Follow-up Blood Cultures in Children With Staphylococcus aureus Bacteremia

We conducted a retrospective cohort study of patients with SAB who were hospitalized at any of the 3 hospitals belonging to the Texas Children’s Hospital (TCH) network in Houston, Texas. Cases were identified from an ongoing S aureus surveillance study at TCH.¹²  Patients included in the study were ≤ 18 years of age, with confirmed SAB between January 1, 2018, and December 31, 2018. In the study, we excluded patients with polymicrobial bacteremia or for whom a positive blood culture was considered to be a contaminant by the provider of record. Institutional review board approval was obtained, with patient consent requirements waived.

Medical records were reviewed for all eligible subjects. The following variables were collected: demographic data (age, sex, ethnicity, and race), infectious disease diagnosis based on documentation of the provider of record, patient variables (underlying medical condition, whether the patient was immunocompromised, and vital signs), hospital variables (ICU versus acute care floor), blood culture variables (dates of collection, identities and susceptibilities of isolates, and time to positivity), laboratory data, risk factors for endovascular infection (eg, central line, thrombus on Doppler ultrasound, or vegetation on echocardiogram), treatment variables (time to starting effective antibiotics, empirical and definitive antibiotics used, duration of antibiotics, and, when relevant, time to source control), and clinical outcome (signs of continued infection on follow-up and whether the patient required readmission).

TCH employs the VersaTREK blood culture system (Thermo Fisher Scientific, Waltham, MA), using blood culture vials without resin. Protocols exist at our institution to collect standardized volumes of blood for culture on the basis of patient weight (Supplemental Table 4). For patients with bacteremia, the common practice at TCH is to obtain blood cultures every 24 hours, until sterility is documented.

A patient was considered to have SAB if they had at least 1 blood culture that grew S aureus. Any repeat cultures collected for the purpose of documenting the treatment response and clearance of bacteremia were designated as FUBCs. In calculating the number of FUBCs collected, only cultures drawn ≥24 hours after the previous culture were included. The duration of bacteremia for each patient was described in terms of calendar days and determined by subtracting the date of the first documented positive culture from the date of the first sustained negative culture (which is presumed at 48 hours and officially documented after 5 days of incubation at our institution). For the purposes of this study, only culture results finalized as no growth (no organisms isolated after 5 days of incubation) were regarded as negative. Intermittent positive cultures were defined as any positive culture that both followed and preceded sets of negative FUBC results (with at least a ≥24-hour period between each set of cultures). A patient was determined to have persistent bacteremia if cultures remained positive ≥48 hours after the first positive culture was drawn.¹³  Fever was defined as any temperature ≥100.4°F; patients were considered to have hemodynamic instability for any hypotension and/or temperature-adjusted tachycardia on the basis of age-defined normal values.¹⁴  Source control was considered any debridement of infected tissues, drainage of purulent collections, and/or removal of infected medical devices (eg, central venous catheters). Endocarditis was defined on the basis of the modified Duke criteria.¹⁵ 

Data were abstracted from the medical record into the Health Insurance Portability and Accountability Act–compliant web software Research Electronic Data Capture to create a deidentified database, which was then exported to Microsoft Excel and Stata version 13 (Stata Corp, College Station, TX) for analyses. Categorical variables were compared by using Fisher’s exact test, whereas nonparametric continuous values were analyzed by using the Mann–Whitney U test, Kruskal-Wallis test, and Spearman’s rank correlation coefficient test. A 2-sided P < .05 was considered to be statistically significant. Log-binomial regression was used to estimate the relative risk (RR) and 95% confidence intervals (CIs).

During the study period, 139 patients with SAB were identified. A total of 17 patients were excluded (3 contaminants and 14 polymicrobial), with 122 being included in the study group (Table 1). In total, 53% of patients had medical comorbidities; the most common diagnoses were osteomyelitis (30%) and central line–associated bloodstream infection (CLABSI) (23%). The median duration of bacteremia was 1 day (interquartile range [IQR]: 1–2 days; Fig 1). Only 19 patients (16%) had bacteremia lasting ≥3 days, all of whom had a diagnosis of CLABSI (n = 6), osteomyelitis (n = 10), or endocarditis (n = 3). Within this group, 6 of 19 (32%) had an ultrasound evaluation for thrombus, which was confirmed in 2 of 6 (33%); echocardiogram was performed on 11 of 19 (58%); however, no vegetations were observed in these cases. The duration of bacteremia for our patient sample on the basis of pathogen (MSSA versus MRSA), infectious disease diagnosis, underlying condition, and level of care is summarized in Table 2. An association was observed between the duration of SAB and infectious disease diagnosis. The duration of bacteremia for patients with comorbidities did not differ from that of patients without an underlying condition nor did it differ for patients with MRSA compared with those with MSSA. Similarly, no correlation was seen between the duration of bacteremia and level of care. Furthermore, there was no correlation between the duration of bacteremia and duration of ICU stay (ρ = 0.15; P = .25), duration of bacteremia and fever (ρ = 0.06; P = .51), or duration of bacteremia and hemodynamic instability after accounting for tachycardia due to fever (ρ = 0.04; P = .69). No statistically significant correlation was seen between the duration of bacteremia and time to blood culture positivity (ρ = −0.12; P = .19) or time to initiation of effective antibiotics (ρ = 0.14; P = .11). The only laboratory value that exhibited a correlation with duration of bacteremia was the initial C-reactive protein level (ρ = 0.21; P = .05).

FUBC data are summarized in Figures 1 and 2. A median of 2 negative FUBC results (IQR: 2–3 cultures; Fig 2) were drawn to document sterility. In 48 of 122 (39%) cases of SAB, ≥3 FUBCs were collected to confirm infection clearance. In this group, 32 of 48 (67%) blood cultures were collected despite resolution of signs or symptoms of infection (fever, tachycardia, hypotension, or provider documentation of clinical instability), whereas 16 of 48 (33%) were collected in the setting of hemodynamic instability and/or persistent fever.

Intermittent positive cultures were observed for 6 of 122 patients (4.9%; Table 3), all of whom had a diagnosis of endocarditis (2 of 6; 33%) or osteomyelitis (4 of 6; 67%). Of these patients, only 2 had underlying medical comorbidities: a 22-month-old infant with Canavan disease and a 1-week-old infant born at 27 weeks’ gestation. Only 1 patient (0.8%; in a case of osteomyelitis) had a positive blood culture after 2 documented negative cultures. On the day that cultures reverted back to positive, this patient was transferred to the ICU from the acute care floor because of concern for septic shock. Patients with intermittent positive blood cultures had a longer time to initiation of effective antibiotics (median: 14.3 and IQR: 6.8–20 hours versus median: 2.6 and IQR: 0.9–9.2 hours; P = .03). No other variables studied (including age, sex, presence of an underlying condition, and level of care) differentiated this group from patients with persistently negative FUBCs. Among the 4 patients with osteomyelitis in this group, 2 had surgical debridement performed >24 hours after initial blood cultures were drawn (3 and 10 days, respectively), and the other 2 required multiple debridements to achieve source control.

All 122 patients were started on empirical antimicrobial coverage appropriate for the S aureus subsequently isolated from the first set of positive blood culture (73% vancomycin, 11% clindamycin, 9% cephalosporin, and 7% nafcillin). All patients received appropriate doses of antimicrobial agents. Patients for whom the infectious diseases service was consulted tended to receive a longer duration of intravenous antibiotics, particularly patients with CLABSI (a median treatment of 14 vs 10 days for those without infectious diseases consultation; P = .007). Although the majority of patients completed a full course of intravenous antibiotics in hospital, 45 of 122 (37%) were transitioned to oral antibiotics at discharge, with cephalexin as the antibiotic of choice for MSSA (33 of 35; 94%) and clindamycin for MRSA (9 of 10; 90%).

A central line was present in 45 of 122 (37%) patients and was identified by the provider of record as the source of infection in 28 of those patients (62%). From this group, the line was pulled in 26 patients (93%). The median number of days from when the first positive blood cultures were drawn to when the line was pulled was 2 (IQR: 1–3 days). The time to central line removal did not have any significant correlation with the duration of bacteremia (ρ = 0.15; P = .47). Surgical debridement was performed for 25 patients. The median duration from the time first bloods were drawn to the first surgical debridement was 1 day (IQR: 0–1 days). A second debridement was performed for 9 of 25 (36%) of the patients, the majority of whom had osteomyelitis; 1 of the patients with osteomyelitis, ultimately, required a total of 5 debridement procedures. These additional procedures were performed at a median of 4 days (IQR: 2–6 days) after the first surgical debridement, in most cases because of persistent fevers and/or limited clinical improvement. No significant correlation was seen between the time to debridement and duration of bacteremia (ρ = 0.18; P = .38).

Follow-up data were available for 107 of 122 (88%) patients. A total of 4 of these patients (3.7%) required readmission within 30 days from discharge because of complications related to their infection. One of these patients, who initially had MSSA CLABSI, was readmitted for mechanical issues with the line but had no recurrence of bacteremia. Another patient was a bone marrow transplant recipient who had recurrent MRSA bacteremia, despite having the original central line removed and having no endovascular nidus of infection identified on imaging. The remaining 2 patients underwent readmission for management of persistent MSSA osteomyelitis, 1 of whom required further debridement and the other of whom did not adhere to their prescribed antibiotics. An additional patient with MRSA osteomyelitis who did not require readmission developed persistent functional limitations; otherwise, no irreversible morbidities related to SAB were observed. No deaths occurred from SAB.

In our sample of 122 children with SAB, the majority of cases had a limited duration of bacteremia (<48 hours), with minimal to no complications. Bacteremia lasting ≥3 days was, exclusively, seen in patients with CLABSI, endocarditis, and osteomyelitis, highlighting that patients with endovascular and closed-space infections are at an increased risk of persistent bacteremia and merit more conservative management in terms of clinical monitoring and follow-up.^7,8,11  Interestingly, although a correlation was seen between an infectious disease diagnosis and the duration of bacteremia, in our study, we did not find any increased risk for persistent SAB based on an underlying medical condition, including immunosuppression.

Intermittent positive cultures were seen in <5% of patients, with repeat positive cultures after 2 negative FUBC results documented in <1% of patients. With these findings, it is suggested that for most patients, 2 days of negative blood cultures are sufficient to document sterility, particularly when clinical improvement has been observed. Furthermore, with our analysis, it was indicated that patients with intermittent positive cultures are at more than twice the risk of having a diagnosis of endocarditis or osteomyelitis, compared with patients without intermittent positive cultures (P = .004). This observation is consistent with the results of another pediatric study, in which researchers found that musculoskeletal and endovascular infections due to MRSA bacteremia tended to be associated with a higher rate of treatment failure, possibly because of an increased bacterial burden or decreased antibiotic penetration at these sites.⁷  Although the number of patients with intermittent positive cultures was small, our data suggest that source control could be a critical variable in terms of the risk for intermittent positive cultures: for every such patient in the osteomyelitis group, surgical debridement occurred >24 hours after the initial blood cultures were drawn or a single debridement was clinically inadequate. In comparison, for the patients with FUBC results that remained negative, only 2 of 33 (6%) had debridement >24 hours after the initial blood cultures were drawn, and only 6 of 33 (18%) required >1 debridement.

Our results also suggest that intermittent positive cultures may occur less frequently in children, compared with adults. In a recent adult study, intermittent positive cultures were documented in 13% (140 of 1071) of episodes of SAB, with ≥2 days of negative blood cultures preceding a repeat positive culture in 4% (44 of 1071).¹⁰  Additionally, in our findings, the 30-day readmission rate for complications due to SAB is placed much lower for the pediatric population than it is for adults (3.7% vs 22%).⁶ 

Our study has limitations, many of which are inherent to its retrospective design. First, 74 of 122 (61%) of the patients in our study had ≤2 FUBCs collected, and 26 of 122 (21%) had ≤1. As such, we are basing our conclusions on the presumption that those 74 patients would not have had any further positive cultures if additional cultures had been drawn. Furthermore, for the patients who had ≥3 negative FUBC results collected, the indications for drawing additional cultures were not always consistently documented in the medical record. For these cases, we had to infer on the basis of what information was available from the provider notes and/or patient vital signs. By relying on provider documentation, cases of bacteremia without an identified source were also likely overrepresented in our patient sample, especially given a previous study, in which researchers suggested that bacteremia without a focus constitutes only 5% of cases of pediatric SAB.¹⁶  Additionally, because of the retrospective nature of our study, our ability to collect any standardized follow-up metrics on our patient sample was limited.

Several characteristics of our patient sample were also likely to affect the quality of our study. This sample population contained a large number of patients with an underlying medical condition as well as premature infants, which may have influenced multiple variables, including data on hemodynamic instability and decisions to collect additional FUBCs. Our subset of patients with intermittent positive FUBCs was small (n = 6), which restricts our ability to make comprehensive analyses of risk factors for intermittent positive FUBC results. In particular, our study was, likely, underpowered to appreciate the impact of source control on bacteremia duration. Finally, although this study offers recommendations for the management of patients with SAB in regard to the collection of FUBCs, we have not demonstrated that this knowledge will improve patient outcomes in terms of treatment duration, timing of source control, medical costs, and length of stay, and, given that our study was conducted in a single center, it may not be broadly applicable to all institutions treating such patients.

The aim of our study was to determine if specific recommendations can be made on the number of FUBCs needed to demonstrate clearance of SAB in the pediatric population. Within our patient sample of 122 children, the frequency of having an intermittent positive culture result after 2 documented negatives was <1%, with the only documented case being a patient with ongoing hemodynamic instability and delayed source control. On the basis of these findings, additional blood cultures to document sterility are not necessary after 2 FUBC results are negative in well-appearing patients. To guide the implementation of this recommendation, if the second FUBC is negative at 24 hours, with the first FUBC negative at 48 hours, we do not recommend collection of a third FUBC unless the patient has ongoing hemodynamic instability. If we were to apply these conclusions to our patient sample, unnecessary FUBCs could have been prevented in 32 patients (26% of our total sample). Additionally, given that 13 of 19 (68%) patients with bacteremia lasting ≥3 days had a diagnosis of either osteomyelitis or endocarditis, we recommend a thorough clinical evaluation for these infection sources in any patients with SAB lasting ≥3 days for whom an infection source has yet to be identified. Further research on this topic is needed to determine the cost reduction from eliminating unnecessary blood cultures and perform a powered analysis of whether complications would increase as a result of these recommendations.

CI

confidence interval

CLABSI

central line–associated bloodstream infection

FUBC

follow-up blood culture

IQR

interquartile range

MRSA

methicillin-resistant Staphylococcus aureus

MSSA

methicillin-susceptible Staphylococcus aureus

RR

relative risk

SAB

Staphylococcus aureus bacteremia

TCH

Texas Children’s Hospital