Initial Treatment Failure in Patients with Complicated Skin and Skin Structure Infections

Abstract

Background:

Consequences of initial antibiotic failure in patients hospitalized for complicated skin and skin-structure infections (cSSSI) are not well understood.

Methods:

Using data from >100 hospitals in the United States, we identified all adults hospitalized for cSSSI between January 1, 2000 and June 30, 2009. We defined “initial therapy” as all parenteral antibiotics administered <24 h of admission, and such therapy was assumed to have failed if the patient (1) received new antibiotic(s) subsequently (excluding similar/narrower spectrum antibiotics or those begun at discharge), or (2) underwent drainage/debridement/amputation >72 h after admission. We limited attention to the 40 most commonly used antibiotic regimens in 2009. We compared clinical and economic outcomes of patients who experienced initial treatment failure and those who did not.

Results:

The rate of initial treatment failure was 16.6% in acute infections (n=13,498), 34.1% in chronic/ulcerative infections (n=1,116), and 26.7% in surgical site infections (SSIs) (n=2,929). Treatment failure was associated with 4.1–7.3 additional days in the hospital and $11,995–$23,655 in additional inpatient charges; the case fatality rate was from 4- to 12-fold higher in patients who experienced treatment failure than in those who did not (all comparisons, p<0.01).

Conclusion:

Initial treatment failure in patients hospitalized for cSSSI is associated with significantly worse clinical outcomes, longer hospital stays, and higher hospital charges than with successful initial treatment.

Although there is no universally accepted definition of complicated skin and skin structure infections (cSSSI), those that require both surgical and antibiotic treatment, or that involve fascia or muscle, are generally considered to be complicated [1, 2]. Management of cSSSIs usually requires parenteral antibiotic therapy, which is often empiric and selected on the basis of patient characteristics, clinical presentation, and treatment guidelines. Treatment is often complicated by the need for surgical intervention, vascular compromise from diabetes or peripheral vascular disease, peripheral neuropathy, the polymicrobial nature of many infections, and growing resistance to many treatment regimens used commonly [1,3 –10].

Although broad-spectrum antibiotic regimens might be most effective as initial empiric therapy for an infection, their benefits must be weighed against their generally higher cost and greater potential to foster drug resistance. Narrow-spectrum regimens, on the other hand, although less likely to foster resistance, may pose higher risks of treatment failure. When initiating therapy, physicians presumably attempt to strike a balance between these competing considerations.

However, relatively little is known, about the risks and consequences of initial treatment failure in cSSSI. In their study of more than 47,000 patients hospitalized for cSSSI, Edelsberg et al. reported that 23% experienced initial treatment failure, defined as drainage, debridement, or receipt of parenteral antibiotics that afforded broader/different coverage than patients' initial regimens at >2 d after their hospital admission [11]. Patients who experienced treatment failure also spent an average of 5.7 additional days in the hospital and incurred $5,285 (2003 dollars) in additional inpatient charges; moreover, their rate of in-hospital mortality was three-fold greater than that of patients who did not experience treatment failure (all p<0.01) [11]. Because the data used by Edelsberg et al. are almost a decade old, we re-examined in the present study the frequency and consequences of initial treatment failure in patients hospitalized for cSSSI, using more recent data from more than 100 acute care hospitals in the United States.

Methods

Data source

Data for the study were obtained from the Cerner Health Facts Database, which contains comprehensive clinical records from approximately seven million inpatient admissions to more than 100 acute care hospitals throughout the United States. Available information for each admission includes patient demographics (age, gender, race/ethnicity); hospital characteristics (geographic region, number of beds, teaching status); discharge diagnoses (principal and all secondary) in the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) format; all antibiotic therapy administered; results of various laboratory tests; and total in-hospital charges. Because all patient-identifying information has been either encrypted or removed from the database, it is compliant with the Health Insurance Portability and Accountability Act of 1996 and with federal guidelines on Public Welfare and the Protection of Human Subjects under Title 45 of the Code of Federal Regulations (CFR). According to the CFR, review and approval by an Institutional Review Board is not needed for a study of this nature, because “subjects cannot be identified, directly or through identifiers linked to the subjects” (45 CFR 46 §46.101).

Source population

We identified all adult (age ≥18 years) patients hospitalized between January 1, 2000 and June 30, 2009 (“study period”), with a principal diagnosis of a cSSSI (ICD-9-CM diagnosis codes 680.XX-683.XX, 685, 686.XX, 707.0X, 707.1X, 707.8, 707.9, 958.3, 997.62, and 998.5X). We limited our attention to patients for whom antibiotic therapy was begun <24 h after hospital admission, and who received such therapy for ≥48 h if still alive. As with prior studies [11, 12], we excluded patients who died within 24 h after hospital admission, and those with evidence of: (1) Infections of body sites/organs other than the skin or skin structures, except bacteremia and systemic inflammatory response syndrome (SIRS); (2) relatively rare or life-threatening infections (e.g., necrotizing fasciitis); (3) osteomyelitis; (4) complications of pregnancy, childbirth, and the puerperium; (5) impetigo (these admissions were believed to represent coding errors); or (6) receipt of plasmapheresis or hemoperfusion.

Among all remaining patients, we identified the initial antibiotic regimen that each patient received, which we defined as comprising all antibiotics that were parenterally administered within 24 h after hospital admission. We then limited our attention to patients who received any of the 40 initial antibiotic regimens used most commonly in 2009 (these patients constituted 78.4% of all cSSSI admissions over the study period). Although somewhat arbitrary, the decision to limit attention to these 40 regimens (of a total of 1,036 regimens identified) was driven by the need to define what would constitute “failure” with each regimen (according to the algorithms described below); because the vast majority of patients in our study sample were treated with one of these regimens, we believed that the additional insight afforded by inclusion of these additional regimens would not justify the associated effort.

We stratified patients according to whether they had acute infections (e.g., cellulitis, abscess), chronic/ulcerative infections (e.g., decubitus ulcer), or surgical site infections (SSIs), because treatment and outcomes may differ substantially among the three groups. Chronic/ulcerative infections, for example, are more likely to involve anaerobes and gram-negative organisms.

Measures

We examined the demographic and clinical characteristics of study subjects in terms of age, gender, and various co-morbidities (as secondary discharge diagnoses; see the Appendix for a complete list of these co-morbidities and their corresponding ICD-9-CM diagnosis codes). We also examined the results of all specimens drawn for laboratory tests <24 h after the patient's initial clinical presentation, to see whether the patient had selected other conditions (e.g., leukopenia, thrombocytopenia, uremia, hypoxemia) that may be predictive of treatment outcome; if multiple values were available, we used only the earliest value.

Initial treatment failure was defined as having occurred if a patient (1) received a new (i.e., not previously received) antibiotic (excluding agents begun at discharge) >24 h after hospital admission, excluding substitution of an agent of similar or narrower spectrum for one or more components of the initial regimen (i.e., de-escalation, streamlining), and also excluding all changes in therapy on the day of (live) hospital discharge or the immediately preceding day (because these changes were deemed unlikely to represent treatment failure); or (2) underwent drainage, debridement, or amputation >72 h after hospital admission. The earliest such event was designated as the time of initial treatment failure. To identify patients receiving “new” antibiotics (as defined above), we developed a unique “failure algorithm” for each of the 40 initial regimens of interest (algorithms are available from the authors upon request). For example, patients receiving regimens that did not provide coverage for methicillin-resistant Staphylococcus aureus (MRSA) were deemed to have experienced initial treatment failure if they began treatment with another antibiotic that provided such coverage >24 h after hospital admission.

Analyses

We used multivariable logistic regression to identify potentially important predictors of initial treatment failure. Covariates in the model included age, gender, race, selected co-morbidities (as described above), history of hospitalization for cSSSI in the prior year, hospitalization (for any reason), hemodialysis or wound care in the prior 30 d, and selected hospital characteristics (i.e., teaching status, number of beds [<300 beds vs. ≥300 beds]).

We also compared the in-hospital mortality >24 h after admission, time in hospital, and total inpatient charges of patients who experienced treatment failure and those who did not. Significance testing for differences in mortality was based on a multivariable logistic regression model. Models based on analysis of covariance were used to compare length of stay and total inpatient charges. Explanatory variables entered into these models included initial treatment outcome (i.e., success vs. failure) as well as the same set of covariates used in analyses of predictors of treatment failure. These comparisons were undertaken alternatively for patients who experienced initial treatment failure at any time during their hospitalization and for those who experienced failure within the first 72 h after hospitalization. The latter analyses were undertaken because the inclusion of patients who experienced failure >72 h after hospital admission could conflate cause and effect (i.e., patients with long hospital stays may experience changes in antibiotic therapy that are unrelated to failure of their initial regimen). In the latter analyses, patients who did not experience treatment failure within 72 h after hospital admission were designated “not failed,” regardless of whether they experienced treatment failure at any time subsequently. All analyses were done with PC-SAS software version 9.1 (SAS Institute, Cary, NC).

Results

We identified 40,825 patients ≥18 years of age who were admitted to the hospital with a principal diagnosis of cSSSI between January 1, 2000 and June 30, 2009 (Table 1). Of all such patients, 17,543 (43%) received parenterally administered antibiotics during the 24-h period after their hospital admission and which were continued for ≥48 h (except in cases of “early” death), and also met all other entry criteria for the study, including receipt of any of the 40 antibiotic regimens prescribed most commonly in 2009. A total of 13,498 patients (77% of the study sample) had acute infections; 1,116 (6%) had chronic/ulcerative infections; and 2,929 (17%) had SSIs. Co-morbidities were relatively common, including alcohol/drug abuse, congestive heart failure, compromised immune status, and type 2 diabetes (Table 2).

Table 1.

Sample Selection

Criteria	Admissions
Total number of admissions in database with principal diagnosis of SSSI,^a and	46,989
Age ≥18 years and	40,825
Received parenteral antibiotics within 24 h of admission and	34,988
Continued receipt of such antibiotics for ≥48 h^b and	31,689
Did not have
Secondary diagnoses of infection^c or	27,618
Necrotizing fasciitis, gangrene, ecthyma gangrenosum or osteomyelitis, or	22,803
Complications of pregnancy, childbirth, and puerperium, or	22,791
Impetigo, or	22,758
Evidence of plasmapheresis/hemoperfusion during hospitalization, and	22,384
Received one of the 40 regimens of interest, and	17,791
Did not die <24 h of hospital admission	17,543

Composed of ICD-9-CM diagnosis codes 680.XX–683.XX, 685.XX, 686.XX, 997.62, 707.0X, 707.1X, 707.8, 707.9, 958.3, and 998.5X.

Or at least one dose of such therapy, for those who died within 48 h of admission.

Excluding bacteremia, septicemia, and/or systemic inflammatory response syndrome (SIRS).

SSSI=skin and skin structure infection.

Table 2.

Demographic and Clinical Characteristics of Patients Hospitalized for cSSSIs: 2000–2009, by Infection Type ^a,
b

	Infection Type
Characteristic	Acute (N=13,498)	Chronic/ulcerative (N=1,116)	Surgical site (N=2,929)
Mean (SD) age, y	55.2 (19.4)	63.1 (18.7)	56.4 (16.3)
Gender
Male	7,249 (53.7)	596 (53.4)	1,342 (45.8)
Female	6,248 (46.3)	520 (46.6)	1,587 (54.2)
Unknown/missing	1 (0.0)	0 (0.0)	0 (0.0)
Race
Caucasian	10,543 (78.1)	797 (71.4)	2,315 (79.0)
African American	1,702 (12.6)	228 (20.4)	331 (11.3)
Hispanic	311 (2.3)	20 (1.8)	58 (2.0)
Other	82 (0.6)	4 (0.4)	30 (1.0)
Unknown/missing	860 (6.4)	67 (6.0)	195 (6.7)
Principal payor
Medicare	2,154 (16.0)	250 (22.4)	408 (13.9)
HMO/managed care	1,908 (14.1)	97 (8.7)	440 (15.0)
Other	1,674 (12.4)	100 (9.0)	293 (10.0)
Unknown/missing	7,762 (57.5)	669 (59.9)	1,788 (61.0)
Comorbidities
Alcohol/drug abuse	2,498 (18.5)	139 (12.5)	319 (10.9)
Burns	50 (0.4)	1 (0.1)	4 (0.1)
Congestive heart failure	1,435 (10.6)	150 (13.4)	224 (7.6)
Type 2 diabetes	3,869 (28.7)	398 (35.7)	746 (25.5)
Evidence of an immunocompromising condition^c	836 (6.2)	85 (7.6)	288 (9.8)
Hypoalbuminemia	46 (0.3)	5 (0.4)	13 (0.4)
Liver disease	441 (3.3)	42 (3.8)	67 (2.3)
Lupus	26 (0.2)	3 (0.3)	4 (0.1)
Lymphedema	458 (3.4)	23 (2.1)	14 (0.5)
Malnutrition	118 (0.9)	83 (7.4)	46 (1.6)
PVD	619 (4.6)	187 (16.8)	161 (5.5)
Renal disease	914 (6.8)	134 (12.0)	150 (5.1)
Bacteremia	273 (2.0)	55 (4.9)	161 (5.5)
SIRS	55 (0.4)	28 (2.5)	78 (2.7)
Clinical characteristics noted at admission^d
Leukopenia	190 (1.8)	23 (2.5)	48 (2.1)
Thrombocytopenia	285 (2.8)	21 (2.3)	31 (1.4)
Hyperglycemia	855 (8.7)	78 (9.2)	98 (4.9)
Hyponatremia	372 (3.7)	35 (4.1)	75 (3.9)
Acidosis	51 (31.7)	8 (44.4)	13 (31.7)
Uremia	1,406 (13.7)	213 (23.6)	185 (8.6)
Anemia	705 (6.7)	174 (19.1)	398 (17.2)
Hypoxemia	25 (16.1)	0 (0.0)	7 (15.9)
Receipt during first 48 h
Drainage/debridement	1,795 (13.3)	153 (13.7)	499 (17.0)
Parenteral nutrition	3 (0.0)	2 (0.2)	5 (0.2)
Swan-Ganz catheterization	0 (0.0)	0 (0.0)	2 (0.1)
Vasoactive medications	19 (0.1)	6 (0.5)	19 (0.6)
Amputation	4 (0.0)	4 (0.4)	1 (0.0)
Receipt during 30-d period prior to admission of
Hospitalization	1,060 (7.9)	232 (20.8)	1,475 (50.4)
Hemodialysis	155 (1.1)	39 (3.5)	72 (2.5)
Cancer chemotherapy	35 (0.3)	2 (0.2)	21 (0.7)
IV antibiotic therapy	1,661 (12.3)	219 (19.6)	1,389 (47.4)
Wound care	72 (0.5)	48 (4.3)	85 (2.9)
Admission source
From ED	10,664 (79.0)	716 (64.2)	1,850 (63.2)
From other acute care facility	108 (0.8)	36 (3.2)	94 (3.2)
From other healthcare facility	31 (0.2)	19 (1.7)	22 (0.8)
Other	2,695 (20.0)	345 (30.9)	963 (32.9)
Admitted to ICU within 24 h of admission	237 (1.8)	39 (3.5)	135 (4.6)
Hospital characteristics
Teaching facility	7,028 (52.1)	583 (52.2)	2,029 (69.3)
Urban setting	13,475 (99.8)	1,115 (99.9)	2,927 (99.9)
Geographic region
New England	1,425 (10.6)	85 (7.6)	279 (9.5)
Mid-Atlantic	4,874 (36.1)	392 (35.1)	931 (31.8)
West North Central	220 (1.6)	50 (4.5)	43 (1.5)
East North Central	2,958 (21.9)	217 (19.4)	722 (24.7)
East South Central	433 (3.2)	62 (5.6)	95 (3.2)
South Atlantic	1,453 (10.8)	147 (13.2)	519 (17.7)
West South Central	1,279 (9.5)	88 (7.9)	188 (6.4)
Mountain	48 (0.4)	3 (0.3)	4 (0.1)
Pacific	808 (6.0)	72 (6.5)	148 (5.1)
Bed size
<200	3,543 (26.2)	225 (20.2)	516 (17.6)
200–299	3,685 (27.3)	316 (28.3)	545 (18.6)
300–499	3,030 (22.4)	235 (21.1)	735 (25.1)
≥500	3,240 (24.0)	340 (30.5)	1,133 (38.7)

Unless otherwise indicated, all values are number of patients (%).

January 1, 2000–June 30, 2009.

Defined as cancer, HIV/AIDS, other primary or secondary immunodeficiency, or receipt within first 48 h of admission of oral/IV corticosteroids, anti-TNFs, and/or drotrecogin alfa.

Based on first recorded value of all relevant labs ordered within first 24 h in hospital.

AIDS=acquired immune deficiency syndrome; cSSSI=complicated skin and skin structure infection; CY=calendar year; ED=emergency department; HIV=human immunodeficiency virus; HMO=health maintenance organization; ICU=intensive care unit; IV=intravenous; PVD=peripheral vascular disease; SD=standard deviation; SIRS=systemic inflammatory response syndrome; TNF=tumor necrosis factor.

The distribution of patient admissions according to initial antibiotic regimen is shown in Table 3. Of note is that the proportion of patients receiving treatment that provided coverage against MRSA (i.e., vancomycin, linezolid, trimethoprim-sulfamethoxazole [TMP–SMX], daptomycin, tigecycline, clindamycin, doxycycline, minocycline, and dalfopristin-quinupristin) ranged from 41.3% (for chronic/ulcerative infections) to 52.8% (for SSIs).

Table 3.

Frequency Distribution of Initial Antibiotic Regimens among Patients Admitted for cSSSIs: 2000–2009^a

	Number of admissions (%) ^b
Acute infections	N=13,498
Cefazolin	3,019 (22.4)
Vancomycin	2,060 (15.3)
Ampicillin-sulbactam	1,852 (13.7)
Piperacillin-tazobactam	832 (6.2)
Piperacillin-tazobactam and vancomycin	732 (5.4)
Clindamycin	710 (5.3)
Levofloxacin	555 (4.1)
Ceftriaxone	486 (3.6)
Ampicillin-sulbactam and vancomycin	324 (2.4)
Clindamycin and vancomycin	306 (2.3)
All others	2,622 (19.4)
Chronic/ulcerative infections	N=1,116
Piperacillin-tazobactam	179 (16.0)
Ampicillin-sulbactam	149 (13.4)
Cefazolin	149 (13.4)
Vancomycin	138 (12.4)
Piperacillin-tazobactam and vancomycin	111 (9.9)
Levofloxacin	61 (5.5)
Ceftriaxone	41 (3.7)
Clindamycin	27 (2.4)
Ampicillin-sulbactam and vancomycin	21 (1.9)
Levofloxacin and metronidazole	18 (1.6)
All others	222 (19.9)
Surgical site infections	N=2,929
Vancomycin	659 (22.5)
Cefazolin	436 (14.9)
Piperacillin-tazobactam	309 (10.5)
Ampicillin-sulbactam	296 (10.1)
Piperacillin-tazobactam and vancomycin	225 (7.7)
Levofloxacin and metronidazole	91 (3.1)
Levofloxacin	86 (2.9)
Cefazolin and vancomycin	75 (2.6)
Clindamycin	66 (2.3)
Ceftazidime and vancomycin	61 (2.1)
All others	625 (21.3)

January 1, 2000–June 30, 2009.

Percentages calculated within each infection subgroup.

cSSSI, complicated skin and skin structure infection.

Acute infections

Among patients with acute infections, the rate of initial treatment failure was 16.6%. Most failures (79%) were defined on the basis of a change in antimicrobial coverage; the remainder (21%) were defined through the need for late surgical intervention. Approximately two thirds of all treatment failures occurred within the first 72 h after admission; the median (interquartile range [IQR]) time to failure was 50 h (range 34 h–86 h). In multivariable logistic regression analysis, significant predictors of initial treatment failure included: bacteremia/SIRS (OR 2.91 [95% CI 2.27–3.73]); malnourishment (OR 2.38 [95% CI 1.61–3.53]); admission to the intensive care unit (ICU) within the first 24 h after hospital admission (OR 1.86 [95% CI 1.39–2.49]); uremia (OR 1.44 [95% CI 1.23–1.68]); anemia (OR 1.44 [95% CI 1.19–1.75]); hyponatremia (OR 1.40 [95% CI 1.09–1.80]); hyperglycemia (OR 1.35 [95% CI 1.12–1.63]); and drainage/debridement during the first 48 h in the hospital (OR 1.19 [95% CI 1.04–1.36]) (Table 4).

Table 4.

Multivariable Logistic Regression of Factors Associated with Failure of Initial IV Antibiotic Therapy in cSSSI, by Infection Type

	Infection type
	Acute			Chronic/ulcerative			Surgical site
Factors	OR	95% CI	p value	OR	95% CI	p value	OR	95% CI	p value
Age (vs. <65 y)
<65 y	1.00	—	—	1.00	—	—	1.00	—	—
65–74 y	0.95	(0.81–1.10)	0.49	1.35	(0.92–1.96)	0.12	1.17	(0.93–1.47)	0.17
≥75 y	0.98	(0.86–1.11)	0.73	1.35	(0.97–1.89)	0.08	1.07	(0.84–1.37)	0.58
Gender (male vs. female)	0.93	(0.85–1.02)	0.13	0.88	(0.67–1.16)	0.36	1.04	(0.87–1.23)	0.68
Race (Caucasian vs. all others)	1.07	(0.94–1.22)	0.32	1.18	(0.85–1.62)	0.32	0.93	(0.73–1.20)	0.59
Comorbidities (yes vs. no)
Alcohol/drug abuse	1.03	(0.91–1.17)	0.68	1.08	(0.70–1.65)	0.73	1.09	(0.83–1.45)	0.53
Burns	0.95	(0.44–2.04)	0.89	—	—	—	0.90	(0.09–8.89)	0.92
Congestive heart failure	1.13	(0.97–1.32)	0.12	1.36	(0.92–2.00)	0.13	1.79	(1.32–2.42)	<0.01
Type 2 diabetes	1.04	(0.93–1.16)	0.54	0.92	(0.68–1.23)	0.56	1.17	(0.96–1.43)	0.11
Immunocompromising conditions	1.01	(0.84–1.22)	0.89	1.21	(0.76–1.94)	0.42	1.04	(0.79–1.38)	0.78
Hypoalbuminemia	1.06	(0.50–2.23)	0.88	—	—	—	6.63	(2.00–21.97)	<0.01
Liver disease	1.28	(0.99–1.64)	0.06	1.36	(0.63–2.92)	0.43	0.85	(0.47–1.53)	0.58
Lupus	1.56	(0.62–3.94)	0.35	1.08	(0.09–13.16)	0.95	2.46	(0.34–17.84)	0.37
Lymphedema	1.16	(0.90–1.48)	0.25	1.33	(0.53–3.31)	0.54	0.36	(0.08–1.66)	0.19
Malnutrition	2.38	(1.61–3.53)	<0.01	2.09	(1.28–3.41)	<0.01	2.09	(1.13–3.89)	0.02
PVD	1.09	(0.88–1.35)	0.42	1.48	(1.04–2.10)	0.03	1.19	(0.83–1.71)	0.34
Renal disease	1.15	(0.95–1.37)	0.14	1.24	(0.82–1.89)	0.31	0.98	(0.67–1.43)	0.91
Bacteremia/SIRS	2.91	(2.27–3.73)	<0.01	2.33	(1.29–4.20)	<0.01	1.67	(1.18–2.37)	<0.01
Clinical characteristics noted at admission (present vs. not present)
Leukopenia	0.84	(0.56–1.26)	0.41	0.67	(0.21–2.08)	0.48	1.50	(0.80–2.83)	0.21
Thrombocytopenia	1.07	(0.78–1.47)	0.68	2.77	(0.83–9.31)	0.10	1.29	(0.57–2.90)	0.54
Hyperglycemia	1.35	(1.12–1.63)	<0.01	1.26	(0.73–2.17)	0.41	1.18	(0.75–1.86)	0.47
Hyponatremia	1.40	(1.09–1.80)	<0.01	2.30	(1.11–4.78)	0.03	1.86	(1.14–3.05)	0.01
Acidosis	1.15	(0.59–2.25)	0.68	0.23	(0.03–2.02)	0.19	1.19	(0.36–3.98)	0.78
Uremia	1.44	(1.23–1.68)	<0.01	0.91	(0.63–1.30)	0.59	1.34	(0.96–1.89)	0.09
Anemia	1.44	(1.19–1.75)	<0.01	1.34	(0.94–1.92)	0.10	1.09	(0.85–1.40)	0.50
Hypoxemia	0.73	(0.24–2.21)	0.58	—	—	—	0.75	(0.14–4.04)	0.74
Receipt during first 48 h of (yes vs. no)
Drainage/debridement	1.19	(1.04–1.36)	0.01	0.90	(0.61–1.32)	0.58	1.17	(0.94–1.46)	0.16
Parenteral nutrition	2.03	(0.16–26.53)	0.59	—	—	—	0.45	(0.05–4.29)	0.49
Swan-Ganz catheterization or vasocative medications	0.75	(0.24–2.36)	0.63	0.19	(0.02–1.87)	0.16	0.68	(0.25–1.86)	0.45
Amputation	3.99	(0.54–29.81)	0.18	—	—	—	—	—	—
Receipt during 30-d period prior to admission of (yes vs. no)
Hospitalization	1.12	(0.91–1.37)	0.29	1.10	(0.65–1.84)	0.72	1.06	(0.84–1.35)	0.62
Hemodialysis/cancer chemotherapy/IV antibiotic therapy or wound care	0.87	(0.74–1.03)	0.11	1.19	(0.72–1.97)	0.51	1.05	(0.83–1.32)	0.70
Admission source (All other vs. ED)	1.09	(0.97–1.22)	0.16	1.11	(0.84–1.46)	0.48	1.09	(0.91–1.30)	0.34
Admitted to ICU (vs. not admitted)	1.86	(1.39–2.49)	<0.01	2.23	(1.10–4.53)	0.03	2.12	(1.44–3.11)	<0.01
Hospital characteristics
Teaching vs. non-teaching	0.98	(0.88–1.09)	0.71	1.31	(0.96–1.78)	0.09	0.82	(0.67–1.00)	0.05
Bed size (<300 vs. ≥300)	0.94	(0.85–1.05)	0.30	0.81	(0.59–1.11)	0.20	0.95	(0.77–1.16)	0.58

CI=confidence interval; ED=emergency department; ICU=intensive-care unit; OR=odds ratio; PVD=peripheral vascular disease; SIRS=systemic inflammatory response syndrome.

Mortality beyond 24 h after admission was about 12-fold higher in patients who experienced initial treatment failure than in those who did not (1.2% vs. 0.1%, respectively; p<0.01). Initial treatment failure was also associated with an additional 4.1 d in the hospital (mean [±SD], 8.0±6.0 d vs. 3.9±2.8 d, respectively) and with $11,995 in additional hospital charges ($22,821±$41,289 vs. $10,826±$21,976) (both comparisons, p<0.01). Treatment failure within the 72-h period after hospital admission as compared with the success of treatment within this period was found to be associated with higher mortality (0.8% vs. 0.2%), longer hospital stays (6.8±4.7 d vs. 4.3±3.6 d), and higher total inpatient charges ($19,924±$40,920 vs. $11,943±$24,073); (all comparisons, p<0.01).

Chronic/ulcerative infections

Among patients with chronic/ulcerative infections, the rate of initial treatment failure was 34%. Although 43% of these failures were defined on the basis of late surgical intervention, 57% were defined as a change in antibiotic regimen or expansion of antibiotic coverage. Relative to that in patients with acute infections, treatment failure in patients with chronic/ulcerative infections occurred later during hospital stays; the median (IQR) time to failure was 77 h (range 44 h–118 h). In multivariable logistic regression, predictors of initial treatment failure included bacteremia/SIRS (OR 2.33 [95% CI 1.29–4.20]), and malnourishment (OR 2.09 [95% CI 1.28–3.41]).

Mortality beyond 24 h after hospital admission was approximately four-fold higher in patients who experienced initial treatment failure than in those who did not (4.5% vs. 1.2%, respectively; p<0.01). Initial treatment failure was also associated with an additional 7.3 d in the hospital (mean [±SD]; 13.2±11.7 d vs. 5.9±4.7 d, respectively) and with $23,655 in additional hospital charges ($39,961±$49,111 vs. $16,306±$21,029); both comparisons, p<0.01). Moreover, treatment failure within the first 72 h after hospital admission was associated with longer hospital stays (10.3±7.3 d vs. 8.0±8.7 d) and higher total inpatient charges ($32,147±$43,424 vs. $22,978±$33,333); both comparisons, p<0.01); however, it was not associated with a higher rate of mortality (2.4% vs. 2.3%; p>0.99).

Surgical site infections

Among patients with SSIs, the rate of initial treatment failure was 26.7%. Most such failures (72%) were defined on the basis of a change in antimicrobial coverage. In multivariable logistic regression analysis, significant predictors of initial treatment failure included ICU admission within the first 24 h after hospital admission (OR 2.12 [95% CI 1.44–3.11]); hyponatremia at admission (OR 1.86 [95% CI 1.14–3.05]); congestive heart failure (OR 1.79 [95% CI 1.32–2.42]; and bacteremia/SIRS (OR 1.67 [95% CI 1.18–2.37]).

Mortality beyond 24 h after hospital admission was about six-fold higher among patients who experienced initial treatment failure than among those who did not (1.2% vs. 0.2%, respectively; p<0.01). Initial treatment failure was also associated with an additional hospital stay of 5.8 d (mean [±SD], 10.4±9.7 d vs. 4.6±3.6 d, respectively), and with $14,523 in additional hospital charges ($26,744±$43,499 vs. $12,221±$19,694; both comparisons, p<0.01). Treatment failure within the first 72 h after hospital admission was associated with a significantly higher mortality (0.8% vs. 0.4%), longer hospital stays (9.0±9.6 d vs. 5.6±5.4 d), and higher total inpatient charges ($24,916±$42,527 vs. $14,420±$25,044); (all comparisons, p<0.01).

Discussion

In selecting an initial antibiotic regimen for hospitalized patients with cSSSIs, there may be conflicting concerns about risks of treatment failure, antibiotic resistance, safety, and cost. Although broad-spectrum regimens may be more effective than those with narrower spectra, their benefits must be weighed against their generally higher cost and the risk of fostering drug resistance. Conversely, although the use of agents with narrower antimicrobial spectra might reduce concerns about antibiotic resistance and costs, their use may carry a greater risk of treatment failure. The risks and consequences of treatment failure are undoubtedly an important consideration in therapeutic decision making.

In our study of more than 17,000 hospitalized patients with cSSSIs, approximately one in five (19.4%) patients were found to have experienced initial treatment failure, ranging from 16.6% among patients with acute infections to 34.1% among those with chronic/ulcerative infections. The designation of initial treatment failure was typically based on a change in antimicrobial coverage, usually within 2–3 d after hospital admission. Significant predictors of initial treatment failure included malnourishment, bacteremia/SIRS, and admission to the ICU within the first 24 h after hospital admission.

To the best of our knowledge, the only other study of initial treatment failure in patients hospitalized for cSSSI is that of Edelsberg et al. [11]. Our findings are similar in most respects to those of this earlier investigation, including the frequency of treatment failure (19.4% vs. the earlier estimate of 22.8%) and the number of additional days of hospitalization associated with this outcome (5.0 d vs. 5.7 d). Although the methods in the two studies were virtually identical, the hospitals were different and the databases encompassed different time periods. Presumably, these differences account for the nominal differences in findings.

Our definition of failure of initial therapy, consisting of receipt after 24 h of a new (i.e., not received previously) antibiotic, excluding agents of similar/narrower spectrum and those begun at hospital discharge, or “late” (i.e., >72 h after admission), drainage, debridement, or amputation, is deserving of comment. To the best of our knowledge, there are no consensus statements on what constitutes treatment failure in hospitalized patients with cSSSI. A definition similar to ours was used in earlier retrospective studies in patients with complicated intra-abdominal infections [13,14] and in Edelsberg's prior study of patients with cSSSI [11]. We believe that the elements in our definition represent clinical deterioration as perceived by the treating physician. Changes in antibiotic therapy are often motivated by this perception, and not by culture results, because most patients with cSSSIs do not have cultures drawn, have negative results of culture, or receive new antibiotics before the availability of culture results. Clinical deterioration may be caused by organisms resistant to the antibiotics administered, the severity of infection, or a patient's inability to withstand infection because of compromised immunity (e.g., advanced age). The relative importance of each of these factors cannot be determined from our study. However, the presence of an organism resistant to the initial antibiotic regimen has been identified as a major risk factor for treatment failure in various infections, including cSSSIs [15 –24]. In a study of 137 patients with either intra-abdominal or skin/soft-tissue infections, a mismatch between antibiotic(s) administered and causative microorganisms was reported to be the most important predictor of clinical failure (OR 14.9; p=0.001) [15]. In another study involving 717 patients with cSSSIs, 22.0% were found to have been given inappropriate initial therapy, defined as antibiotics without activity against the identified pathogen(s) [12].

Although we found that initial treatment failure was associated with significantly worse clinical and economic outcomes, including higher mortality, additional days of hospitalization, and higher inpatient charges, it is important to note that these comparisons represent an “improper” subgroup analysis, in that the comparison groups were defined not on the basis of a baseline characteristic but rather on an outcome (i.e., treatment success vs. failure) [25]. Such comparisons may result in conflations of cause and effect (e.g., the length of hospital stay may be longer for patients who experience treatment failure in part because longer stays may increase the risk of secondary infection and hence the likelihood of a change in antibiotic therapy). To address this issue, we conducted a subgroup analysis in which we compared the outcomes of patients who experienced treatment failure within the first 72 h with those who did not, thereby controlling (at least partly) for this problem. In this subgroup analysis, initial treatment failure remained highly predictive of mortality (beyond 24 h after hospitalization), length of hospital stay, and total inpatient charges. We therefore believe that our findings are not an artifact of an outcomes-based comparison.

Additional limitations of our study merit mention. We focused attention exclusively on patients who received any of the 40 most commonly used initial antibiotic regimens (in calendar year 2009). Although these patients constituted almost 80% of all admissions for cSSSI in our study database, our exclusion of patients who received relatively uncommon regimens may limit the generalizability of our findings. Also, although our study reflects the experience with more than 17,000 patients at more than 100 hospitals throughout the United States, the study database consists of a convenience rather than a random sample, which may result in estimates with unknown biases.

In conclusion, our findings suggest that initial treatment failure is a frequent occurrence in patients admitted to the hospital for cSSSIs, and that it is associated with significantly worse clinical outcomes, longer hospital stays, and higher cost of care than is the success of initial antibiotic treatment. Our findings may be important in helping to guide therapeutic decision making in this setting.

Footnotes

Acknowledgments

Funding for this research was provided by Forest Research Institute. Funding for editorial assistance was provided by Forest Laboratories, Inc. We thank Scientific Therapeutics Information, Inc., for their editorial assistance. We also thank Kenneth LaPensee, PhD, of The Medicines Company, and Yun Su, MD, MPH of Bristol-Myers Squibb, both of whom were employed by Forest Research Institute at the time this study was conducted.

Author Disclosure Statement

Gerry Oster, Ariel Berger, and John Edelsberg are employed by Policy Analysis Inc., in Brookline, Massachusetts. David J. Weber is employed by the University of North Carolina School of Medicine, at Chapel Hill, North Carolina. Xingyue Huang is employed by Forest Research Institute. Forest Research Institute reviewed the study research plan and the study manuscript. The authors conducted the data management, processing, and analyses for the study, and made all final decisions with respect to the analyses and manuscript preparation.

Appendix

Comorbidity	ICD-9-CM diagnosis codes
Alcohol/drug abuse	265.2, 291.0, 291.1, 291.2, 291.3, 291.4, 291.5, 291.81, 291.89, 291.9, 303.XX–305.XX, 648.4, 980.9, V11.3
Burns	941.XX–949.XX
Congestive heart failure	398.91, 402.01, 402.11, 402.91, 404.01, 404.03, 404.11, 404.13, 404.91, 404.93, 428.0X
Type 2 diabetes	250.XX
Evidence of immunocompromise^a	140.XX–209.XX, 230.XX–238.XX (cancer), 42 (HIV/AIDS), 279.X (primary/secondary immunodeficiency)
Hypoalbuminemia	273.8
Liver disease	006.3, 070.22, 070.32, 070.33, 070.44, 070.54, 456.0–456.21, 571.XX, 572.0, 572.2–572.8, 573.X, 996.82, V42.7
Lupus	373.34, 695.4, 710.1
Lymphedema	457.1, 757.0, 997.99, 457.0
Malnutrition	261–263.X, 648.9, 995.52, 995.84
Peripheral vascular disease	440.2X, 440.3X, 443.9X, 444.22
Renal failure	584.XX–586.XX
Bacteremia	038.XX, 790.7
Systemic inflammatory response syndrome	995.9X

In addition to listed diagnoses, receipt within first 48 h of admission of oral/IV corticosteroids, anti-tumor necrosis factor agents, or drotrecogin alfa (activated) was deemed evidence of immunocompromise.

Note: “X” is a wildcard that may range from 0 to 9.

AIDS=acquired immune deficiency syndrome; HIV=human immunodeficiency virus.

References

Nichols

. Optimal treatment of complicated skin and skin structure infections. J Antimicrob Chemother, 1999; 44:19–23.

Center for Drug Evaluation and Research (CDER). Guidance for industry: Uncomplicated and Complicated skin and skin structure infections—developing antimicrobial drugs for treatment. Washington, DC: Food and Drug Administration, US Department of Health and Human Services, 1998.

Deery

II . Outpatient parenteral anti-infective therapy for skin and soft-tissue infections. Infect Dis Clin North Am, 1998; 12:935–949.

Nichols

, Florman

. Clinical presentations of soft-tissue infections and surgical site infections. Clin Infect Dis, 2001; 33:S84–S93.

Fung

, Chang

, Kuczynski

. A practical guide to the treatment of complicated skin and soft-tissue infections. Drugs, 2003; 63:1459–1480.

Doern

, Jones

, Pfaller

et al.

Bacterial pathogens isolated from patients with skin and soft tissue infections: Frequency of occurrence and antimicrobial susceptibility patterns from the SENTRY Antimicrobial Surveillance Program (United States and Canada, 1997)

Diagn Microbiol Infect Dis, 1999; 34:65–72.

Summanen

, Talan

, Strong

et al. Bacteriology of skin and soft-tissue infections: Comparison of infections in intravenous drug users and individuals with no history of intravenous drug use. Clin Infect Dis, 1995; 20:S279–S282.

Nathwani

. Impact of methicillin-resistant Staphylococcus aureus infections on key health economic outcomes: Does reducing the length of hospital stay matter? J Antimicrob Chemother, 2003; 51:ii37–ii44.

Boyce

. Methicillin-resistant Staphylococcus aureus in hospitals and long-term care facilities: Microbiology, epidemiology, and preventive measures. Infect Control Hosp Epidemiol, 1992; 13:725–737.

10.

Stevens

, Bisno

, Chambers

et al. Infectious Diseases Society of America. Practice guidelines for the diagnosis and management of skin and soft-tissue infections. Clin Infect Dis, 2005; 41:1373–1406.

11.

Edelsberg

, Berger

, Weber

et al. Clinical and economic consequences of failure of initial antibiotic therapy for hospitalized patients with complicated skin and skin-structure infections. Infect Cont Hosp Epidemiol, 2008; 29:160–169.

12.

Zilberberg

, Shorr

, Micek

et al. Epidemiology and outcomes of hospitalizations with complicated skin and skin-structure infections: Implications of healthcare-associated infection risk factors. Infect Control Hosp Epidemiol, 2009; 30:1203–1210.

13.

Cattan

, Yin

, Sarfati

et al. Cost of care for inpatients with community-acquired intra-abdominal infections. Eur J Clin Microbiol Infect Dis, 2002; 21:787–793.

14.

Edelsberg

, Berger

, Schell

et al. Economic consequences of failure of initial antibiotic therapy in hospitalized adults with complicated intra-abdominal infections. Surg Infect (Larchmt), 2008; 9:335–347.

15.

Falagas

, Barefoot

, Griffith

et al. Risk factors leading to clinical failure in the treatment of intra-abdominal or skin/soft tissue infections. Eur J Clin Microbiol Infect Dis, 1996; 15:913–921.

16.

Dupont

, Mentec

, Sollet

, Bleichner

. Impact of appropriateness of initial antibiotic therapy on the outcomes of ventilator-associated pneumonia. Intensive Care Med, 2001; 27:355–362.

17.

Ruiz

, Torres

, Ewig

et al. Noninvasive versus invasive microbial investigation in ventilator-associated pneumonia: Evaluation of outcome. Am J Respir Crit Care Med, 2000; 162:119–125.

18.

Kollef

, Sherman

, Ward

, Fraser

. Inadequate antimicrobial treatment of infections: A risk factor for hospital mortality among critically ill patients. Chest, 1999; 115:462–474.

19.

Kollef

, Ward

. The influence of mini-BAL cultures on patient outcomes: Implications for the antibiotic management of ventilator-associated pneumonia. Chest, 1998; 113:412–420.

20.

Sanchez-Nieto

, Torres

, Garcia-Cordoba

et al. Impact of invasive and noninvasive quantitative culture sampling on outcome of ventilator-associated pneumonia: A pilot study. Am J Respir Crit Care Med, 1998; 157:371–376.

21.

Rello

, Rué

, Jubert

et al. Survival in patients with nosocomial pneumonia: Impact of the severity of illness and the etiologic agent. Crit Care Med, 1997; 25:1862–1867.

22.

Luna

, Vujacich

, Niederman

et al. Impact of BAL data on the therapy and outcome of ventilator-associated pneumonia. Chest, 1997; 111:676–685.

23.

Alvarez-Lerma

. Modification of empiric antibiotic treatment in patients with pneumonia acquired in the intensive care unit. ICU-Acquired Pneumonia Study Group. Intensive Care Med, 1996; 22:387–394.

24.

Mosdell

, Morris

, Voltura

et al. Antibiotic treatment for surgical peritonitis. Ann Surg, 1991; 214:543–549.

25.

Yusuf

, Wittes

, Probstfield

, Tyroler

. Analysis and interpretation of treatment effects in subgroups of patients in randomized clinical trials. JAMA, 1991; 266:93–98.