American Journal of Emergency Medicine (2013) 31, 360-364

Original Contribution

Identifying patients with cellulitis who are likely to require inpatient admission after a stay in an ED observation unit^?,??

Kathryn A. Volz MD?, Louisa Canham MD, Emily Kaplan MD, Leon D. Sanchez MD, MPH, Nathan I. Shapiro MD, MPH, Shamai A. Grossman MD

Department of Emergency Medicine, Beth Israel Deaconess Medical Center, Boston, MA 02215, USA

Received 29 June 2012; revised 27 August 2012; accepted 5 September 2012

Abstract

Background: Emergency department Observation Units (EDOU) are often used for patients with cellulitis to provide Intravenous antibiotics followed by a transition to an oral regimen for discharge. Because institutional regulations typically limit EDOU stays to 24 hours, patients lacking a Clinical response within this period will often be subsequently admitted to the hospital for further treatment. Objective: The aim of this study was to determine the rate of hospital admission and characteristics predictive of admission in patients with cellulitis who are initially placed in an ED observation unit. Methods: A retrospective cohort study of patients placed into EDOU with a diagnosis of skin infection was conducted. Age, sex, history of diabetes mellitus, immunosuppression, intravenous drug use, location of cellulitis, presence of abscess, laboratory infectious markers, vital signs, and outpatient antibiotic treatment were recorded. The primary outcome was a hospital admission due to failure to respond to treatment within the 24-hour observation time window. Significant variables on univariate analysis were used to create a multivariate analysis, which identified predictive characteristics.

Results: Four hundred six patient charts were reviewed, with 377 meeting inclusion criteria; the inpatient admission rate from EDOU was 29.2%. Using logistic regression techniques, we created a model of independent predictors for need of admission after 24 hours: cellulitis of the hand (odds ratio [OR], 2.9; 95% confidence interval [CI], 1.8-4.9), measured temperature higher than 100.4?F (OR, 2.5; 95% CI, 1.1-5.5), and lactate greater than 2 (OR, 3.1; 95% CI, 1.3-7.3) were predictive of failure of ED observation.

Conclusions: Patients with cellulitis placed into ED Observation status were more likely to fail an observation trial if they had an objective fever in the ED, an elevated lactate, or a cellulitis that involved the hand.

(C) 2013

^? No financial support was obtained for this investigation.

^?? Prevoiusly presented: Poster presentation at SAEM 2012 Annual Meeting.

* Corresponding author. Beth Israel Deaconess Medical Center,

Department of Emergency Medicine, West Campus Clinical Center, Boston, MA 02215. Fax: +1 617 754 2350.

E-mail address: [email protected] (K.A. Volz).

Introduction

In an era where health care costs are increasingly scrutinized, and with a growing need to limit health care expenditure, emergency department observation units

0735-6757/$ - see front matter (C) 2013 http://dx.doi.org/10.1016/j.ajem.2012.09.005

(EDOU) are more commonly being used as an alternative to inpatient hospital admission in select patients, with strict observation admission criteria attempting to ensure that patients placed into observation status are likely to be dis- charged home within 24 hours. Common diagnoses admitted to EDOU include chest pain, asthma, congestive heart failure, syncope, and cellulitis. Given that observation unit resources are finite, it is important to appropriately identify patients for this disposition to maximize use, efficiency, and cost-effectiveness.

Cellulitis or skin infection is noted in prior studies to represent 1% to 5% of all ED observation patients [1-3]. Patients are commonly placed into ED observation status for

IV antibiotics and then discharged on oral antibiotics. Previous literature suggests that admission rates to the hos- pital after an ED observation stay for cellulitis vary between 15 and 38% [3,4], whereas the national average for admis- sion of all ED observation patients is closer to 18% [5]. There are few studies that help physicians predict which patients with cellulitis are likely to fail a trial of EDOU. Because it remains unclear which characteristics can reliably predict which patients with cellulitis require prolonged hospitaliza- tions, this study was designed to identify characteristics in patients with cellulitis that are predictive of EDOU failure.

Methods

A retrospective cohort study of consecutive patients presenting to the ED of an urban academic level 1 trauma center with more than 55000 annual visits, with an annual ED observation volume of 5600 patients per year, was conducted. The project was approved by the hospital insti- tutional review board. Consecutive patients admitted to the observation who were older than 18 years during the study period of August 2009 to May 2011 with a diagnosis of celllulitis or skin infection were included. These patients were identified by final ED diagnosis using International Classification of Diseases, Ninth Revision codes 680-686.

Patients were excluded if they had a diagnosis of cellulitis but were discharged directly from the ED or admitted to an Inpatient unit without a stay in ED observation, as this was beyond the scope of this study. Patients were excluded further from analysis if they were placed in observation primarily for other reasons, they were erroneously placed into observation, the chart was missing, the patient did not receive IV antibiotics during the EDOU stay, or the patient left against medical advice, eloped, or was transferred.

A chart review was performed by trained reviewers, and characteristics including patient age, sex, presence of diabetes, intravenous drug use, immunocompromised status (defined by the presence of HIV/AIDS or use of immuno- suppressant medications), subjective history of fever, loca- tion of cellulitis, previous treatment with oral antibiotics, type of oral antibiotics received, and final disposition

(subsequent hospital admission vs Discharge home from the observation unit) were recorded. Vital signs during the ED stay including heart rate, blood pressure, and tempera- ture, as well as laboratory results including white blood cell count and lactate level, if sent, were noted. Presence of abscess with incision and drainage performed was also recorded. Although data including presence of crepitus, bullae, size of cellulitis, and length of symptoms were also collected when available, these characteristics were not included in our analysis because they were not consistently documented in patient charts.

When evaluating patients with an outpatient antibiotic course, we considered only those who had received at least 1 dose of oral antibiotics that reviewers deemed directed against cellulitis. Patients who had received 1 dose of IV antibiotics and then were transferred to our tertiary care center were not considered to have been treated with outpatient antibiotics. We also characterized the patients’ outpatient antibiotics as methicillin resistant staph aureus (MRSA) appropriate or non- MRSA appropriate. Patients treated with oral trimethoprim/ sulfamethoxazole, doxycycline, or clindamycin were deemed to have received MRSA coverage.

Descriptive statistics including confidence intervals (CIs) are reported where appropriate. Continuous covariates were turned into dichotomized variables at standard accepted cutoffs as determined a priori. For analysis, we first per- formed a univariate analysis comparing admission rates for each covariate of interest using Fisher exact test, t test, or Wilcoxon rank sum, as appropriate. Variables with unad- justed univariate significance of at least 0.2 were eligible for inclusion into our multivariate model. Next, we built a stepwise, forward selection, multivariate logistical regression model, which retained covariates reaching a P b .05. We adjusted for missing data with imputation using previously described techniques [6]. We reported the area under the receiver operator characteristic curve as a measure of model accuracy. For Power calculation, we powered the study to yield a model to support 5 predictors. Using the guideline of 20 outcomes per predictor in a multivariate model and a 35% admission rate, we estimated that at least 350 patients would be required. We assessed a period of 22 months. Data were entered into a Microsoft Excel 2003 (Redmond, WA) database. Data were analyzed using SASm, version 8.

Results

A total of 406 consecutive patients with a diagnosis of skin infection/cellulitis were identified. Twenty-nine patients were excluded: 10 had been observed for other reasons (chest pain, psychiatric disease, alcohol intoxication, or need for social work); 8 did not receive IV antibiotics; 5 were erroneously placed into observation status; 5 left against medical advice, eloped, or were transferred to another institution; and 1 had a missing chart. Of 377 patients who met study criteria and were

included in the final analysis, 110 (29.2%) were ultimately admitted from EDOU to an inpatient unit. Of the 377 patients, 45.9% were female. The average age of the study group was 46.1 years. There was no significant difference in average age of admitted or discharged patients (47.2 vs 45.6 years; P = .39). The admission rate did not differ based on sex, presence of diabetes, intravenous drug use, or immunocompromised status (see Table 1). One hundred eighteen patients were noted to have abscesses associated with their cellulitis necessitating Incision and drainage procedures. Of these, there was no difference in the rate of admission; 28.0% of incision and drainage patients were admitted vs 29.7% of patients without abscess (P = .73).

We found 113 (30.0%) patients were admitted to ED observation with cellulitis of the hand vs 24 (6.4%) torso/ buttock, 49 (13.0%) head/neck, 69 (18.3%) arm, 35 (9.3%) foot, and 87 (23.1%) leg (Table 2). There was a higher admission rate for Skin infections of the hand (43.4%) vs other body locations (23.1%; P b .001). Admission rates varied by body location, including torso/buttock (25.0%), head/neck (30.6%), arm (23.2%), foot (20.0%), and leg (19.5%).

There was no significant difference in admission rate of patients who reported fever by history (30.4%) vs those who denied history of fever (29.0%; P = .83). With regard to vital signs in the ED, patients with documented temperature higher than 100.4?F were more likely to be admitted (45.2% vs 28.2%; P = .048). Similarly, those with heart rate higher than 100 (39.5% vs 26.8%; P = .035) during their ED course were also more likely to be admitted from obser- vation. Blood pressure in the ED did not appear to sig- nificantly contribute to admission rates, as 41.7% of patients with a systolic blood pressure lower than 100 during their ED stay were admitted as opposed to 28.8% of patients whose systolic blood pressure remained higher than 100 mm Hg (P = .33).

A complete blood count was obtained on 89.4% of patients with cellulitis placed in observation status, whereas 50.4% of all observation patients had lactate levels drawn. In those patients who did have a WBC count sent, there was no significant difference between those with WBC greater than

Location	No.	(%)	Admitted (%)
Hand	113	(30)	43.4
Torso/buttock	24	(6.4)	25.0
Head/neck	49	(13)	30.6
Arm	69	(18.3)	23.2
Leg	87	(23.1)	19.5
Foot	35	(9.3)	20.0

11 K/uL (35.9% admission rate) vs WBC less than 11 (27.5% admission rate; P = .13). However, when comparing lactate levels, 50.0% of patients with a lactate greater than 2 mmol/L were admitted vs a 27.9% admission rate for those with lactate less than 2 (P = .02).

Table 2 Admission rate based on cellulitis location

In total, 133 (35.2%) patients were documented as having received some course of oral antibiotics as an outpatient before their EDOU stay. These patients did not have a higher rate of EDOU failure (31.6% rate of admission vs 27.9% in those who had not received PO antibiotics; P = .47). Of patients treated with prior anti- biotics, 61% had received MRSA coverage. Patients treated with oral MRSA-appropriate antibiotics did not have different admission rates than those without MRSA cover- age (28.6% vs 36%; P = .37).

A multivariate analysis was then performed. After ad- justing for multiple factors, objective temperature (odds ratio [OR], 2.5; 95% CI, 1.1-5.5), lactate greater than 2 (OR, 3.1; 95% CI, 1.3-7.3), and location of infection on the hand (OR, 2.9; 95% CI, 1.8-4.9) remained predictive of the need for admission; however, the model demonstrated only fair discrimination with an area under the receiver operating characteristic curve of 0.65. If a patient had one of these risk factors, the admission rate was 42.1% vs 19.7% in those without any independent risk factors (Table 3).

Discussion

Although traditional teaching suggests that factors such as age and underlying comorbidities such as diabetes and

Table 1 Percentage of patients requiring inpatient admission

Table 3 Multivariate model for independent predictors of admission

This table represents the final logistic regression model that retained 3 covariates and had an area under the curve (c statistic) of 0.65 showing only fair discriminatory ability. Missing lactate was forced into the model such that patients with a missing lactate had a 0.86 (0.52-1.42; P = .65) risk of cellulitis.

Risk factor	Yes (%)	No (%)	P
Location in hand	43.4	23.1	.001
Lactate N2	50.0	27.9	.018
Heart rate N100	39.5	26.8	.035
Documented temperature N100.4	45.2	28.2	.048
WBC N11	35.9	27.5	.13
Blood pressure b100	41.7	28.8	.33
IVDU	41.7	28.8	.33
Immunocompromised	35.1	28.5	.4
Diabetes	26.7	29.5	.69
Presence of abscess	28.0	29.7	.73
Subjective fever	30.4	29.0	.83

Risk factor	OR (95% CI)	P
Temperature N100.4	2.5 (1.1-5.5)	b.03
Lactate N2.0 mmol/L	3.1 (1.3-7.3)	b.01
Cellulitis of the hand	2.9 (1.8-4.9)	b.001

peripheral vascular disease would predict worse Short-term outcomes (including need for admission) in patients with cellulitis, recent studies have not supported these pre- sumptions. Shrock [4] reported that only Female gender and a WBC greater than 15000 K/uL were predictive of the need for admission, whereas factors such as diabetes, IV drug use (IVDU), age, and presence of medical insurance did not contribute to rates of EDOU failure in that study. In contrast, Sabbaj et al [7] found that only the presence of fever (either in the ED or by patient report) was predictive of the need for admission of more than 24 hours. A study by Roberts [8] suggested that certain high-risk categories such as patients with severe pain; bite wounds; foreign bodies; orbit, neck, scrotum, and hand involvement; or peripheral vascular disease be excluded for consideration of ED observation. Female gender, reported fever by history, and an elevated WBC count did not prove to be predictive of admission in our study, as they have in others.

Our data suggest that patients with an elevated temper- ature in the ED and elevated lactate levels are likely to require admission. These markers may be evidence that the infection has progressed systemically, and thus, these patients may be more ill and will require a longer hospital treatment period. We believe that providers should consider directly admitting patients who have these signs of Systemic Infection to an inpatient ward because they are less likely to be successfully treated and sent home within a 24-hour ED observation period.

Hand infections were also much more likely to fail ED observation, with a rate of 43.4% requiring inpatient ad- mission. This suggests that providers should also consider directly admitting patients with hand cellulitis to maximize the efficiency of ED observation. Of note, we had a low rate of lower extremity cellulitis placed into observation, although it was the most common location of cellulitis in at least 1 study [9]. In patients with diabetes, this is an area that is commonly associated with a severe infection. We hypothesize that providers already bypass observation and place patients such as lower extremity cellulitis or the most ill appearing of patients with cellulitis directly into full inpatient status. Further studies should evaluate patients with cellulitis who are directly admitted to the inpatient unit and evaluate inpatient length of stays to see if ED providers are appropriately identifying patients who need a full in- patient admission.

We found no significant difference in failure of obser- vation by sex as opposed to the previous study by Shrock [4]. We believe that sex is unlikely to represent a true risk factor of failure of ED observation. We found that patients’ report of fever at home not predictive of the need for admission; however, we did find objective fever, as measured in the ED, to be significant in our analysis. Subjective fever ultimately may prove to be a predictive factor, as discrepancies regarding the ability of patients to report a fever history may be related to patients’ interpretation of what is a fever at home. We evaluated an abnormal WBC count (N11000 K/

uL at our institution) and did not find this variable to be predictive of admission, although Shrock’s study found a WBC count higher than 14000 to be significant [4]. It may be that markedly abnormal WBC counts are predictive as similar to objective fever and elevated lactate, and this may be evidence of systemic infection; however, our study was underpowered to demonstrate this discrepancy.

Along these lines, we found no significant difference in a number of the other factors that logically might be expected to be predictors of admission, including age, sex, diabetes, immunosuppression, IVDU, presence of abscess, patient history of fever, and blood pressure lower than 100. Additional investigations will further clarify the impact of these factors on the need of prolonged hospital admission in cellulitis.

Lastly, we also looked at recent outpatient antibiotic use among patients in this study group, hypothesizing that failure of outpatient antibiotics might predict the need for a longer admission and thus EDOU failure. However, we did not find that previous treatment with oral antibiotics predicted a failure of ED observation. There was also no difference in the rate of ED observation failure based on previous coverage with an MRSA-appropriate regimen. These comparisons are limited because choice of outpatient antibiotics and number of doses of oral antibiotics were not controlled. Further studies with control of type and length of outpatient anti- biotics may better evaluate this question.

Limitations

Limitations of our retrospective study include enrollment bias and incomplete records. We did attempt to look at factors that could be important, such as size, length of symptoms, and characteristics such as bullae and crepitus; however, these were not consistently documented in charts, and we were therefore unable to include them in our data analysis. Not every patient in our study had a WBC count or lactate drawn. Physicians may be more likely to order these tests on patients who appear sicker, thus creating bias. Further limitations include that failure of observation was defined as admission from EDOU to the inpatient unit, which was the decision of the treating provider, as was the initial decision to admit the patient to EDOU. Individual practice patterns may be inconsistent and may influence results. Similarly, the decision of which antibiotic is to be used during a patient’s ED observation stay was the decision of the individual provider. We did not compare admission rates based on different antibiotic regimens used during the EDOU stay. We did not follow up the patients’ inpatient course or length of admission or evaluate for patients who returned to the ED within 24 hours of being discharged. As a single- center study, these results may be unable to be generalized to other institutions. We did not evaluate patients who were diagnosed with cellulitis and discharged or directly admitted

to the inpatient units because this was beyond the scope of the study. Further prospective studies are warranted. Further studies should look at all patients with a diagnosis of cellulitis in the ED who are admitted, discharged, and placed into EDOU to identify patients with less than 24-hour inpatient admissions as well as those who fail outpatient treatment to evaluate the patient population best served with a stay in an EDOU.

Conclusions

In this study group, almost one third of patients admitted to the EDOU with cellulitis ultimately were admitted to an inpatient unit. We found no significant difference in admission rates based on age, sex, diabetes, immunosup- pression, IVDU, presence of abscess, elevated WBC count, patient history of fever, blood pressure lower than 100, or prior course of PO antibiotics. Patients with hand infections had the highest admission rates compared with other body locations. Patients with fevers while in the ED or an elevated lactate were also more likely to fail ED observation. Cellulitis is more likely than other EDOU diagnoses to require inpatient care; therefore, decisions to admit patients based on signs of systemic infection such as documented

fever, lactate levels, or location on the hand may result in a more judicious use of the ED observation unit.

References

1. Venkatesh AK, Geisler BP, et al. Use of observation care in US emergency departments, 2001 to 2008. PLoS One 2011;6(9):e24326.
2. Ross MA, et al. The use and effectiveness of an emergency department observation unit for elderly patients. Annals Emerg Med 2003;41: 668-77.
3. Martinez E, et al. The observation unit: a new interface between inpatient and outpatient care. Am J Med 2001;110:274-7.
4. Shrock JW, et al. Predicting observation unit treatment failures in patients with skin and soft tissue infections. Int J Emerg Med 2008;1: 85-90.
5. Wiler JL, et al. National study of emergency department observation services. Acad Emerg Med 2011;18:959-65.
6. Greenland S, Finkle W. A critical look at methods for handling missing covariates in epidemiologic regression analysis. Am J Epidemiol 1995; 142:1255-64.
7. Sabbaj A, et al. soft tissue infections and emergency department disposition: predicting the need for inpatient admission. Acad Emerg Med 2009;16:1290-7.
8. Roberts R. Management of patients with infectious diseases in an emergency department observation unit. Emerg Med Clin North Am 2001;19:187-207.
9. Simonsen SM, et al. Cellulitis incidence in a defined population. Epidemiol Infect 2006;134:293-9.