a b s t r a c t

Background: Infectious Diseases Society of America guidelines recommend that patients hospitalized for acute bacterial skin infections after failure of outpatient antibiotic therapy be managed as “severe” infections; however, the clinical relevance of apparent failure of outpatient therapy is not clear.

Methods: This was a secondary analysis of a multicenter, retrospective cohort of adults and children hospitalized for cellulitis, abscess, or Wound infection. We compared clinical features, laboratory and microbiology findings, antibiotic treatment, and outcomes among patients who received outpatient antibiotics prior to admission and those who did not.

Results: Of 533 patients, 179 (34%) received outpatient antibiotics prior to admission. Compared with those who did not, patients who received antibiotics prior to admission less frequently had fever (18% vs 26%, P = .04) and leukocytosis (33% vs 51%, P b .001). In the 202 cases where a microorganism was identified, Staphylococcus aureus was more common among those who received antibiotics prior to admission (75% vs 58%, P = .02), particularly methicillin-resistant S aureus (41% vs 27%, P = .049), whereas aerobic gram-negative bacilli were less common (3% vs 13%, P = .03). After hospitalization, clinical failure occurred with similar frequency between the 2 groups (12% vs 11%, P = .73).

Conclusions: Patients hospitalized with skin infections after apparently failing outpatient therapy had clinical features suggestive of less severe infection and similar outcomes compared with patients who did not receive antibiotics prior to admission. Our results suggest that inpatient treatment for patients not responding to outpatient therapy should focus on methicillin-resistant S aureus, not gram-negative pathogens.

(C) 2016

Introduction

? Funding: National Institute of Allergy and Infectious Diseases, National Institutes of Health (K23 AI099082). The sponsor had no role in the design of the study, data collection, the analysis and interpretation of data, the writing of the report, or the decision to submit the manuscript for publication.

?? Potential conflicts of interest: T.C.J.: Durata Therapeutics; D.M.P.: Optimer, Cubist, and

Forest Pharmaceuticals. All other authors have no conflicts of interest to disclose.

? Prior presentations: None.

* Corresponding author: Denver Health, 660 Bannock St, Denver, co 80204. Tel.: +1 303 602 5041; fax: +1 303 602 5055.

E-mail address: timothy.jenkins@dhha.org (T.C. Jenkins).

Acute bacterial skin infections result in approximately 14 million ambulatory care visits [1] and 900 000 hospitalizations [2] per year and are among the leading indications for antibiotic use in hospitals [3]. In prior cohorts of patients requiring hospitalization for acute bacte- rial skin infections, 28% to 51% had failed an initial attempt at outpatient antibiotic therapy [4,5]. However, the determination of whether a patient is responding to outpatient therapy can be challenging for emer- gency department (ED) providers, particularly in cases of cellulitis, given that the skin may remain inflamed despite eradication of the causative pathogen. Whether failure of outpatient therapy as judged

http://dx.doi.org/10.1016/j.ajem.2016.02.013

0735-6757/(C) 2016

by clinicians is associated with an increased breadth of microbial path- ogens or affects clinical outcomes has not been previously studied.

Current Infectious Diseases Society of America (IDSA) guidelines for the management of skin infections stratify treatment recommendations based on a clinical classification of mild, moderate, or severe infection [6]. Patients who have failed outpatient antibiotic therapy, have system- ic signs of infection, are immunosuppressed, or have clinical signs suggestive of necrotizing infection are classified as having “severe” infection. In the case of severe nonpurulent cellulitis, an antibiotic regi- men with a broad spectrum of gram-positive and gram-negative activity is recommended (eg, vancomycin plus piperacillin-tazobactam). Given the frequency of failure of outpatient therapy, adherence to this suggested treatment approach would lead ED providers to start broad- spectrum antibiotics in a sizeable proportion of cases. It is thus impor- tant to gain a better understanding of the clinical relevance of failure of outpatient therapy as judged by clinicians. We used a large cohort of patients requiring hospitalization for treatment of an acute bacterial skin infection to determine whether an initial attempt at outpatient therapy prior to admission impacted the microbiology, treatment, and clinical outcomes.

Methods

Study design

We previously described a multicenter, retrospective cohort of patients requiring hospitalization for acute bacterial skin infection between June 1, 2010, and May 31, 2012 [7]. Using this cohort, we performed a secondary analysis comparing patients who were initially attempted to be treated with outpatient antibiotic therapy prior to admission with those who were not.

Study setting and population

The study was performed in 7 Colorado hospitals including a Veteran’s Affairs hospital, tertiary referral center, public safety-net hospital, children’s specialty hospital, and 3 private community hospitals.

Detailed study entry criteria have been described previously [7]. Briefly, adults and children with a principal discharge diagnosis of cellulitis, Cutaneous abscess, or wound infection were identified through Interna- tional Classification of Diseases, Ninth Revision, Clinical Modification codes (680.*, 681.*, 682.*, 686, 035). Cases were manually reviewed to confirm eligibility. The main exclusion criteria were infections with deep tissue involvement (eg, necrotizing fasciitis, osteomyelitis, and myositis) and clinical scenarios that require specialized management, including the following: infected ulcers, surgical site infection, periorbital or perineal infection, infected human or animal bites, health care-associated infec- tions, and hidradenitis suppurativa. At each hospital, for cases meeting the study entry criteria, a clinician with experience in the management of skin infections abstracted clinical and pharmacy data from the medical record using a standardized data collection form. Each clinician underwent training to standardize data collection across sites. The study was approved by each hospital’s institutional review board.

Definitions and study outcomes

Outpatient antibiotic therapy prior to admission was defined as the receipt of one or more systemic antibiotic for the current episode of skin infection leading up to hospitalization, as documented in the med- ical record (Figure A). Both parenteral and oral antibiotics prescribed at ambulatory care visits prior to admission were included in this defini- tion; this information was obtained from provider documentation and pharmacy fill data. When it was specifically noted in the medical record that a patient was prescribed an antibiotic but either did not fill the prescription or did not take any doses of the antibiotic, this was not clas- sified as antibiotic therapy prior to admission. Antibiotic courses for a previous, distinct episode of skin infection were also not classified as antibiotic therapy prior to admission. The specific antibiotics received leading up to hospital admission were recorded; however, the number of days of therapy and the timing in relation to admission were not collected as this information was frequently not available in the medical record.

Fever and leukocytosis were defined as a core body temperature of at least 38.0?C and a serum white blood cell count greater than

Figure. Conceptual framework of the process of care for patients requiring hospitalization for treatment of acute bacterial skin infections. A, Patients who appear to fail an initial attempt at outpatient antibiotic therapy and a list of potential reasons for the lack of response to therapy. B, Patients hospitalized for initial treatment with no attempt at outpatient antibiotic therapy. Ambulatory care refers to EDs, urgent care, and outpatient clinics.

10,000 cells/mm³, respectively. The Charlson age-comorbidity index (a measure of the presence of underlying medical conditions) and Laboratory Risk Index for Necrotizing Fasciitis (LRINEC) scores were calculated as previously defined [8,9]. Vancomycin, daptomycin, linezolid, clindamycin, trimethoprim-sulfamethoxazole, and doxycycline were

Table 1

Demographic and clinical characteristics

Outpatient antibiotic therapy prior to admission (n = 179)

No antibiotics P

prior to admission (n = 354)

classified as antibiotics with activity against methicillin-resistant Staphy- lococcus aureus (MRSA). ?-Lactam/?-lactamase inhibitor combinations,

Type of infection .20

Nonpurulent cellulitis 98 (55) 222 (63)

second- through fifth-generation cephalosporins, fluoroquinolones, carbapenems, tigecycline, aminoglycosides, and colistin were classified	Wound infection or purulent cellulitis Cutaneous abscess	21 (12) 60 (34)	23 (7) 109 (31)
as antibiotics with a broad spectrum of gram-negative activity. From the	Age (y), mean (SD)	42 (23)	46 (25)	.08
time of hospitalization to 45 days after admission (ie, excluding the period prior to admission), clinical failure was defined as a Composite end point of any of the following: (1) treatment failure, defined as a change in	Male Comorbid conditions/ risk factors Diabetes mellitus	101 (56) 39 (22)	216 (61) 68 (19)	.31 .48

antibiotic therapy or unplanned drainage procedure due to inadequate

Clinical response more than 5 days after hospital admission; (2) recur- rence, defined as reinitiation of antibiotics for skin infection after comple- tion of the initial treatment course; or (3) rehospitalization due to skin infection [10].

vascular impairment^a

Injection drug use	11 (6)	25 (7)	.70
HIV infection	3 (2)	9 (3)	.76
Dialysis dependence	4 (2)	3 (1)	.23
Cirrhosis	3 (2)	18 (5)	.06
Lower extremity	25 (14)	71 (20)	.08

Prior skin infection 51 (28) 94 (27) .63

Statistical analysis

Cases were stratified by the presence or absence of outpatient anti- biotic therapy prior to admission. For the analysis of clinical characteris- tics, laboratory and microbiology data, in-hospital and discharge antibiotic therapy, and clinical outcomes, comparisons were performed using the ?² test of association or Fisher exact test where appropriate for categorical variables and the Wilcoxon nonparametric rank sum

Immunosuppressing medication or	17 (10)	52 (15)	.09
condition Charlson age-	1 (0-3)	2 (0-4)	.03
comorbidity index,
median (IQR) Anatomical location Head and neck	23 (13)	31 (9)	.14

Prior MRSA infection or colonization

18 (10) 42 (12) .53

test for continuous variables. Key analyses were stratified by adult	Upper extremity	27 (15)	75 (21)	.09
(N 18 years old) vs Pediatric cases. A P value less than .05 was considered	Trunk	13 (7)	25 (7)	.93
significant. We used SAS Version 9.3 (SAS Institute, Cary, NC) for data analysis.	Buttock, inguinal, or groin	13 (7)	27 (8)	.88

Results

Lower extremity	107 (60)	218 (62)	.69
Site from which			.002
admitted to hospital
ED	141 (79)	309 (87)
Urgent care center	9 (5)	20 (6)
Outpatient clinic	29 (16)	25 (7)
Core body temperature	37.1 (0.9)	37.4 (0.9)	.002
(?C), mean (SD) Feverb	32 (18)	91 (26)	.04
Serum WBC count (cells/mm3), mean	9.7 (4.2)	11.6 (6.4)	b.001
(SD)

Of the 533 patients included in the cohort who were hospitalized for management of an acute bacterial skin infection, 179 (34%) received outpatient antibiotic therapy prior to admission, whereas the remaining 354 (66%) did not (Figure). The proportion of cases involving outpatient antibiotic therapy prior to admission was similar among adults and children (Supplementary Table 1). For the entire cohort, the demo-

graphic characteristics, types of skin infections, risk factors, and anatom-	Leukocytosisc	59 (33)	182 (51)	b.001
ic location of infection were largely similar between those who received	Fever or leukocytosis	76 (42)	215 (61)	b.001
outpatient antibiotic therapy prior to admission and those who did not	CRP obtained	93 (52)	200 (57)	.32
(Table 1). Those who received prior antibiotics had lower Charlson age-comorbidity index scores (median, 1 [interquartile range {IQR}, 0-	CRP (mg/dL), mean (SD) LRINEC score 0-5	24.5 (35.3) (n = 69) 68 (99)	43.4 (59.4) (n = 126) 116 (92)	.15 .06
3] vs 2 [IQR, 0-4], P = .03) and less frequently had fever (18% vs 26%,	6-7	1 (1)	8 (6)
P = .04) and leukocytosis (33% vs 51%, P b .001). C-reactive protein	>= 8	0	2 (2)
(CRP) levels were performed in similar proportions in each group Data presented as n (%) unless otherwise noted. (52% vs 57%, P = .32). Mean CRP levels (24.5 mg/dL vs 43.4 mg/dL, Abbreviation: WBC, white blood cell. P = .15) and LRINEC scores (P = .06) were lower in the group that a Chronic venous stasis, lymphedema, peripheral arterial disease, saphenous vein

received outpatient antibiotic therapy prior to admission; however, these comparisons did not reach statistical significance. In addition, those who received antibiotics prior to admission were more frequently admitted from outpatient clinics (16% vs 7%) and less frequently admit- ted from the ED (79% vs 87%; P = .002).

Microbiology

Microbiological cultures were less likely to be obtained in patients who received outpatient antibiotics prior to admission (77% vs 85%, P = .02), primarily due to less frequent use of blood cultures (55% vs 69%, P b .001; Table 2). At least one microorganism was identified in a similar proportion of cases in the 2 groups (35% vs 39%, P = .36). Of the 202 cases in which a microorganism was identified, S aureus was more common among those who received outpatient antibiotics prior to admission (75% vs 58%, P = .02), particularly MRSA (41% vs 27%,

harvest.

^b Core body temperature >=38.0?C.

^c Serum WBC N 10 000 cells/mm³.

P = .049). In contrast, ?-hemolytic streptococci (10 vs 22%, P = .03) and aerobic Gram-negative organisms (3% vs 13%, P = .03) were identi- fied less frequently among those who received antibiotics prior to admission. The microbiological findings were similar when the analysis was stratified by adult vs pediatric cases (Supplementary Table 1) and when limited to cases involving purulent infections (Supplementary Table 2).

Outpatient antibiotic therapy prior to admission

Of the 179 patients who received outpatient antibiotic therapy prior to admission, 142 (79%) had received oral antibiotics alone, whereas 37

Table 2 Microbiology
	Outpatient antibiotic therapy prior to admission (n = 179)	No antibiotics prior to admission (n = 354)	P
Any microbiological culture obtained	138 (77)	302 (85)	.02
Surface culture	30 (17)	52 (15)	.53
Abscess material	47 (26)	91 (26)	.89
Tissue culture	10 (6)	25 (7)	.52
Blood culture	98 (55)	246 (69)	b.001
Any microorganism identified	63 (35)	139 (39)	.36
Any gram-positive organism	61 (97)	129 (93)	.35
S aureus	47 (75)	80 (58)	.02
Methicillin-susceptible	20 (32)	38 (27)	.52
Methicillin-resistant	26 (41)	38 (27)	.049
Susceptibility not performed	1 (2)	4 93)	1.00
Streptococcal species	16 (25)	43 (31)	.42
Streptococcus anginosus group	5 (8)	10 (7)	1.00
?-Hemolytic streptococcus	6 (10)	31 (22)	.03
Other ?-hemolytic streptococcus	4 (6)	4 (3)	.26
Other	1 (2)	1 (1)	.53
S aureus or streptococci	57 (90)	114 (82)	.12
Coagulase-negative staphylococcus	11 (17)	22 (16)	.77
Aerobic gram-negative	2 (3)	18 (13)	.03
Anaerobic organism(s)	4 (6)	13 (9)	.48
Enterococcus species	0	2 (1)	1.00
Other	4 (6)	6 (4)	.51

Data presented as n (%).

(21%) received at least one dose of a parenteral agent (Table 3). Seventy-five (42%) received more than one antibiotic prior to admis- sion. The outpatient antibiotic regimen lacked an MRSA-active agent in 67 (37%) cases, including 19 (32%) of 60 cases involving a cutaneous abscess.

Antibiotic therapy after hospital admission

Once hospitalized, patients who received outpatient antibiotic ther- apy prior to admission were more likely to be started on an MRSA-active agent (92% vs 85%, P = .04)–most commonly vancomycin–by the provider assuming care for the patient in the hospital (Table 4). Of the 67 patients who received antibiotics that lacked MRSA activity prior to admission, 58 (87%) were started on MRSA-active therapy in the hospi- tal. Patients who received antibiotics prior to admission were somewhat more likely to be prescribed combination antibiotic therapy upon

hospitalization (41% vs 33%, P = .06). Use of antibiotics with broad gram-negative activity upon hospitalization was similar between the groups (38% vs 32%, P = .18). At the time of hospital discharge, more patients who had received outpatient antibiotic therapy prior to admis- sion were prescribed an MRSA-active agent (79% vs 67%, P = .003) and Combination therapy (22% vs 15%, P = .049).

Clinical outcomes after hospitalization

From the time of hospitalization to the end of the 45-day follow-up period, there were no significant differences in the composite end point of clinical failure (12% vs 11%, P = .73) or the individual components of this end point between the 2 groups (Table 5). For patients who received outpatient antibiotic therapy prior to admission, the median length of hospital stay was shorter (3 [IQR, 3-5] vs 4 [IQR, 3-5] days, P = .07), although this difference did not reach statistical significance.

Table 3

Antibiotics administered or prescribed at ambulatory care visits for the 179 patients who received outpatient antibiotic therapy prior to admission

	Nonpurulent cellulitis (n = 98)	Purulent cellulitis or wound infection (n = 21)	Cutaneous abscess (n = 60)	Total (n = 179)
Received parenteral antibiotic	20 (20)	7 (33)	10 (17)	37 (21)
Received oral antibiotic(s) only	79 (80)	14 (67)	50 (83)	142 (79)
Received >= 2 antibiotics	41 (42)	12 (57)	22 (37)	75 (42)
Any MRSA-active agent	56 (57)	15 (71)	41 (68)	112 (63)
Trimethoprim-sulfamethoxazole	35 (36)	14 (67)	29 (48)	78 (44)
Clindamycin	11 (11)	0	6 (10)	17 (10)
Doxycycline	6 (6)	2 (10)	5 (8)	13 (7)
Linezolid	2 (2)	0	0	2 (1)
Parenteral MRSA-active agenta	10 (10)	4 (19)	8 (13)	22 (12)
Any ?-lactam	68 (69)	15 (71)	35 (58)	118 (66)
Cephalexin	41 (42)	12 (57)	26 (43)	79 (44)
Amoxicillin-clavulanate	16 (16)	1 (5)	3 (5)	20 (11)
Other oral ?-lactam	8 (8)	2 (10)	3 (5)	13 (7)
Parenteral ?-lactamb	12 (12)	4 (19)	5 (8)	21 (12)
Fluoroquinolone	5 (5)	0	3 (5)	8 (4)
Other or unknown	4 (4)	1 (5)	2 (3)	7 (4)
Drainage procedure prior to presentation	4 (4)	4 (19)	29 (48)	37 (21)

Data presented as n (%).

^a Parenteral vancomycin (17) or daptomycin (5).

^b Parenteral cephalosporin (17), ertapenem (3), or ampicillin-sulbactam (1).

Table 4

Antibiotic therapy after hospital admission

Outpatient antibiotic therapy prior to admission (n = 179)

Antibiotics initiated by the hospital provider assuming care for patient

No antibiotics prior to admission

(n = 354) P

were not attempted to be treated with outpatient antibiotics prior to admission, those who appeared to have failed outpatient therapy had features of less severe illness: they had less frequent fever and leukocyto- sis, a trend toward lower CRP levels and LRINEC scores, and fewer comor- bidities. S aureus was more likely to be isolated from those who received outpatient therapy prior to admission, but aerobic gram-negative or anaerobic pathogens were not. Initiation of antibiotic therapy with

MRSA activity in the hospital and prescription of an MRSA-active agent

MRSA-active agent 164 (92) 302 (85) .04 Agent with broad gram-

negative activity	68 (38)	114 (32)	.18	outpatient therapy prior to admission. Clinical outcomes between the 2
Combination therapya	74 (41)	117 (33)	.06	groups were similar.

at hospital discharge were more common among those who received

There are a number of reasons why patients with acute bacterial skin infections may not appear to have an appropriate clinical response to

Vancomycin	128 (72)	224 (63)	.06
Clindamycin	39 (22)	82 (23)	.72
?-Lactam/?-lactamase

inhibitor combination	34 (19)	62 (18)	.67	initial outpatient antibiotic therapy (Figure). Among such reasons are
Higher generation				failure to drain abscesses, insufficient antibiotic levels at the site of
cephalosporins	14 (8)	29 (8)	.88	infection, antimicrobial-resistant pathogens, patient nonadherence to

antibiotic therapy, ongoing inflammation despite eradication of the infecting pathogen, or incorrect diagnosis. Although the lack of clear improvement in signs and symptoms of infection after the initiation of outpatient antibiotics is often considered by clinicians to represent failure of outpatient therapy and is referred to as such in IDSA guide- lines, our results raise the question of whether the lack of response is

Cefazolin	7 (4)	27 (8)	.10
Carbapenem	12 (7)	12 (3)	.08
Otherb	26 (15)	45 (13)	.56
Antibiotics prescribed at
hospital discharge
MRSA-active agent	142 (79)	237 (67)	.003
Agent with broad gram-
negative activity	37 (21)	85 (24)	.39

Combination therapy	39 (22)	53 (15)	.049	due to a true failure of antibiotic therapy or a process that mimics ongo-
Clindamycin	43 (24)	99 (28)	.33	ing infection. One might expect that patients who truly fail outpatient

Trimethoprim-
sulfamethoxazole	51 (28)	67 (19)	.01
Cephalexin	17 (10)	49 (14)	.15
Amoxicillin-clavulanate	17 (10)	47 (13)	.20
Doxycycline	26 (15)	42 (12)	.38
Fluoroquinolone	14 (8)	27 (8)	.94
Linezolid	13 (7)	14 (4)	.10
Vancomycin or daptomycin	17 (9)	18 (5)	.052
Otherc	12 (7)	30 (8)	.47
Total duration of therapy,
median (IQR)	12 (10-15)	12 (10-15)	.59
Duration of inpatient
therapy, median (IQR)	3 (3-4)	4 (3-5)	.16
Duration of therapy after
discharge, median (IQR)	9 (7-10)	8 (7-10)	.58

Data presented as n (%) unless otherwise noted.

^a Defined as receipt of 2 or more antibiotics concurrently.

^b Includes fluoroquinolone (23), daptomycin (18), nafcillin (7), linezolid (7), trimeth-

oprim-sulfamethoxazole (9), doxycycline (3), other ?-lactam (5), and other (9).

^c Includes nafcillin (1), ertapenem (5), meropenem (1), ceftriaxone (4), rifampin (1),

cefuroxime (1), cefdinir (2), ceftaroline (1), amoxicillin (5), dicloxacillin (14), penicillin

VK (1), metronidazole (1), and erythromycin (1).

Discussion

Similar to previous studies [4,5], in this multicenter cohort, we found that approximately one-third of patients requiring hospitalization for acute bacterial skin infections had been attempted to be treated as outpatients prior to admission. To our knowledge, this is the first study to evaluate the clinical relevance of what is frequently referred to as “failure of outpatient therapy” by evaluating the effects of outpatient antibiotic therapy prior to admission on the microbiology, treatment, and clinical outcomes. We found that compared with patients who

antibiotic therapy are more likely to have underlying comorbid condi- tions or severe infection that reduce the probability of initial treatment success; the findings from this study are therefore notable. First, compared with patients who did not receive outpatient antibiotics prior to admission, those with apparent failure of outpatient therapy actually had lower Charlson comorbidity scores, fewer comorbidities that might delay the clinical response to antibiotics such as lower extremity vascular impairment and immunosuppression, less frequent fever and leukocytosis, and a trend toward lower CRP levels and LRINEC scores. In addition, they were more likely to have presented to an outpatient clinic and had a trend toward shorter hospital stays.

Taken together, these findings suggest that the group of patients with apparent failure of outpatient therapy had milder, less complicated infections. One logical explanation for this is that patients who were hospitalized without receiving antibiotics prior to admission had associ- ated comorbid conditions or severe infections that compelled ED pro- viders to hospitalize immediately rather than attempt outpatient treatment. However, an alternative explanation is that the patients with apparent failure of outpatient therapy had evidence of less severe infection because the outpatient antibiotics were effective to some degree. That is, the outpatient antibiotics had at least partially treated the infection by the time of hospitalization despite the clinical findings that prompted hospitalization. This is plausible because residual inflam- mation (erythema) of the skin after eradication of the infecting patho- gen can be difficult to distinguish from active infection. This underscores the potential importance of research to identify biomarkers that correlate with Treatment response that could be used as a tool to augment the clinical evaluation of the response to therapy.

The differences in the microbiology between the 2 groups have po- tentially important treatment implications. First, as discussed previously,

Table 5 Clinical outcomes after hospitalization
	Outpatient antibiotic therapy prior to admission (n = 179)	No antibiotics prior to admission (n = 354)	P
Survived to discharge	179 (100)	353 (99.7)	1.00
Clinical failure	21 (12)	38 (11)	.73
Treatment failure	8 (4)	22 (6)	.40
Recurrence	12 (7)	15 (4)	.22
Rehospitalization due to skin infection	11 (6)	21 (6)	.92
Rehospitalization for reason other than skin infection	9 (5)	22 (6)	.58
Length of stay (d), median (IQR)	3 (3-5)	4 (3-5)	.07

Data presented as n (%) unless otherwise noted.

current IDSA guidelines suggest that failure of outpatient therapy should prompt the classification of severe infection and, in the case of nonpurulent cellulitis, treatment with broad-spectrum antibiotic regi- mens [6]. In our cohort, patients who had apparently failed outpatient therapy had a significantly lower frequency of gram-negative pathogen involvement and similar frequency of anaerobic involvement. In addi- tion, S aureus, particularly MRSA, was actually more common among those who failed outpatient therapy. Although the findings from this study must be interpreted cautiously, the clinical and microbiological data suggest that in patients with skin infections who require hospitali- zation due to an inadequate response to outpatient antibiotic therapy, expanding therapy to include gram-negative and anaerobic activity is not likely to be beneficial and, thus, is not routinely indicated. Further- more, these data suggest that in the IDSA guideline, grouping failure of outpatient therapy with clinical factors such as immunosuppression and evidence of necrotizing infection as criteria for severe infection may not be appropriate. Additional studies are needed to confirm these preliminary findings and to investigate the underlying reasons for appar- ent lack of clinical response to outpatient therapy. Furthermore, in future revisions to national guidelines, treatment recommendations for this common clinical scenario should be reevaluated.

That S aureus was isolated more commonly in cases that involved apparent failure of outpatient therapy warrants additional discussion. There are several potential explanations for this finding. First, skin infec- tions caused by S aureus may be more difficult to treat or may be slower to clinically respond compared with those caused by streptococci and other pathogens (eg, due to increased virulence and antimicrobial resis- tance). Therefore, among patients not responding to outpatient antibi- otic therapy who require hospitalization, this could result in a disproportionate number of cases involving S aureus compared with those who are hospitalized without an attempt at outpatient treatment. Second, in our cohort, the outpatient antibiotic regimen lacked MRSA activity in more than 30% of cases, including in cutaneous abscesses where MRSA is known to be the most common pathogen [10,11]. This frequent lack of MRSA-active antibiotic therapy may have contributed to the disproportionate identification of MRSA among patients who failed outpatient therapy as compared with those who did not receive antibiotics prior to admission. It is also important to note that of patients with a cutaneous abscess who failed outpatient therapy, approximately half had not undergone incision and drainage. Because drainage is the mainstay of therapy for abscesses, this may explain the lack of clinical response to antibiotic therapy in such cases. These findings suggest that ensuring primary drainage of skin abscesses and prescription of antibiotics with MRSA activity (when the decision is made to prescribe adjunctive antibiotics) are opportunities to improve the initial manage- ment of acute bacterial skin infections that may potentially reduce the Need for hospitalizations.

Limitations

This study has a number of limitations. First, as a secondary analysis with multiple comparisons, there is the risk of type 1 error. We did not adjust for multiple comparisons because this was an exploratory analy- sis that was intended to raise questions for the field about a common clinical scenario. Second, because the data were collected retrospective- ly, the exposure variable of receipt of outpatient antibiotic therapy prior to admission (failure of outpatient therapy) was subject to misclassifica- tion. The most likely misclassification is that antibiotics were taken prior to admission but not reported by the patient or not documented by the provider in the medical record; however, this would have biased the comparison between the 2 groups toward the null. In addition, the specific rationale for the decision to hospitalize patients who were receiving outpatient antibiotic therapy prior to admission was not known. Because this study was limited to cases where the skin infection

was the primary reason for admission (ie, the principal diagnosis), it is likely that most cases were judged by providers not to be responding adequately to the outpatient antibiotic therapy; however, there may have been alternative reasons for admission in a subset of cases. Third, we were not able to evaluate the underlying reasons for the apparent lack of response to outpatient therapy with this study design. Fourth, the study included only patients who were hospitalized for treatment. It would have also been of interest to have a comparison group of patients who were treated as outpatients and responded appropriately to therapy. Fifth, due to the observational nature of the study, the comparisons made were subject to confounding, and thus, the observed differences between the groups may have been due to factors other than the receipt of outpatient antibiotic therapy prior to admission. Sixth, we did not have data on the dose, duration, and adherence to antibiotics prescribed prior to admission and thus cannot make inferences about their effects. Finally, because the infecting pathogen is uncommonly identified in patients with cellulitis, the inferences from the microbio- logical data may not be able to be extrapolated to cellulitis.

Conclusions

This study highlights that among patients requiring hospitalization for acute bacterial skin infection, an apparent lack of clinical response to initial outpatient therapy was very common. Compared with patients who did not receive antibiotics prior to admission, those who appeared to fail outpatient therapy had fewer comorbidities, clinical features of less severe illness, and similar outcomes. These findings have potential implications for ED providers, raising the question of whether such cases could represent partially treated infection rather than true failure of therapy. In our cohort, the key interventions among patients who had received prior antibiotic therapy were to assure coverage for MRSA and identify drainable pus. Additional research is needed to better under- stand the reasons for the frequent apparent lack of clinical response to outpatient therapy and to develop tools to augment clinical judgment when evaluating response to therapy.

References

1. Hersh AL, Chambers HF, Maselli JH, Gonzales R. National trends in ambulatory visits and Antibiotic prescribing for skin and soft-tissue infections. Arch Intern Med 2008; 168(14):1585-91.
2. Edelsberg J, Taneja C, Zervos M, Haque N, Moore C, Reyes K, et al. Trends in US hos- pital admissions for skin and soft tissue infections. Emerg Infect Dis 2009;15(9): 1516-8.
3. Magill SS, Edwards JR, Beldavs ZG, Dumyati G, Janelle SJ, Kainer MA, et al. Prevalence of antimicrobial use in US acute care hospitals, May-September 2011. JAMA 2014; 312(14):1438-46.
4. Jenkins TC, Knepper BC, Sabel AL, Sarcone EE, Long JA, Haukoos JS, et al. Decreased Antibiotic utilization after implementation of a guideline for inpatient cellulitis and cutaneous abscess. Arch Intern Med 2011;171(12):1072-9.
5. Wargo KA, McCreary EK, English TM. Vancomycin combined with clindamycin for the treatment of acute bacterial skin and skin-structure infections. Clin Infect Dis 2015;61(7):1148-54.
6. Stevens DL, Bisno AL, Chambers HF, Dellinger EP, Goldstein EJ, Gorbach SL, et al. Practice guidelines for the diagnosis and management of skin and soft tissue infec- tions: 2014 update by the Infectious Diseases Society of America. Clin Infect Dis 2014;59(2):e10-52.
7. Jenkins TC, Knepper BC, Moore SJ, O’Leary ST, Brooke C, Saveli CC, et al. Antibiotic prescribing practices in a multicenter cohort of patients hospitalized for acute bacterial skin and skin structure infection. Infect Control Hosp Epi 2014; 35(10):1241-50.
8. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prog- nostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987;40(5):373-83.
9. Wong CH, Khin LW, Heng KS, Tan KC, Low CO. The LRINEC (Laboratory Risk Indicator for Necrotizing Fasciitis) score: a tool for distinguishing necrotizing fasciitis from other soft tissue infections. Crit Care Med 2004;32(7):1535-41.
10. Jenkins TC, Sabel AL, Sarcone EE, Price CS, Mehler PS, Burman WJ. Skin and soft- tissue infections requiring hospitalization at an academic medical center: opportuni- ties for Antimicrobial stewardship. Clin Infect Dis 2010;51(8):895-903.
11. Moran GJ, Krishnadasan A, Gorwitz RJ, Fosheim GE, McDougal LK, Carey RB, et al. Methicillin-resistant S. aureus Infections among patients in the emergency depart- ment. N Engl J Med 2006;355(7):666-74.