Results
Demographic and Clinical Characteristics
Using ICD-9-CM search, 2842 eligible cases were initially identified. Of 1030 cases manually reviewed, 533 met study entry criteria. Of these patients, 102 were 18 years old or less and included in the pediatric cohort: 43 (42%) had non-purulent cellulitis, 19 (19%) had wound infection or purulent cellulitis and 40 (39%) had cutaneous abscess (see Fig., Supplemental Digital Content 1, http://links.lww.com/INF/B837). Eighty-four of these cases were reported from the freestanding children's hospital. The median age was 5 years (range 45 days to 18 years; see Table, Supplemental Digital Content 2, http://links.lww.com/INF/B838). Comorbid conditions assessed in this study were infrequent; however, prior skin infection was documented in 15% of cases and was particularly common among those with cutaneous abscess. Anatomical sites of infection varied, with the lower extremities and head or neck being most common (44% and 23%, respectively). Upper extremity involvement was more common in cases of non-purulent cellulitis (P = 0.02), whereas inguinal involvement was more common in cases of cutaneous abscess (P = 0.008). Most patients were admitted to a medical service (87 General Pediatrics and 4 Medicine), and surgical consultation was frequent. Infectious Diseases consultation did not occur in any cases.
Resource Utilization and Microbiology
Laboratory, radiographic and microbiological data are presented in Table, Supplemental Digital Content 3, http://links.lww.com/INF/B839. A microorganism was identified in 59 (67%) of 88 cases where at least 1 microbiological culture was obtained. Among cases with a positive culture, S. aureus was the most frequently identified pathogen (41 of 59, 69%), with a similar frequency of methicillin-susceptible S. aureus (32%) and MRSA (34%). Only 1 case was complicated by bacteremia ([beta]-hemolytic streptococcus). In total, a Gram-positive organism was identified in 56 (95%) cases. A Gram-negative organism was identified in 5 (8%) cases; only 1 of these cases involved culture of a single organism. In the other 4 cases, Gram-negative organisms were isolated along with staphylococci (2), streptococci (1) or normal skin flora (1).
Antibiotic Utilization
In 35% of cases, an initial attempt at outpatient antibiotic therapy before hospitalization had failed, most commonly with cephalexin (17%), trimethoprim-sulfamethoxazole (10%) or clindamycin (10%; see Table, Supplemental Digital Content 4, http://links.lww.com/INF/B840). During the hospitalization, the initial empiric antibiotic regimen consisted of a combination of antibiotics in 21% of cases. Clindamycin or vancomycin was included in the initial regimen in 67% and 20% of cases, respectively. Vancomycin was more likely to be used in cases involving cutaneous abscess (P = 0.005). At the time of hospital discharge, a single antibiotic was prescribed in 87% of cases, most commonly clindamycin (66%), cephalexin (20%) or trimethoprim-sulfamethoxazole (11%). The median total duration of therapy was 11 days (interquartile range 10–12) and was similar for all 3 types of infection. Only 3% of patients were treated for 7 or fewer days, and only 7% were treated for >14 days.
The composite primary endpoint of use of antibiotics with broad Gram-negative activity or treatment duration >10 days occurred in 67% of cases and was predominantly driven by treatment duration >10 days (61% of cases; Table 1). Although common among all 3 types of infection, the primary endpoint occurred somewhat more frequently in cases of wound infection or purulent cellulitis (P = 0.08). Antibiotics with broad Gram-negative activity were prescribed in 25% of cases, including only 1 of the 5 cases where a Gram-negative organism was identified. In cases where an antibiotic with broad Gram-negative activity was prescribed, these agents accounted for 32% of the total days of antibiotic therapy. The results of the primary endpoint and its individual components were unchanged when excluding 5 patients with immunosuppression from a hematologic malignancy, connective tissue disease or vasculitis or an immunosuppressing medication (data not shown).
Logistic Regression Models
The multivariate logistic regression model revealed that admission via an emergency department and primary management by a medical (vs. surgical) service were independently associated with the composite primary endpoint (Fig. 1). In separate models, head or neck involvement was independently associated with the use of antibiotics with broad Gram-negative activity (see Fig., Supplemental Digital Content 5, http://links.lww.com/INF/B841), and admission through an emergency department and management by a medical service were independently associated with treatment duration longer than 10 days (see Fig., Supplemental Digital Content 6, http://links.lww.com/INF/B842).
(Enlarge Image)
Figure 1.
Multivariate logistic regression model of factors associated with the composite primary endpoint.
Clinical Outcomes
All patients survived to hospital discharge; follow-up medical encounters were documented most frequently in cases of cutaneous abscess (P = 0.04; Table 2). Clinical failure occurred in 6% of all cases; incidence of treatment failure was significantly higher in cases of abscess than in the other types of ABSSSI (P = 0.04).