Treatment Failure in Emrgency Department Patients with Cellulitis
By: Murray, Heather; Stiell, Ian; Wells, George. CJEM: The Journal of the Canadian Association of Emergency Physicians, Jul2005, Vol. 7 Issue 4, p228-234, 7p
To identify the rate of treatment failure in emergency department patients with cellulitis.
This prospective observational convenience study enrolled adult patients with uncomplicated cellulitis. Physicians performed a standardized assessment prior to treatment To calculate the interrater reliability of the assessment, duplicate data collection forms were completed on a small subsample of patients. Treatment failure was defined as the occurrence of any one of the following events after the initial emergency department visit: incision and drainage of abscess; change in antibiotics (not due to allergy/intolerance); specialist consultation; or, hospital admission.
Comparison of means and proportions between the 2 groups was performed with univariate associations, using parametric or non-parametric tests where appropriate. Results: Seventy-five patients were enrolled; 57% were male, the mean age was 48 (standard deviation 19), 71(95%) patients had extremity cellulitis and 10 (13%) had abscess with cellulitis. Fourteen episodes (18.7%, 95% confidence interval [Cl] 11 %-28%) were classified as treatment failures, with an oral antibiotic failure rate of 6.8% (95% Cl 2%-22%) and an emergency department-based intravenous antibiotic failure rate of 26.1% (95% Cl 16%-40%). Patients with treatment failure were older (mean age 59 yr v. 46 yr, p = 0.02) and more likely to have been taking oral antibiotics at enrolment (50% v. 16.4%, p = 0.01). Patients with a larger surface area of infection were also more likely to fail treatment (465.1 cm² v. 101.5 cm², p <>
Interrater agreement was high for the presence of fever (kappa 1.0) and the size of surface area of infection (intraclass correlation coefficient 0.98), but low for assessments of both severity (kappa 0.35) and need for admission (kappa 0.46).
The treatment of cellulitis with daily emergency department-based intravenous antibiotics has a failure rate of more than 25% in our centre. Cellulitis patients with a larger surface area of infection and previous (failed) oral therapy are more like to fail treatment. Further research should focus on defining eligibility for treatment with emergency department-based intravenous antibiotics.
The primary outcome was the rate of treatment failure.
There are currently no validated definitions of treatment
failure for cellulitis in the published literature. We devised
an evidence-based definition relying on expert opinion and
on our own observation of treatment patterns in this
centre.'"'"" Patients were labelled treatment failures if they
required any of the following: specialist consultation, hospital
admission, a change in antibiotics or a surgical procedure
(e.g., incision and debridgement). A change in antibiotics
was defined as an "upgrade" to IV antibiotics from
oral, or a change from one IV antibiotic to another; however,
patients who required a change in antibiotics due
only to an adverse effect from their antibiotic were classified
as clinical responders if they did not have any of the
other criteria listed under treatment failure.
Study personnel attempted telephone contact of all enrolled
patients 1 week after discharge to ensure that their
infection had continued to resolve as expected. Patients
without complete resolution of their symptoms or who reported
a worsening in symptoms were asked to return to
the ED for a re-evaluation. The hospital medical records of
all study patients not contacted were reviewed 1 month after
their final visit to ensure that they had not had further
ED visits for re-evaluation of their infection.
Univariate associations between historical features, co-existent
illness, clinical characteristics and the primary outcome
of treatment failure were calculated using SPSS.
Proportions were compared using chi-squared and Eisher's
exact tests, and the means of continuous variables were
compared using Student's r tests and Mann-Whitney U
tests where appropriate.
The reliability of the selected cellulitis measurements
was evaluated on the sample of patients who had duplicate
data forms completed. The percentage agreement for each
variable was calculated. Kappa statistics were calculated
for dichotomous variables and intraclass correlation coefficients
(one way random effects model) for the continuous
During the 8-month study period 86 patients were enrolled.
Eleven patients were subsequently excluded from analysis
because they had infections deemed ineligible (1 bite
wound, 3 post-op. 2 diabetic foot infections) or were referred
or admitted at the first ED visit (5 patients), leaving a
study population of 75 patients. A summary of treatment decisions
and patient outcomes is displayed in a flow chart in
Figure 1 (see page 230). Of the 70 patients ultimately discharged
from ED care, 48 (68.6%) were successfully contacted
and reported their infection to be completely resolved
or significantly improved. None of the remaining 22 patients
returned to either centre within 1 month of their treatment,
and thus these patients were classified as clinical responders.
The demographic data and comorbidities of the 75 study
patients are shown in Table 1. The historical features and
physical exam findings are summarized in Table 1. Of
note, 18 patients (24%) in the study group reported a previous
episode of cellulitis. the majority occurring in
the same anatomic location.
Of the 75 study patients, 29 (38.6%) were treated
initially with oral antibiotics and 46 (61.4%) were
given ED-based IV antibiotics. Fourteen patients
met the definition for treatment failure, for an overall
failure rate of 18.7% (95% CI ll%-28%).
There were 2 treatment failures in the oral antibiotic
group (6.8%, 95% CI 2%-22%). compared
with 12 of the 46 patients (26.1%, 95% CI
16%^0%) in the IV group. Table 2 describes the
type and timing of the treatment failure outcome.
Nineteen patients had duplicate data collection
forms completed, by a total of 11 different physicians.
Calculated measures of agreement for these
data are summarized in Table 3. The agreement
between physicians on the presence of fever, the
presence of systemic symptoms and the likelihood
of abscess were high, with kappas of 1.0 and 0.73,
respectively. There was poor agreement for both
physician assessment of severity (kappa 0.35) and
teristics of ihe treatment failure group compared with those
patients who re.sponded successfully. Older patients were
significantly more likely to fail treatment. Other characteristics
associated with failure included prior antibiotic treatment
and the initial size of the infection, as measured both
by the single largest diameter and the area of erythema.
Our high rate of treatment failure is concerning, and reinforces
the need for further evaluation of this therapy. Iden-
tifying patients who are likely to fail treatment is an important
step in evaluating eligibility for any treatment plan.
Patients who are older or have peripheral vascular disease
seem more likely to fail treatment, as do patients previously
treated with oral antibiotics for cellulitis or who have
a larger size of infection at first ED presentation. These
findings should be verified in future studies. The higher
rate of treatment failure seen in our ED-based IV treatment
group was possibly due to the increased severity of those
infections. We may be underestimating which patients seen
in the ED require hospital admission. Currently there are
no guidelines defining the indications for either IV
therapy or hospital admission. Further study into
the indicators of infection severity and clarification
of which infections require inpatient treatment
Published estimates of treatment failure in cellulitis
vary. The failure rate in a retrospective chart review''
of 170 patients treated with ED-based IV antibiotics
in our centre over a different I -year period
— using the same criteria for treatment failure —
was 27.4%. A 1999 abstract comparing a protocol of
ED-based IV antibiotics versus a single dose of IV
antibiotics followed by oral therapy found a failure
rate of 32.4%.''' Another group reported a treatment
failure rate of only 12% in 346 patients when looking
at the etTicacy of ED-based IV therapy using cefazolin
and probenecid." A Canadian study examined
the treatment strategies in 5 Canadian urban
centres and found an overall treatment failure rate of
12% in 416 patients treated for cellulitis over a Iyear
This study (which included both oral
and IV antibiotic regimens) noted more than 25 different
initial antibiotic choices and dose regimens in
the study patients, and did not pubhsh the IV treatment
failure rate. An ED-based trial comparing IV
cefazolin and probenecid with IV ceftriaxone reported
similar failure rates in both arms (7% v. 8%).
However the cause of cellulitis in the majority of the
trial population was IV drug use and, due to concerns
about follow-up, all patients in both arms were
also treated with oral antibiotics throughout the
trial.' Another clinical trial comparing home-ba.sed
IV cefazolin plus probenecid with IV ceftriaxone reported
a treatment failure rate of 14% in the cefazolin
arm and 4% in the ceftriaxone arm.' In that
study, a large proportion of patients had an "indeterminate
clinical outcome" in both arms. Ifthese indeterminate
patients had been considered treatment
failures, then failure rates would have been 32.5% in
the cefazolin group and 30% in the ceftriaxone group.
Our study is the first to examine the interrater agreement
of the characteristics of celluUtis. The agreement between
physicians on these characteristics is very strong when objective
measurements are used (such as fever or diameter
of erythema), and the agreement appears to be poor to
moderate when physicians are asked for a subjective clinical
impression (such as an impression of severity). This
finding demonstrates the need to record the objective manifestations
of these infections and to clarify the features of
severity and the admission requirements.
Our study has some important limitations. Due to the
availability of the research nurses, patients were not enrolled
during night shifts and weekends. Two-thirds of our
patients were treated with IV antibiotics, suggesting that
many patients treated with oral antibiotics were not enrolled.
Enrolment was slower than expected, and even with
extension of the study time frame, we still did not meet our
target of 100 patients. There is the potential for selection
bias and, consequently, an overestimation of the treatment
failure rates. Our small sample size was not adequate for a
multivariate analysis. Some of the features that appeared to
be associated with treatment failure did not reach statistical
significance in our univariate analysis, and this may also
be due to the small sample size.
Despite these limitations, this is the first ED study to
prospectively identify specific patient and infection characteristics.
Our definition for treatment failure was also
prospectively defined and implemented after an extensive
review of the cellulitis literature. There have not been any
previous attempts to examine the interrater reliability of
the features of cellulitis. and although our sample size was
small, we have been able to show disagreement between
physicians on many of the more subjective features of cellulitis.
These fmdings will be used to assist with the design
of future research to establish which patients should be eligible
and will benefit most from treatment with ED-based IV antibiotics.
The treatment of cellulitis with daily ED-based IV antibiotics
has a treatment failure rate of more than 25% in our
centre. A clinical trial of this practice is needed to determine
which patients may benefit from more intensive IV
therapy, as is possible with admission. Older patients and
patients with previous (failed) oral therapy and those infections
over a larger area are more likely to fail ED treatment
for cellulitis. Physicians show high interrater reliability for
the objective findings of skin and soft-tissue infections
(such as fever and estimated size) but poor interrater reliability
for subjective decision (such as infection severity and
need for hospital admission). Further research should also
be focused on identifying infection characteristics for clinical
decision-making that are reliable and reproducible.
[ABSTRACT FROM AUTHOR]