Children’s Hospital (RCWMCH), a tertiary level children’s hospital in Cape Town, South Africa, which serves as a major referral centre for children up to 13 years. The hospital does not provide routine postnatal care or have a dedicated neonatal intensive care facility.

Study design

A retrospective analysis was conducted of all paediatric patients seen at RCWMCH with S. aureus bloodstream infections between January 2007 and December 2011. All patients from whom S. aureus had been isolated from blood culture specimens over the five year period were identified using the National Health Laboratory Service (NHLS) database. Hospital clinical records were then retrospectively examined. Data were collected on demographics, clinical diagnosis, risk factors for infection, clinical and laboratory markers of infection, antibiotic susceptibility profiles, management of infection and clinical outcome.

Definitions

SAB was defined as the isolation of S. aureus from a blood culture specimen. Isolates were identified and tested for antimicrobial susceptibility at Groote Schuur Hospital NHLS microbiology laboratory, according to the Clinical and Laboratory Standards Institute guidelines appropriate for the years in question [15]. Isolates were identified using the VITEK 2 system (BioMérieux, France). Antibiotic susceptibility was determined using either standard Kirby Bauer disc diffusion, or the automated VITEK 2 system [16]. The D-test for inducible clindamycin resistance was performed on erythromycin resistant isolates, and D-test positive isolates were reported as clindamycin resistant. For MRSA isolates tested by disc diffusion, vancomycin MICs were determined using the E-test (BioMérieux, France). For isolates where susceptibility testing was performed using the Vitek 2, the vancomycin MIC as determined by the Vitek 2 was recorded. All S. aureus organisms cultured from blood specimens were considered to be clinically significant. However, multiple isolates of the same organism from the same patients were excluded from analysis if they were clinically deemed to be related to the incident infection. The isolation of S. aureus from a blood culture drawn more than 48 hours after admission to hospital, or drawn at readmission within 48 hours of discharge, was considered to be a nosocomial infection. Based on the definitions utilized by the Centers for Disease Control and Prevention, healthcare-associated infections were defined as those that occurred within 48 hours of admission to hospital, with at least one of the following healthcare-associated risk factors: hospitalization or surgery within one year of onset of infection, presence of an invasive device at the time of admission, resident of a long-term care facility, and history of MRSA colonization or infection [17]. Community-acquired MRSA infections were those identified within 48 hours of hospital admission, but with none of the aforementioned healthcare-associated risk factors. Clinical diagnoses during the SAB episode were classified according to the assessment of the attending physician in the patient records. HIV infection was determined by HIV PCR testing in children younger than 18 months, and by HIV ELISA testing in children over 18 months. HIV exposure was identified by history of maternal HIV infection, or a positive HIV antibody test in infants less than 18 months who tested HIV PCR negative. Leucocytosis was regarded as a white cell count above the normal range for age according to the NHLS reference standards. Malnutrition was defined according to WHO growth standards as weight-for-height or height-for-age z-score -2 SD below the mean WHO reference values. Severe malnutrition was defined as weight-for-height or height-for-age z score below -3SD of the mean, or by the presence of nutritional oedema [18].

Statistical analysis

Weight-for-age z-scores (WAZ), height-for-age z-scores (HAZ) and weight-for-height z-scores (WHZ) were calculated using Epi-Info, version 3.5.3 (2011). Data were analysed using Stata Statistical Software version 11.1 (StataCorp, College Station, Texas). Non-parametric variables were evaluated with Wilcoxon rank-sum tests and Kruskal-Wallis tests, and categorical data with chi-square tests. Odds ratios, together with 95% confidence intervals, were utilized to describe measures of effect. For all analyses, p <0.05 was considered statistically significant. Logistic regression was performed to assess risk factors for MRSA infection and risk factors for mortality. Variables with p <0.1 on univariate analysis were selected for stepwise forward multivariate analysis.

Ethical considerations

The study was conducted in accordance with the Declaration of Helsinki and was approved by the Departmental Research Committee of the Department of Child and Adolescent Health, and the Research Ethics Committee, Faculty of Health Sciences, University of Cape Town (HREC REF:029/2011).

Incidence of infection

Between January 2007 and December 2011, 365 episodes of S. aureus bacteraemia (SAB) were identified, 270 with MSSA (74%) and 95 with MRSA (26%). S. aureus was isolated on subsequent culture specimens, including blood, bone, tissue, pus, pleural, pericardial, peritoneal and synovial fluids in 143/365 episodes (39%). Basic demographic (age, gender), laboratory and microbiological data were accessible for all SAB episodes. Unfortunately, due to the retrospective design of the study, clinical and outcome data were not available for all cases.

The annual incidence of SAB was 3.28 cases/1000 hospital admissions per year with a mean frequency of 2.43 MSSA cases and 0.85 MRSA cases per 1000 hospital admissions per year. There was no significant change in the frequency of SAB over the study period (p=0.38). While rates of MRSA infection remained fairly constant from 2007-2011, the incidence of nosocomial SAB rose steadily from 2007 and peaked in 2010 (Figure 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Incidence of S. aureus bacteraemia (SAB) per 1000 hospital admissions.

https://doi.org/10.1371/journal.pone.0078396.g001

Characteristics of study population

Demographic characteristics and basic infective markers of the cohort are outlined in Table 1. The median age of patients with staphylococcal bacteraemia was 11 months. Neonates accounted for 36/365 (9.9%) of episodes while infants represented 189/365 (51.8%) of SAB. Children with MRSA infection were significantly younger (p=0.03), and more frequently underweight (p=0.00) and stunted (p=0.00). In those children whose HIV exposure and testing status was known, 51 (19.8%) were HIV-infected and 48 (19.4%) were HIV-exposed but uninfected (Table 1). Notably 107 (29%) children had an unknown HIV status. HIV infection, but not HIV exposure, was significantly associated with MRSA bacteraemia (p=0.00). Analysis of infective markers, including white cell count (WCC), neutrophil count, band count and C-reactive protein (CRP), revealed no differences between MSSA and MRSA bacteraemia (Table 1).

Seventeen patients had two distinct episodes of SAB based on isolation of the second organism more than 30 days after the initial SAB episode and after completing an appropriate course of antibiotic therapy. Six of these patients first had MSSA followed by MRSA bacteraemia with a median of 51 days (range 4-278 days) between episodes. The remaining 11 patients are described in Table 2. Six patients had two episodes of SAB more than 3 months apart (median 221 days; range 101-807 days). The remaining 5 patients had a median of 64 days (range 32-64 days) between the two episodes of bacteraemia. However, as molecular typing was not available, it was not possible to determine if these episodes were due to re-infection with the same type of organism or a completely new S. aureus infection.

Spectrum of S. aureus infections

The origin of SAB, was identifiable in 357/365 episodes: 33% (118/357) were nosocomial, 16% (56/357) healthcare-associated and 51% (183/357) community-acquired. MRSA was responsible for 21% of healthcare-associated and 72% of nosocomial infections. Six MRSA infections were classified as community-acquired with no identifiable healthcare-associated risk factors (one in 2007, one in 2009, three in 2010 and one in 2011). Primary diagnoses for these suspected CA-MRSA cases were bacteraemia without identified source (4), gastroenteritis (1) and superficial skin sepsis (1).

Data on clinical diagnosis, available for 337 episodes of SAB (Table 3), showed that the most common clinical diagnosis was SAB without an identifiable source (33%), followed by SAB with pneumonia ± empyema (22%), SAB with skin and soft tissue infections (17%) and SAB with bone or joint infections (12%). Bloodstream infections without a source were more likely to be MRSA infections, while pulmonary, bone and joint infections were more likely to be due to MSSA. The only two bone or joint infections caused by MRSA were nosocomial in origin. Central venous catheter-related line sepsis was equally likely to be attributable to MSSA or MRSA, although all were nosocomial in origin (Table 3).

Risk factors for MRSA infection

Univariate analysis showed that prior hospitalization within one year, presence of an indwelling medical device, previous MRSA infection, being a resident of a long-term care facility, HIV infection, malnutrition, congenital cardiac disease, concurrent tuberculosis, duration of incident hospitalization and infancy (under 12 months), were significant risk factors for MRSA infection (Table 4). On multivariate analysis, infancy, resident of a long-term care facility, malnutrition, and longer duration of incident hospitalization remained significant risk factors.

Microbiologic profile

Results of antimicrobial susceptibility testing, available for all 365 isolates (Table 5), suggested that most MSSA isolates were susceptible to the majority of antibiotics tested. Only 6% of MSSA were susceptible to penicillin. Most MRSA isolates displayed multi-drug resistance (Table 5). All MRSA were sensitive to vancomycin (minimal inhibitory concentration (MIC) ≤ 2μg/ml); three isolates were reported to have a vancomycin MIC equal to 2μg/ml. All three cases were treated successfully – catheter-related sepsis in a preterm infant with vancomycin alone, pneumonia in an HIV-infected child with vancomycin and clindamycin used sequentially, and bacteraemia without source in a child with short bowel syndrome with a combination of vancomycin and trimethoprim-sulfamethoxazole (TMP/SMX). Compared to HIV-uninfected children, organisms isolated from HIV-infected patients were more likely to be resistant to TMP/SMX (63% vs. 21%) (p=0.00), gentamicin (55% vs. 24%) (p=0.00) and rifampicin (51% vs. 16%) (p=0.00). Of the 32 HIV-infected patients with TMP/SMX resistant isolates, 22 (69%) were on TMP/SMX prophylaxis for Pneumocystis jirovecii at the time of SAB. Antimicrobial resistance patterns were similar in children who were HIV-exposed but uninfected, compared to children who were not exposed to HIV.

Of the six MRSA isolates that were suspected to be community-acquired, antimicrobial susceptibility patterns showed that 4/6 (67%) were resistant to clindamycin and erythromycin, 2/6 (33%) were resistant to TMP/SMX, 3/6 (50%) were resistant to gentamicin, and 2/6 (33%) were resistant to rifampicin. All six isolates were sensitive to vancomycin, ciprofloxacin and fusidic acid.

In 84/365 (23%) of SAB episodes, a concomitant organism was isolated on blood culture. Once typical skin contaminants (coagulase negative staphylococci, Bacillus spp. and Corynebacterium spp.) were excluded, 24/365 cultures (7%) had a clinically significant second organism (polymicrobial bacteraemia) – gram negative bacilli (10), Enterococcus faecalis or faecium (9) and Streptococcus pyogenes or agalactiae (5).. Infants (11%), children with malnutrition (10.5%) and those with MRSA bacteraemia (12.6%) were more likely to have a significant concomitant organism on culture. No association was found between polymicrobial bacteraemia and HIV infection or HIV exposure on chi-square analysis.

Antibiotic management

Data on appropriate antibiotic therapy, which was regarded as the administration of antibiotics to which the organism was sensitive, were available for 322 episodes. At the time the culture was drawn, 31/322 (9.6%) were already on antibiotics to which the organism was sensitive and likely to respond clinically. The majority of these episodes (29/31; 94%) were caused by MSSA and 25/31 (81%) were community-acquired infections. Twelve of these patients received a dose of intramuscular ceftriaxone for suspicion of severe bacterial infection at a primary care clinic prior to transfer to RCWMCH, in keeping with South African Integrated Management of Childhood Illness guidelines [19]. The median time on therapy before the blood culture was drawn was 6 hours (IQR 1.5-22.7 hours).

Excluding those who were already on antibiotics at the time of culture, appropriate antibiotic therapy was instituted in 267/322 (82.9%) episodes of SAB. Time to appropriate therapy was calculated if the precise time of blood draw and time of antibiotic administration were documented. This information was available in 141 episodes. Appropriate therapy was initiated within a median time of 2.5 hours (IQR 0.7-8.5) in MSSA and 46 hours (IQR 4.7-55.3) in MRSA infections (p=0.00). The median time to initiation of therapy was 2.5 hours for community-acquired infections, 5.8 hours for healthcare-associated infections, and 10.6 hours for nosocomial infections (p=0.00). There was no significant difference in time to therapy when comparing episodes with an identified source of bacteraemia and those without.

The remaining 24/322 episodes (7.5%) did not receive appropriate antimicrobials. In 18/24 cases, the S. aureus isolated was clinically regarded to be a contaminant. The remaining six untreated patients died; in 5 cases, death occurred before blood culture results became available while in one case, the MRSA isolated was considered to be a contaminant. Two deaths occurred upon arrival at RCWMCH with pulmonary infections caused by MSSA noted on autopsy. The remaining four deaths were related to MRSA bacteraemia without a source; three of these were also malnourished.

Total treatment duration, which was known for 261 episodes, varied according to the clinical diagnosis. The overall treatment duration was 14 days (IQR 7-29 days). Endocarditis, pulmonary and bone or joint infections were treated for a median of 44 days (IQR 39-51 days), 13 days (IQR 7-23 days) and 40 days (IQR 32-79 days) respectively. Central venous catheter-related infections were treated for 16 days (IQR 13-28 days); 8/11 episodes required line removal due to recurrent isolation of S. aureus. SAB without a source was treated for a median of 9 days (IQR 6-14 days).

Outcomes

According to clinical records, 32 deaths were attributed to SAB, resulting in a case fatality rate of 8.8% over the five year period. Median time to death was 3.5 days (IQR 1.5-6.5 days). SAB without a source accounted for 59% of deaths, while pulmonary infection was responsible for an additional 25% of mortality. There were no deaths attributed to central venous catheter-related infection. MRSA accounted for 53% of deaths. Factors contributing to mortality on univariate analysis included MRSA infection (OR 3.71; CI 1.77-7.66), HIV infection (OR 2.91; CI 1.23-6.87) and malnutrition (OR 2.45; CI 1.17-5.12) (Table 6). An identifiable source of SAB infection (OR 0.3; CI 0.14-0.62), and initiating appropriate antibiotic therapy within eight hours of blood culture, had a protective effect (OR 0.32; CI 0.13-0.77). Age, polymicrobial bacteraemia and the occurrence of multiple positive cultures for S. aureus, were not associated with mortality. On multivariate analysis, only MRSA remained a significant risk factor for mortality (OR 3.76; CI 1.12-12.67).