Variation in Antibiotic Selection and Clinical Outcomes in Infants <60 Days Hospitalized With Skin and Soft Tissue Infections

Study Design and Data Source

We conducted a multicenter, retrospective cohort study of infants <60 days of age with a diagnosis of SSTI. We extracted data from the Pediatric Health Information System (PHIS), an administrative database of 45 free-standing pediatric hospitals in the United States that are affiliated with the Children’s Hospital Association (Lenexa, KS). Patient data are deidentified within PHIS; however, encryption of patient identifiers allows for tracking of individual patients across visits. The current study included data from a total of 36 hospitals, with 9 hospitals excluded for lack of emergency department (ED) data or low hospital volumes (<10 cases). This study was reviewed and approved by the local institutional review board.

Study Population

Antibiotic Classification

Empirical antibiotic selection was defined as antibiotics administered during the first 2 days of hospitalization. Because billing code data in the PHIS do not distinguish between administration in the ED or inpatient setting, our definition inherently included antibiotics given in the ED. This definition was chosen to capture the window of time when microbiologic test results are not available to guide the decision for antibiotic choice. Empirical antibiotic selection was divided into 2 broad categories: parenteral antibiotics and oral antibiotics (Supplemental Table 5). Parenteral antibiotics were further subdivided into antibiotics typically administered for staphylococcal infection, antibiotics typically administered for neonatal sepsis, and combination therapy (staphylococcal infection and neonatal sepsis). Antibiotics typically administered for staphylococcal infection included ampicillin-sulbactam, cefazolin, clindamycin, oxacillin, nafcillin, and vancomycin. Because the PHIS does not contain data on local susceptibilities, we chose to include agents with activity against methicillin-susceptible S aureus and/or methicillin-resistant S aureus with the assumption that providers would choose an agent on the basis of their local patterns of resistance. Antibiotics typically administered for neonatal sepsis included ampicillin and gentamicin, ampicillin and a third- or fourth-generation cephalosporin, or a third- or fourth-generation cephalosporin alone. Combination therapy was defined as concomitant use of neonatal sepsis antibiotics and staphylococcal antibiotic(s). The oral antibiotic category was defined as use of oral preparations alone during the first 2 days of hospitalization. Categorization was reviewed and confirmed by 1 of the board-certified pediatric infectious diseases physicians in the study group (R.J.M).

Outcome Measures

Outcome measures included hospital LOS (in days), returns (ED revisits and/or hospital readmission) within 30 days, and cost. The time frame of 30 days was chosen to measure subsequent visits associated with treatment failure, antibiotic-associated adverse effects, or IBI. Return visits were further classified as same cause, related cause, and/or all cause (Supplemental Table 6). A same-cause return was defined as a return for any of the diagnoses in our case identification strategy. A related-cause return was defined a priori and by group consensus as a return for reasons that could reasonably be attributed to management of an SSTI (eg, bacteremia, fever, and diarrhea). All-cause returns were defined as returns for any reason. Cost of index hospitalization included use from the ED visit and hospitalization. Costs are presented as standardized costs by using methodology previously described by Keren et al.²⁶

Demographics and Covariates

Demographic characteristics included age, sex, race and/or ethnicity, primary payer, and region of the United States. All records were assessed for billing codes for fever,²⁷ site of infection, IBI, incision and drainage procedures, obtainment of cultures (blood, cerebrospinal fluid [CSF], and wound), ICU services, and peripherally inserted central catheter (PICC) placement. IBI was defined as bacteremia and/or sepsis, meningitis, osteomyelitis, and pyogenic arthritis and was identified through ICD-9-CM codes (Supplemental Table 4). Additionally, we examined case mix index (CMI), which is a relative weight assigned to each discharge on the basis of All-Patient Refined Diagnosis Group (APR-DRG) assignment and ARP-DRG severity of illness. The weights are derived by Truven Health Analytics (Ann Arbor, MI) as the ratio of the average charge for discharges within a specific APR-DRG and severity of illness combination to the average charge for all discharges in the database. For simplicity of reporting, we split the weights at the median and combined the middle 2 quartiles into a single group to produce 3 categories: minor, moderate and/or major, and extreme.

Validation

An internal validation study was performed through chart review at 2 PHIS hospitals to assess the accuracy of our case identification strategy. Of the 131 medical records reviewed, our case identification strategy was associated with a positive predictive value of 90.8% (N = 119) for identifying included SSTI diagnoses. Two infants (1.5%) with billing codes for cellulitis had documentation to support a diagnosis of neonatal fever, and 10 infants (7.6%) with billing codes for cellulitis had documentation to support a diagnosis of perirectal abscess. Of the 131 charts reviewed, fever (temperature ≥38°C reported or documented) was identified for 18 infants (13.7%). However, fever was coded for only 9 infants (6.9%) in the PHIS. On chart review of these 9 infants, 4 had a documented fever at ≥38°C, 1 had a temperature of 37.9°C, 1 had a subjective fever reported, and 3 infants did not have documentation to support fever. On chart review, 6 infants (4.6%) had positive results on blood cultures, all of which were considered to be contaminants by the medical team (4 coagulase-negative staphylococci, 1 Streptococcus parasanguinis, and 1 Bacillus species [not anthracis]). No cases of meningitis, osteomyelitis, or pyogenic arthritis were identified.

Statistical Analysis

We calculated summary statistics for continuous variables with medians and interquartile ranges (IQRs), and categorical variables were summarized with frequencies and percentages. We made comparisons between infants who did and did not have a return using the χ² test for categorical measures or the Kruskal-Wallis test for continuous measures. We examined the association between antibiotic groups (eg, staphylococcal and combination therapy) and outcomes using generalized linear mixed-effects models with a random intercept for each hospital. We performed age-stratified, diagnosis-stratified, and overall analyses. For diagnosis-stratified analyses, adjustments were made for age, sex, hospital, census region, CMI, incision and drainage, fever code, culture obtainment (blood, CSF, and wound), and ICU use. For age-stratified and overall analyses, we additionally adjusted for infection type. For the outcomes of LOS and cost, we used an exponential distribution, and we present the results as rate ratios with 95% confidence intervals (CIs). For all other outcomes, we used a binomial distribution, and we present the results as odds ratios with 95% CIs. Finally, we performed a sensitivity analysis, removing children with CSF cultures to isolate infants who did not have a complete evaluation for serious bacterial infection on the basis of their initial presentation. All statistical analyses were performed by using SAS 9.4 (SAS Institute, Inc, Cary, NC), and P values <.05 were considered statistically significant.