Study design
This was a retrospective cohort study of children between 30 days and 18 years of age admitted to two freestanding tertiary children’s hospitals within the same institution in the Southeastern United States with acute hematogenous osteomyelitis and/or septic arthritis over an 8-year period between January 2009 and December 2016. The study was approved by our hospital’s Institutional Review Board.
Selection of participants
Patients were included if they were treated for osteomyelitis and/or septic arthritis based on clinical documentation. Patients were excluded if ultimate management and treatment were not consistent with osteomyelitis or septic arthritis based off imaging findings, cultures, and treatment course (e.g. pyomyositis alone). Patients were also excluded if they had recent trauma such as fracture or foreign body (minor trauma such as bumps or bruises noted in history were still included), had post-operative infection or hardware at the site of the infection, had chronic symptoms (> 6 weeks in duration), had chronic recurrent multifocal osteomyelitis (CRMO), were immunocompromised (e.g. malignancy, sickle cell, chronic immunosuppressive or immunomodulator therapy, transplant recipient, or primary immune disorder), if they were a neonate who had never left the hospital, or if they had the majority of their workup at an outside hospital.
Data collection
Subjects were identified via International Classification of Diseases, Ninth Revision and Tenth Revision (ICD-9 and ICD-10) queries of our electronic medical record (EMR), Epic. We included the ICD-9 codes 730, 730.2, 730.3, and 730.9 for osteomyelitis and 711.0 and 711.9 for pyogenic arthritis, as well as the ICD-10 codes M86.9 for osteomyelitis and M00.9 for pyogenic arthritis. Charts of patients identified via these queries were reviewed for admission history and physical, progress notes, subspecialty consult notes, and discharge summaries to determine if the patient was treated for osteomyelitis and/or septic arthritis. Patients were included if they were treated for one of these diagnoses and excluded if these diagnoses were clinically ruled out. Data collected on these patients included socio-demographics, presenting symptoms, timing of antibiotics, laboratory results, and results of microbiologic and radiographic studies. Data were entered into Research Electronic Data Capture (REDCap, Vanderbilt University, Nashville TN), a secure web application for managing databases, and were de-identified. The type of culture (blood, joint aspirate, bone, “other”) was based off the specimen type recorded in the EMR. “Other” cultures were cultures from wound, abscess, or fluid other than joint fluid collected from the suspected site of infection (eg. subperiosteal abscess) in patients with other clinical, laboratory, or radiographic evidence of osteomyelitis and/or septic arthritis. Time to initiation of antibiotics was defined as the time of admission in our EMR to the time of the first documented administration of an antibiotic. Time to culture was defined as the time of admission in our EMR to the time the culture was obtained. Culture positivity was defined as having a culture which grew a pathogen and pretreated cultures were defined as having been exposed to antibiotics as an outpatient, in an outside emergency room, or at our institution, within 1 week of admission, prior to cultures being obtained. Duration of antibiotic exposure was defined as time of the first dose of antibiotics to the time the culture was obtained. If documentation stated a definitive number of days the patient had received antibiotics instead of exact timing, we presumed those were full days of treatment.
Statistical analysis
All statistical analyses were performed using SAS 9.4 (Carey, North Carolina). Descriptive statistics were calculated for variables of interest, including medians and interquartile ranges (25th and 75th percentiles) for continuous variables, and counts and percentages for categorical variables. Logistic regression was used to assess the effect of antibiotic pretreatment and duration on culture positivity. All models were adjusted for age, gender, race, days of illness, and C-reactive protein. Odds ratios with 95% confidence intervals were presented. A p-value of < 0.05 assumed statistical significance.
Outcome measures
Primary outcome was culture positivity. Symptoms and laboratory values at presentation are described and were used to attempt to control for illness severity. The effect of antibiotic duration on culture positivity was also analyzed.