Background

Staphylococcus aureus bacteremia is a serious illness with potential for adverse outcomes. The rate of this infection in children ranges from 1.5 to 3.5 per 1000 hospitalizations.1,2 Among pediatric patients, S. aureus bacteremia poses risk for significant morbidity, including metastatic infection and prolonged hospitalization.1,2 Moreover, the mortality incidence in children with S. aureus bacteremia ranges from 2 to 6%,1,3,4,5 and is as high as 15% in patients with cardiac conditions.6

Despite the morbidity and mortality associated with pediatric S. aureus bacteremia, factors driving outcomes remain understudied. In contrast, modifiable risk factors are better understood for adults with S. aureus bacteremia, for whom fatality rates range from 15 to 25%.7,8,9 An important factor impacting outcomes in adults with S. aureus bacteremia is consultation by an infectious diseases (ID) specialist, which is associated with a 50% reduction in morbidity and mortality.7,8,10,11 In adult populations, thorough diagnostic evaluation, targeted antibiotic treatment, and an appropriate length of antibiotic therapy have been demonstrated to improve clinical outcomes for patients with S. aureus bacteremia.7,9,11,12 Hence, several transdisciplinary, multinational groups of clinicians have derived quality-of-care measures for the management of S. aureus bacteremia. These quality-of-care measures focus on diagnostics, monitoring, and treatment and are associated with improved outcomes (Table 1).11,13 In adult patients, ID consultation confers a greater likelihood of management adhering to these quality-of-care indicators (QCIs).7,9,11,12

Table 1 Quality-of-care indicator definitions.
Full size table

In the pediatric population, data are lacking regarding the best management strategies for children with S. aureus bacteremia, including the impact of ID consultation. Two single-institution studies of children with S. aureus bacteremia conducted in Australia (n = 100) and the United Kingdom (n = 63) evaluated the impact of ID consultation.4,14 Both studies demonstrated that ID consultation was associated with improved diagnostics, laboratory monitoring, and targeted antibiotic therapy; the larger study demonstrated reduced mortality in children receiving ID consultation.4,14 An additional multicenter study, also conducted in Australia, provided further evidence that ID consultation was associated with reduced mortality.15 While these results suggest there are multiple benefits from ID consultation, these studies may have limited generalizability due to innate differences in geography, epidemiology, and medical systems.

The objective of this study was to investigate the impact of ID consultation on adherence to six established QCIs in US children with S. aureus bacteremia. We also sought to estimate the effect of ID consultation on treatment failure, defined as a composite of all-cause mortality or hospital readmission within 90 days. Importantly, we used propensity score methodology (underutilized in prior studies) in the treatment failure analysis to account for potential confounding by severity of illness and other important factors.

Methods

Setting and patients

We conducted a cohort study comprised of 306 patients aged 0–24 years with S. aureus bacteremia hospitalized at St. Louis Children’s Hospital (SLCH), a 402-bed tertiary care center, from January 2011 to May 2018. Children with one or more blood cultures positive for S. aureus for which antibiotic treatment was administered were eligible. Patients with positive blood cultures for which antibiotics were not prescribed (per provider notes and laboratory comments) were excluded. In August 2016, SLCH instituted a formal handshake stewardship antibiotic stewardship program. This study was approved by the Washington University institutional review board.

Data collection

Electronic medical records were reviewed to abstract demographic and clinical factors including primary clinical service; ID consultation; clinical diagnosis; blood culture results; S. aureus susceptibility (methicillin-susceptible or -resistant); location of infection onset (community or hospital); antibiotic administration; vital signs; imaging and laboratory studies; presence of a central venous catheter or instrumentation; surgical procedures for source control; ICU admission; ventilator, inotropic, or extracorporeal membrane oxygenation support; and outcomes.

Exposures and outcomes

Exposures

The primary exposure was the receipt of timely ID consultation, defined as a formal ID consultation within 6 days of collection of the initial positive blood culture. In clinical practice, the large majority of ID consultations occur within this 6-day time period. Our secondary exposure was the receipt of targeted antibiotic therapy (i.e., the antibiotic administered while the patient was bacteremic). The definition of targeted therapy was dependent on S. aureus susceptibility: for MSSA, targeted therapy included cefazolin, nafcillin, or oxacillin. For MRSA, targeted therapy included vancomycin, ceftaroline, or daptomycin.16,17

Outcomes

We assessed the adherence to six factors previously established as QCIs in patients diagnosed with S. aureus bacteremia (Table 1) to test the hypothesis that ID consultation would improve adherence to these QCIs.11,13 We also tested the hypothesis that ID consultation is associated with improved clinical outcomes in pediatric patients with S. aureus bacteremia. Specifically, we used the clinical outcome of treatment failure, defined as a composite of 90-day all-cause mortality or 90-day all-cause hospital readmission. In addition, we tested whether exposure to targeted antibiotic therapy, based on S. aureus susceptibility, reduced treatment failure.

Other definitions

Blood culture proof-of-cure was defined as two consecutive negative blood cultures following a positive blood culture.3 Complicated bacteremia was defined as the duration of bacteremia ≥3 days, fever >72 h, metastatic disease (defined as a deep, distal, or secondary17 infection, anatomically unrelated to the primary site of infection), or endocarditis.17 Source control was defined as a composite of surgical intervention, central venous catheter removal, or instrumentation removal, when applicable. Community infection was defined as a positive blood culture for S. aureus obtained within the first 48 h of hospital admission (see Supplementary methods for further detail).

Statistical analysis

We compared the distribution of adherence to QCIs and treatment failure by exposure groups. Differences were assessed using χ2 tests for categorical variables, and independent sample t-tests or Mann–Whitney U tests for continuous variables. Differences in proportions of ID consultation by year of hospital admission were performed using χ2 for trend analysis. P values ≤0.05 were considered significant.

To examine the relationship between ID consultation and treatment failure, we used Cox proportional hazards models to estimate unadjusted and propensity score-weighted hazard ratios (HRs). Standardized mortality ratio (SMR) weights were calculated from a propensity score using multivariable logistic regression, where the propensity score represented the probability that a patient received ID consultation versus no ID consultation, accounting for baseline covariates selected a priori based on expert knowledge (age, race, sex, primary service, year of admission, staphylococcal susceptibility, infection entity, and targeted antibiotic therapy, Supplementary Table 1).

We used similar methods to examine the relationship between receipt of targeted antibiotic therapy and treatment failure. Specifically, SMR weights were calculated from two separate propensity scores representing the probability that targeted antibiotic therapy was initiated (1) –1 to 2 days in relation to the index date (defined here as the date the first positive blood culture was drawn) versus 3–6 days from the index date; or (2) –1 to 2 days versus no initiation of targeted antibiotic therapy between 0 and 6 days from the index date, accounting for baseline covariates (age, race, sex, primary service, year of admission, ID consult performed, comorbidities, infection entity, and infection onset).

SMR weights were computed from propensity scores, equal to 1 for children in the reference group and equal to the propensity odds (i.e., propensity score/[1 − propensity score]) for children in the other group. SMR weights allowed us to standardize the covariate distribution in each comparator cohort to that in the reference cohort. Thus, under the assumption of no unmeasured confounding, our exposure comparisons would be unconfounded with respect to their effect on the outcome.

We calculated absolute standardized mean differences of baseline covariates in the unweighted and SMR-weighted populations to determine whether weighting the population reduced imbalances of observed covariates and made the treatment groups more exchangeable (defined by absolute standardized mean difference <0.10; Supplementary Tables 2 and 3).18 We used robust variance estimators to calculate 95% confidence intervals (CIs). We verified the proportional hazards assumption with an interaction term between (log) time and treatment. In the analysis comparing patients with and without timely ID consultation, patients categorized as no timely ID consultation (i.e., no ID consultation within 6 days of collection of the initial positive blood culture) were censored upon receipt of late ID consultation (i.e., ID consultation occurring 7 days or more after blood draw). In the analysis comparing patients for whom targeted antibiotic therapy was initiated –1 to 2 days versus no antibiotic therapy, patients were censored upon receipt of antibiotics after day 2.

Results

Of 306 patients hospitalized with S. aureus bacteremia from January 2011 to May 2018, 193 (63%) received ID consultation. The median time from the date of collection of the initial positive blood culture to ID consultation was 3 days for patients receiving ID consultation (interquartile range [IQR] 0, 3). Patient sex and race did not differ between patients who did or did not receive ID consultation (Table 2); however, children receiving ID consultation were older (mean 7.3 ± 5.9 years) compared to those not receiving consultation (5.3 ± 6.3 years; P = 0.005). Seventy-three percent of children receiving ID consultation had a community-onset infection compared to 58% without ID consultation (P = 0.005). While 84% of patients in the general pediatric and surgery services received ID consultation, this was less frequent for children in the PICU (63%), NICU (43%), and in the hematology/oncology service (20%; P = 0.001).

Table 2 Patient characteristics by receipt of infectious diseases consultation.
Full size table

The proportion of patients with S. aureus bacteremia who received ID consultation increased between 2011 and 2018 from 50 to 90% (χ2 for trend, P = 0.001, Fig. 1). ID consultation was more likely for patients with musculoskeletal infection (93%) and endovascular foci (94%) and less likely for patients with central venous catheter-associated infections (46%) and in cases of S. aureus bacteremia with no identifiable source (33%). Of the children receiving ID consultation, 72% were diagnosed with complicated bacteremia versus 36% of those without ID consultation (P < 0.001); 49% of the ID consultation group had underlying conditions versus 87% of those without ID consultation (P < 0.001; Table 2).

Fig. 1: Children with Staphylococcus aureus bacteremia receiving pediatric infectious diseases consultation by the year.
figure 1

The year 2018 includes only January through May. A formal handshake antibiotic stewardship program was initiated in August 2016.

Full size image

Children receiving ID consultation were more likely to have fulfilled all 6 QCIs when compared to the group without ID consultation (Fig. 2a), including proof-of-cure blood cultures (95% with consultation versus 84% without; P = 0.002), all indicated laboratory studies (46% versus 8%; P < 0.001 [Fig. 2b]), echocardiography when indicated (e.g., structural heart anomalies or complicated bacteremia; 32% versus 15%, P = 0.03), source control when indicated (e.g., removal of infected central venous catheter or instrumentation; 57% versus 28%; P < 0.001), targeted antibiotic therapy (92 versus 71%; P < 0.001), and appropriate duration of targeted antibiotic therapy (evaluated in patients with bacteremia without a focus, median [IQR] 14 days [2–16] versus 9 [5–23], P = 0.004 [Table 2]). We also assessed the relationship between QCI fulfillment and treatment outcomes (Table 3). Treatment success was associated with proof-of-cure blood cultures (93% versus 87% in those with treatment failure, P = 0.05) and obtaining all indicated laboratory studies (39% versus 20% in failure; P = 0.001).

Fig. 2: ID consultation results in QCI fulfillment.
figure 2

a Adherence to quality care indicators by receipt of pediatric infectious diseases consultation. b Individual indicated laboratory studies by receipt of pediatric infectious diseases consultation. ID infectious diseases, QCI quality-of-care indicators. Proof-of-cure was considered two negative blood cultures following a positive blood culture. The indicated laboratory studies included CBC, ESR, and CRP for all patients and vancomycin trough and creatinine for children who received at least 3 doses, or 2 days, of vancomycin (N = 248). An echocardiogram was indicated if the patient had complicated bacteremia (defined as the duration of bacteremia ≥3 days, fever >72 h, metastatic disease, or endocarditis) or presence of structural heart condition (N = 174). Source control was a composite of a surgical procedure to debride the source of infection (e.g., osteomyelitis) or removal of an infected central venous catheter or instrumentation (N = 172). Targeted antibiotic therapy was defined for MSSA as nafcillin, oxacillin, or cefazolin and for MRSA bacteremia as vancomycin, ceftaroline, or daptomycin.

Full size image
Table 3 Adherence to quality-of-care indicators and other measures by treatment success.
Full size table

Patients receiving ID consultation were more likely to receive targeted antibiotic therapy (92%) compared to those without ID consultation (71%; P < 0.001, Fig. 2a). The composite outcome of treatment failure was significantly higher in patients not receiving ID consultation (46%) compared to those receiving consultation (31%). When performing propensity score analysis our unweighted, crude model reflected an increased HR (1.56; 95% CI 1.03–2.38) for children who did not receive consultation.

Univariate analysis of factors associated with treatment failure demonstrated no significant difference in demographic factors or S. aureus susceptibility (Table 3). Patients with treatment failure were more likely to have hospital-onset infection (40%, versus 29% of those with treatment success; P = 0.05), central venous catheter-associated infections (31% versus 20%; P = 0.03), underlying conditions (87% versus 50%; P < 0.001), and to require ICU-level care (63% versus 43%; P = 0.001).

Propensity score-weighted analysis

We observed similar weighted risks of treatment failure between children who did, versus those who did not, receive ID consultation (HR, 1.03; 95% CI 0.69–1.53) after accounting for age, race, year of admission, sex, staphylococcal susceptibility, infection entity, targeted antibiotic therapy, and primary service (Table 4). In considering the timing of initiation of targeted antibiotic therapy, we accounted for age, sex, race, ID consultation, primary service, comorbidities, infection onset, infection entity, and year of admission. Similar weighted risks of treatment failure were observed between patients for whom targeted antibiotic therapy was initiated –1 to 2 days in relation to the index date (date the first positive blood culture was collected) and those for whom targeted therapy was initiated 3–6 days after the index date (HR, 1.05; 95% CI 0.59–1.87). However, the number of events was small and risk estimates were imprecise.

Table 4 Propensity score-weighted hazard ratio estimates for composite outcome of 90-day hospital readmission or mortality.
Full size table

Discussion

ID consultation has been demonstrated to be an important part of the management of adults and children with S. aureus bacteremia. This study is the first to demonstrate that ID consultation resulted in a greater likelihood of adherence to multinational derived QCIs leading to appropriate diagnostics, monitoring, and treatment, factors that have been associated with fewer rehospitalizations and deaths.11,13

We also estimated the effect of ID consultation on the risk of treatment failure while accounting for potential confounding variables such as the severity of illness, through the use of propensity score analysis. Crude results showed modest benefit of ID consultation; however, weighted results did not demonstrate that ID consultation was associated with decreased treatment failure. Effect estimates for these analyses were imprecise given the small numbers of events in our single-institution study. Furthermore, this study included both community-onset and hospital-onset infections which may differ in their potential for adverse outcomes. Future multicenter studies are needed to investigate this research question.

Although QCIs have not been ascertained for pediatric patients, we adapted 6 QCIs established for the adult population derived by Hadano et al. to assess diagnostic evaluation and treatment practices performed among children with S. aureus bacteremia.11 In a multicenter study of children with S. aureus bacteremia, each day of bacteremia was associated with 50% increased odds of sequelae, including hematogenous complications or metastatic foci.19 Thus, obtaining daily blood cultures to monitor response to antibiotic therapy, and ultimately demonstrating clearance of bacteremia, is imperative to reduce this risk. Moreover, as prolonged bacteremia may reflect endocarditis or another focus of infection, investigation of the source of prolonged bacteremia, including echocardiogram or other imaging, is necessary to ensure sources are identified and controlled. In the present study, the length of bacteremia in patients receiving ID consultation was longer than in children without ID consultation, which may be reflective of a delay in further diagnostic evaluation until consultation was requested. Children receiving ID consultation were more likely to undergo the indicated monitoring laboratory studies as well as proof-of-cure blood cultures, echocardiography, source control, and targeted antibiotic therapy. Proof-of-cure blood cultures and monitoring laboratory studies were, in turn, associated with a reduction in treatment failure.

ID consultation has been demonstrated to improve morbidity and mortality among adult patients with S. aureus bacteremia.7 Three prior studies, two in Australia and one in the United Kingdom, evaluated the impact of ID consultation for pediatric patients with S. aureus bacteremia.4,14,15 Despite differences in healthcare systems, cohort size, study design, and statistical methodology, there are common findings among these studies and the present study. ID consultation was less frequent among patients cared for by ICU services as well as the hematology/oncology service. Moreover, patients with hospital-onset infections and those with central-line-associated infections were also less likely to receive ID consultation. Lastly, patients not receiving ID consultation were more likely to be diagnosed with bacteremia without source, potentially reflecting an incomplete diagnostic workup. Across studies, ID consultation improved diagnostic evaluation and management, including echocardiography and proof-of-cure blood cultures, identified focus of infection and/or optimized source control, and improved choice and duration of directed antibiotic therapy. Among both Australian cohorts, children receiving ID consultation had a reduced mortality or had a higher likelihood of cure, defined as the absence of death or documented bacteremia relapse at 30 days. Importantly, in the present study, children who were diagnosed with hospital-onset infections, needed ICU-level care, had a central venous catheter-associated infection, and had underlying conditions (e.g., malignancy) were all more likely to suffer treatment failure, highlighting the need for collaboration between ID and other subspecialty services.

Among pediatric patients, mortality due to S. aureus bacteremia is fortunately uncommon; in the present cohort, only six deaths (2%) were attributable to S. aureus bacteremia. Thus, specifying appropriate outcome measures is an important consideration when evaluating the impact of a component of care, such as ID consultation. As reflected in this study, QCIs may represent the best markers of improved care. Indeed, patients with proof-of-cure blood cultures and monitoring laboratory studies had a decreased likelihood of treatment failure. To our knowledge, these findings are unique and bring to light the value of monitoring objective labs in critically ill children. In particular, patients for whom CRP and ESR were measured had improved outcomes. Monitoring these objective inflammatory markers may signal the need for further source control or other intervention, and may guide antibiotic management, which may in turn contribute to better outcomes.

The present investigation used propensity score modeling in an attempt to control for the severity of illness and thus minimize potential confounding bias. We undertook this method with the hypothesis that ID physicians were more likely to be consulted for children with higher severity of illness, a factor that may be associated with an increased likelihood of treatment failure. Although our unadjusted analysis demonstrated benefit with ID consultation, our propensity score analysis demonstrated a similar incidence of treatment failure between groups. However, our analyses are complicated by several issues. The non-randomized nature of the exposure allows for the possibility of unmeasured confounders, such as those related to the severity of illness, due in part to the complex nature of the study with a variety of human elements, which may have biased the results to the null.

Given the potential consequences of S. aureus bacteremia, measures to improve management, and thus outcomes, are urgently needed. Lloyd et. al. conducted a quasi-experimental study to assess an electronic health record Best Practice Advisory (BPA) that recommended ID consultation and targeted antibiotic therapy (based on MRSA or MSSA) for all patients with S. aureus bacteremia.20 One year following BPA implementation, 100% of patients with S. aureus bacteremia received ID consultation. Moreover, time to targeted antibiotic therapy significantly decreased following the intervention. Previous studies have demonstrated that the implementation of antimicrobial stewardship programs (ASPs) increases ID consultation.21 Indeed, at our institution, a formal ASP was implemented in August of 2016, after which over 90% of children with S. aureus bacteremia received ID consultation. Lastly, additional measures to implement quality care measures and evidence-based practices include institutional clinical practice guidelines and standardized order sets. These interventions can be especially valuable in rural and other underserved areas where access to specialists is limited.22

This large pediatric study quantifying the impact of ID consultation for patients with S. aureus bacteremia is unique in its use of propensity score methodology to account for clinical and demographic differences (e.g., severity of illness) between exposure groups. This study also has several limitations. While this is a large pediatric cohort, our sample size (N = 306) was smaller than optimal for propensity score methodology, resulting in imprecise effect estimates, demonstrated by the wide CIs. Data from adult studies, several pediatric studies, and now this study, support ID consultation. A mortality benefit in these prior studies and an improvement in recognized quality indicators for optimal outcomes support ID consultations for children with S. aureus bacteremia. Additional analyses are needed that incorporate larger patient populations to further support these recommendations. Furthermore, our single-center study may not be fully generalizable regarding clinical practice and patient outcomes.

In conclusion, we found that ID consultation improved diagnostics, monitoring, and treatment, which were associated with improved outcomes in pediatric patients with S. aureus bacteremia. Additional strategies such as improved collaboration between subspecialty teams and development of multispecialty clinical practice guidelines can improve clinical management of patients with S. aureus bacteremia, limit complications, and minimize morbidity and mortality.