In this single-centre pilot study, we classified SAB in children into broad clinical categories to aid the paediatrician in thinking about their patients, and shown significant differences between these groups, including CRP max, duration of hospitalisation, total number and duration of antibiotics. This study is the first to our knowledge to categorize SAB in children according to complexity, which is an essential first step when considering clinical trials in both simple (e.g. the role of short course therapy) and complex SAB (e.g. the role of combination therapy).
It is clearly demonstrated that there is a high degree of variability in SAB phenotype, and an according variability in management, with simple infections receiving median 34 antibiotic days and 10 hospital days, compared to 43 and 35 days respectively for complex infections. The extended duration of antibiotics is consistent with skeletal infection comprising almost two-thirds of both groups, for which the recommended duration of antibiotics is 3–4 weeks .
Whilst SAB severity and mortality risks have been well defined in adults , there are limited data to guide classification in children. Indeed, the a priori definition of complexity was determined by clinician experience, and then confirmed using markers such as CRP max, duration of hospital stay and treatment. This is important as children with SAB rarely die, making death a non-discriminatory outcome for clinical trials. Mortality in children is much lower than adults with SAB, with rates of around 5% [4, 16,17,18] vs 20%  respectively, reported. Our small cohort had no deaths attributed to SAB. Therefore, we used surrogate outcome markers of complexity including treatment duration, length of stay and CRP max to test our phenotypic definition of simple versus complex infections, each of which proved significant. Larger multi-centred studies are needed to confirm these findings and increase our knowledge of factors associated with morbidity and mortality in children.
By the time SAB is confirmed at 48 h, the CRP max may help identify children with simple (CRP max < 75 mg/L) and complex (CRP max> = 75 mg/L) infections. This time point is important in paediatrics, as prior studies of bone and joint infections confirmed the efficacy of 3 days of intravenous antibiotics followed by step down to oral antibiotics . The ability to use CRP to differentiate at 48 h between simple and complex SAB gives the clinician some reassurance that in a patient with a CRP below 75 mg/L, step down to oral antibiotics may be a safe option. This finding would need to be supported by further, prospective studies with lager patient populations, however, as finding early predictors of ultimate clinical outcome is challenging in SAB we have reported this finding to guide clinicians in their clinical practice.
The difference in antibiotic use between groups was significant both in number and duration. The treatments were generally longer than expected, particularly for simple SAB. In addition, a large number of different antibiotics were prescribed during the course of treatment in both groups. This finding may be explained by the gradual detection over the first few days of hospitalisation of different foci of infection. However, even in the simple group the number of antibiotic switches was surprising. Multiple antibiotic exposures may result in increasing antimicrobial resistance and cost. The numbers of antibiotics may be excessive, and this was during a period prior to the initiation of guidelines for empiric treatment of SAB, introduced as part of the antimicrobial stewardship bundle that commenced in November 2013. Furthermore, ID consults for SAB were added as a mandatory activity in July 2015 and all hospital-acquired SAB were investigated with a root cause analysis from January 2015. In the current environment of global fears regarding antibiotic resistance , any trend towards over-use of antimicrobial therapy is worth highlighting, and positions infectious disease specialists ideally to help target therapy more effectively.
In Australia, Indigenous children have been reported to have increased rates, but not worse outcomes from SAB . We were not able to detect differences in Indigenous status between simple and complex SAB. This may be due to small numbers, however the duration of the study capturing children throughout the whole state (which has the second highest proportion of Indigenous children at 6.8% (www.abs.gov.au, last accessed 26 September 2016)), suggests that this was a representative sample. This is reassuring in supporting the evidence that outcomes for Indigenous children in Australia are not worse than other ethnic groups for SAB, despite their rates being slightly higher .
Despite widespread use of vancomycin amongst this small cohort, the 10% MRSA rate was lower than reported elsewhere. Other studies have shown between 15 and 34% of methicillin resistance in similar populations [17, 18], and 13.2% nationally, albeit with a gradient from north (higher) to south (lower) Australia . Whilst MRSA has been shown to be associated with increased mortality and nosocomial infections in children [13, 18] we had a high proportion of children with complex SAB infections despite our low rate of MRSA isolates, and MRSA was not over-represented in the complex group.
Malignancy, a risk factor for SAB in adults , was the only significant risk factor at baseline for complex infections. Children with malignancy have a higher use of central venous access devices resulting in central line associated blood stream infection [21, 22] and have significant iatrogenic immunosuppression  making this observation unsurprising. Our cohort showed relatively low rates of hospital-acquired (10.2%) or healthcare-associated (14.3%) infections when compared to 29% and 20% respectively in other studies . These groups combined together were not found to be associated with complex infections and none were MRSA.
Despite a large number of patients with a medical device in situ at the time of SAB (20.4%), we had low rates of device-focussed SAB (12.2%). This is consistent with other studies which have shown rates of venous catheter associated infections between 3.3–11% [17, 18, 23], although rates of up to 60% have been reported .
Bone and joint infection was the most common focus affecting almost 40% of the children. This is comparable to rates of 12–37% in other studies [4, 18, 23]. Whilst 18.4% of our cohort had no focus for their infection, other articles have reported rates as high as 35% [13, 23], making this the most common cause of SAB in their studies. The relatively high proportion of patients with no focus is significant as some studies have shown increased mortality in this group [4, 13]. As such, we had categorized this phenotype of infection as complex. It should be noted that it is not clear whether the increased mortality rate is a true reflection of the severity of SAB with no focus, or due to other associated factors such as MRSA infections, an occult source, or delays in treatment. We note patients with malignancy made up a significant proportion of our patients with SAB with no focus (n = 7, 77.8%), many of whom had a medical device in situ as a possible source of infection not clinically determined or recorded.
There are several limitations to our study. This small, pilot study was conducted retrospectively from patient files and electronic records with only 49 cases identified. Due to this low number, it is not possible to draw definite conclusions about some groups, including Indigenous patients and MRSA bacteraemia. As the study was retrospective, we were unable to type the S. aureus isolates. Due to the paucity of pre-existing data, we defined our cohorts based on phenotypic characteristics rather than performing regression analysis. The significant difference of outcome measures across the two groups gives weight to this definition as being clinically appropriate. In addition, clinicians are interested in defining these groups early in the admission – all of our points of significance are recorded later in the admission, with the exception of CRP max being at median 2 days in complex patients. In this pilot study, we were exploring our ability to differentiate between simple and complex SAB, a distinction that we are familiar with clinically, but do not as yet have the algorithms or diagnostic tools to describe. In this exploratory, retrospective audit, we did not seek ethical approval to contact patients to ascertain long term outcomes which would be a point of interest for future studies. We also did not collect data on combination therapy, including the use of lincosamides alongside beta lactams, in cases of complex SAB.