Diagnostic accuracy in patients admitted to hospitals with cellulitis
- Author(s): David, Consuelo V
- Chira, Sandy
- Eells, Samantha J
- Ladrigan, Manasi
- Papier, Art
- Miller, Loren G
- Craft, Noah
- et al.
Diagnostic accuracy in patients admitted to hospitals with cellulitis1. Los Angeles Biomedical Research Institute at Harbor-UCLA Medical Center, Torrance, California
Consuelo V David1, Sandy Chira1, Samantha J Eells1, Manasi Ladrigan2, Art Papier2*, Loren G Miller1,3, Noah Craft1,3*
Dermatology Online Journal 17 (3): 1
2. Univ. of Rochester School of Medicine and Dentistry, Rochester, New York
3. David Geffen School of Medicine at the Univ. of California, Los Angeles, Los Angeles, California. email@example.com
Misdiagnosis of non-infectious conditions as cellulitis is a common error and can result in unnecessary hospitalization and antibiotic use. We sought to prospectively determine the misdiagnosis rate of cellulitis among hospitalized patients and to determine if a visually-based computerized diagnostic decision support system (VCDDSS, also named VisualDx) could generate an improved differential diagnosis (DDx) for misdiagnosed patients. In two separate institutions, attending dermatologists or infectious disease specialists evaluated all consecutive patients hospitalized for “cellulitis” by the emergency department. Among 145 subjects enrolled, misdiagnosis occurred in 41 (28%) patients. The diagnosis most commonly mistaken as cellulitis was stasis dermatitis (37%). At one center, in cases that were misdiagnosed by the emergency department, the VCDDSS included the correct diagnosis in the DDx more frequently than the admitting team (18/28 cases (64%) compared to 4/28 cases (14%), p=0.0003). These results demonstrate the capability of this VCDDSS to assist primary care physicians with generating a more accurate DDx when confronted with patients presenting with possible skin infections. Misdiagnoses may result in a significant source of healthcare costs and misdiagnosis-related patient harm. Inclusion of decision support tools early in the diagnostic workflow may reduce misdiagnosis and result in more efficient healthcare management.
Improving healthcare information technology and reducing medical errors are important initiatives receiving increased national attention. These initiatives have potential to improve patient safety. Most media coverage is directed at medication errors and electronic healthcare record implementation. However, approximately 40 – 80 thousand hospital deaths occur each year because of diagnostic errors . The resulting healthcare costs, lost income, lost household productivity, and disability costs from medical errors are estimated to cost between $17 and $29 billion annually in the United States . Diagnostic errors do not always result in serious injury, but the magnitude of misdiagnosis-related harm and the financial costs are great. Recent studies have demonstrated the tendencies for diagnostic error propensities to develop in senior residents  and also the ability of educational interventions to improve on known processes that lead to error in medical students .
Cellulitis is a common infectious process affecting the skin and subcutaneous tissues. Along with other skin and soft tissue infections, it results in significant morbidity and exacts considerable healthcare costs. More than $3.7 billion were spent on approximately 240,000 adult inpatient admissions for cellulitis in the United States in 2004 . Outpatient and Emergency Department (ED) visits for skin and soft tissue infections add to this impressive cost. In 2005, approximately 14.5 million persons visited physician offices, hospital outpatient departments, and EDs for skin infections in the United States. This number represents an approximately 50 percent increase over that in 1997 . In spite of this enormous burden, there is a paucity of data helping clinicians to differentiate cellulitis from other conditions that have similar clinical presentations. At the present time, evidence-based clinical guidelines for diagnosis of cellulitis are lacking.
Overdiagnosis of cellulitis can result in the unnecessary costly hospitalizations and potentially dangerous use of intravenous antibiotics that may further fuel the emergence of antibiotic resistant bacteria. Underdiagnosis of cellulitis, though seemingly uncommon, can result in worsening infection and serious complications if not treated promptly. Thus, the need for improved diagnostic accuracy of cellulitis is clear and a structured approach to the diagnostic process may decrease misdiagnoses and improve quality of care.
In a previous pilot study, 15 consecutive admissions for cellulitis from the University of Rochester Strong Memorial Hospital’s ED showed 100 percent diagnostic concordance between dermatologists and infectious disease physicians. However, compared to the admission diagnosis of cellulitis, examination by these specialists revealed a 20 percent misdiagnosis rate by the ED. To further address the accuracy of cellulitis diagnosis, we performed investigations at two major medical centers. The goals were to quantify the magnitude of the problem of cellulitis misdiagnosis, determine which diseases commonly mimic cellulitis, and define factors that could improve diagnostic accuracy.
Cognitive science studies have previously demonstrated that including the correct diagnosis in the DDx within the first 6 minutes of cognitive processing greatly increases the likelihood of arriving at the correct diagnosis . Thus, we sought to evaluate the performance of a visually-based, computerized diagnostic decision support system (VCDDSS), called VisualDx™, to quickly generate a robust and accurate DDx based on a patient’s individual presenting signs and symptoms.
Materials and methods
The primary aim of this investigation was to quantify diagnostic errors among physicians admitting patients to the hospital for uncomplicated “cellulitis.” This aim was accomplished by comparing the admitting diagnosis of “cellulitis” (made by the ED attending physicians) with the expert opinion of either a dermatologist or an infectious disease specialist. Because there was 100 percent concordance between dermatologists and infectious disease doctors in our pilot study, the complete study was not designed to compare inter-rater variability between different specialties (since this would have required two separate visits by specialist attendings for every subject). Two separate studies were performed prospectively at two separate medical centers and both sets of results are presented here in aggregate where appropriate. In both centers, dermatologic evaluation occurred within 24 hours of admission. Evaluation by the specialist consultants was performed independently from the ED or admitting team staff members. In one hospital (Harbor-UCLA), the opinion of the admitting team was also compared. Cases where the final diagnosis made by the specialist physician (the “specialist’s diagnosis”) was cellulitis or erysipelas were labeled “true cellulitis.” Cases given an alternative diagnosis by the specialists were termed “non-cellulitis.” Specialists’ opinions were not considered in management of these subjects unless a consult request was made independently by the primary admitting team.
B. Data collection details
Harbor-UCLA Medical Center is a public county hospital affiliated with the University of California Los Angeles School of Medicine located in Torrance, California. Strong Memorial Hospital is a private hospital associated with the University of Rochester and located in Rochester, New York. We identify throughout the text any small protocol differences between the hospitals.
Patients were admitted through the Adult and Pediatric EDs. Patients of all ages were included in the study. Patients with potential post-operative surgical site infections were excluded. A study coordinator identified and screened patients with an admission diagnosis of cellulitis made by the ED physician. Informed consent was obtained. The study coordinator: a) photographed the affected area of skin, b) obtained a brief medical history, c) administered a survey regarding cellulitis risk factors, d) collected laboratory data and e) coordinated a dermatology evaluation for the patient.
Evaluation by the specialist occurred within 24 hours of admission, but 15 patients were excluded because this was not possible in the time limits. In Rochester, patients were evaluated in the Observation Unit and at Harbor-UCLA, they were admitted to the inpatient wards. A Dermatology or Infectious Disease attending performed a physical exam and reviewed relevant clinical data. The leading diagnosis and a DDx list were recorded. Index of certainty of cellulitis as well as an opinion regarding whether hospitalization and parenteral antibiotics were warranted were graded on a Likert-type scale. Results were dichotomized into “agree” or “disagree” where appropriate.
For comparison, at Harbor-UCLA, the admitting senior resident physician was also asked to identify a primary diagnosis (decide if they agree with the admitting diagnosis of cellulitis from the ED) and list a DDx. Residents also input characteristics of the patient’s lesions, medical history, or hospital course (e.g., gender, anatomic location, unilateral vs. bilateral, febrile vs. afebrile) into the VCDDSS, VisualDx™. A portable tablet computer and a unique module designed to assist in cases of suspected cellulitis were used for each patient. Clinical criteria were input and a DDx was generated and recorded. Residents were compensated with a $5.00 gift certificate.
C. Institutional Review Board
Both studies were performed under the approval of their respective institutional review boards. The EDs were unaware of patient recruitment to minimize any potential Hawthorne effect and therefore, alterations in clinical practice.
D. Data analysis
Data management and analysis was performed using SAS (version 9.1.3; SAS Institute, Cary, NC). Bivariate analysis was used to compare variables from the medical history, evaluation, and risk factor questionnaire hypothesized to be associated with “true cellulitis.” Odds ratios (OR) and 95 percent confidence intervals (CI) were calculated using the chi-square test and the associated p-values, accounting for multiple analyses, were calculated using the Fisher’s exact test. For proportions, the 2-proportion z-test was used to calculate the p-values. All variables were considered significant at the α= 0.05 level.
A. Overall diagnostic outcomes
Specialists evaluated 145 patients as part of the study, 80 at Harbor-UCLA and 65 in Rochester. Overall, 41 patients (28%) were given a diagnosis other than “cellulitis” by attending specialists. At Harbor-UCLA, 52 of the 80 cases (65%) admitted for “cellulitis” were diagnosed as “true cellulitis” by a dermatology or infectious disease attending. In Rochester, 52 of the 65 patients (80%) were diagnosed with “true cellulitis.” An example of a patient with true cellulitis is shown in Figure 1a.
The most common alternate diagnoses provided by the dermatologist or infectious disease attendings in this study were stasis dermatitis, (37%, 15/41), trauma related inflammation (5%, 2/41), deep vein thrombosis (5%, 2/41), non-specific dermatitis (5%, 2/41), and thrombophlebitis (5%, 2/41) (Table 1). Examples of bilateral stasis dermatitis mimicking cellulitis are shown in Figure 1b (acute on chronic stasis dermatitis) and Figure 1c (acute stasis dermatitis). An example of contact dermatitis mimicking cellulitis is shown in Figure 1d. Of the four patients admitted with bilateral “cellulitis,” all four were considered to be stasis dermatitis by a specialist in this study.
B. Computer-assisted Differential Diagnosis (DDx)
Prior studies in cognitive science demonstrate that a physician’s ability to include the correct diagnosis in the DDx is critical to ultimately arriving at the correct diagnosis. Indeed, the single best predictor of diagnostic success was the occurrence of the correct diagnosis as a hypothesis early in the encounter . This means that when considering multiple diagnoses, if the working DDx of a physician does not contain the correct diagnosis at an early time point, subsequent cognitive processes to confirm any diagnosis may be flawed or misguided. At Harbor-UCLA we sought to compare the DDx generated by the admitting teams with that of the specialists. The senior residents from the admitting service were asked to identify their top diagnosis. The admitting residents indicated an alternative top diagnosis for 6 of the 80 admitted patients (8%). The alternative top diagnoses suggested were stasis dermatitis (three times), diabetic ulcer (once), bursitis (once) and gangrene (once).
Both the residents and the specialists were also asked in all cases to generate a DDx for each patient. Disregarding the diagnosis of cellulitis, the DDx of the specialist attending and admitting resident did not contain any of the same alternative diagnoses in 59 percent of all cases (30% of cases misdiagnosed as cellulitis and 77% of true cellulitis).
Because a robust DDx is known to improve the final diagnostic accuracy, we sought to determine the potential for a visually-based computerized diagnostic decision support system (VCDDSS, trade named VisualDx™) to suggest alternative diagnoses based on available clinical data (lesion morphology, symptoms, etc.). A separate module in the VCDDSS called the “Cellulitis DDx Module” was used for this study. Before using the VCDDSS, the admitting residents were asked to list a DDx for the patient. Of the 28 cases misdiagnosed at Harbor-UCLA, the admitting residents included the correct diagnosis in the DDx only 4 times (14%). Then, they were asked to enter pertinent patient findings into the “Cellulitis DDx Module” so the VCDDSS could generate a DDx list. They were not guided during this process. Overall, the residents from the admitting team entered an average of 3.8 patient findings (range 3-6). Compared to the admitting primary care physician, the VCDDSS more frequently included the correct diagnosis in the DDx (70/80 patients or 88%, p=0.007). Among non-cellulitis cases (n=28), the VCDDSS included the alternative correct diagnoses in 18 cases (64%, e.g. stasis dermatitis, thrombophlebitis, or vasculitis) compared to 4/28 (14%) by the admitting physician (p=0.0003). Ten (36%) of the non-cellulitis alternative diagnoses were absent from the DDx developed by the VCDDSS.
The reason for the absence of these diagnoses in the DDx was explored further: six of the diagnoses were listed within the VCDDSS but part of other modules that were not accessed as part of this study (e.g., International Travel Module or Child Abuse Recognition Module); two of the diagnoses were described in the VCDDSS but the location of the lesion (e.g. stasis ulcer on the foot) was not listed as a characteristic location of the diagnosis and was therefore excluded; two of the diagnoses were not included in the VCDDSS as distinct diagnostic entities (erosio interdigitalis blastomycetica and diabetic muscle infarction). Perhaps obviously, in patients with “true cellulitis,” cellulitis was included 100 percent of the time in the DDx generated by the VCDDSS and by the admitting residents.
In summary, when considering cases misdiagnosed and admitted to the hospital as cellulitis by the ED, the VCDDSS included the correct diagnosis in the DDx 64 percent of the time compared with the admitting residents, who included the correct diagnosis in the DDx only 14 percent of the time without using the VCDDSS (p=0.0003).
C. History and clinical findings
At the University of Rochester, the average length of time from the initial manifestation of the disease process to emergency department presentation was 2.8 days for true cellulitis cases and 5.5 days for non-cellulitis cases (p=0.01). Thirty-three percent of those with true cellulitis gave a history of fever versus 25 percent of those with other diagnoses (p=0.71). However, only one person in the entire study had a measurable fever. Thirty-one percent of those with true cellulitis gave a history of recent trauma at the site of infection vs 25 percent of non-cellulitis cases (p=0.85). No significant differences were detected between groups for any clinical features measured (Table 2). However, misdiagnosed patients were less likely to have a leukocytosis (24% vs 55% in true cellulitis cases, p=0.001).
D. Antibiotic Administration
Overall, 26 percent of patients had previously used antibiotics at home, but there was no significant difference detected between true cellulitis patients and non-cellulitis patients. All patients admitted with suspected cellulitis in this study received intravenous antibiotics. Clindamycin was the most commonly prescribed antibiotic among in both the ED and inpatient settings. Other common regimens were combined beta-lactam/beta lactamase inhibitors or trimethoprim-sulfamethoxazole (Table 3).
E. Diagnostic confidence and management outcomes
Because there was occasional disagreement between the ED, the admitting teams, and the specialist attendings, we sought to determine the confidence level of participants with their leading diagnoses. Recognizing that the risks involved in misdiagnosis by different health care providers could have important impacts on care, we sought to determine the diagnostic certainty of the participants and whether or not it would impact patient care.
At Harbor-UCLA, participating physicians were asked about their degree of diagnostic certainty. Among the 41 cases determined to be “non-cellulitis” by the specialists, the specialist attendings were “very certain” the patient did not have cellulitis 83 percent of the time (34 cases).
To determine the hypothetical impact of the specialist’s diagnostic intervention on the actual treatment, specialists were also asked about management recommendations. With these same 41 “non-cellulitis” patients, the specialist agreed with hospitalization in 39 percent of cases and agreed with the use of parenteral antibiotics in 44 percent of cases (Table 4).
The proportion of patients admitted for cellulitis that received a misdiagnosis of cellulitis at two institutions was 28 percent. The most common diagnosis of patients admitted to the hospital in error was stasis dermatitis.
Cellulitis is a clinical diagnosis and a potential medical emergency. Clinicians are given the challenging task of diagnosing a disease based on limited epidemiologic data and poorly characterized presenting symptoms. There is data to suggest that misdiagnosis of cellulitis is common. One group found that 13.6 percent of outpatients with a diagnosis of cellulitis did not have cellulitis . Others found in a retrospective chart review, that of 196 inpatient dermatology consultations requested over 2 years, 21 were for evaluation of “cellulitis.” Seven of these (33%) were misdiagnosed as dermatitis (4), herpes simplex (2), and drug eruption (1) . However, clinical and epidemiologic details as well as predictors of misdiagnosis in these studies were not explored. Our results correlate with these related studies and demonstrate that less acute, non-infectious pathologies are frequently misdiagnosed as cellulitis. These conditions usually can be effectively treated in an outpatient setting without administration of antibiotics. The result is unnecessary hospitalization, misallocation of resources and antibiotic administration. A systematic approach to the diagnosis of cellulitis may reduce the number of patients who receive a misdiagnosis.
Among cases misdiagnosed as cellulitis at Harbor-UCLA the DDx of dermatology/infectious disease attendings and internal medicine residents shared at least one diagnosis in 70 percent of cases. This suggests there is room for improvement in expanding or refining the DDx for non-specialists. Diagnosis improvement may be maximized if focus is shifted from diagnosing cellulitis to recognizing common mimickers (i.e., improving the DDx for inflamed erythematous skin). This study was not designed to address the underdiagnosis of cellulitis.
In our study, there were no clinical signs that were obviously associated with either cellulitis or the misdiagnosis thereof. However, lack of leukocytosis was associated with mimickers of cellulitis. Thus, the creation of guidelines to increase the accuracy of diagnosing cellulitis must focus on improving the DDx and consultation with subspecialists when the diagnosis is in question.
Proper and accurate diagnosis of the patient presenting with suspected cellulitis is based on clinical features and acquired expertise. There are no laboratory or imaging studies to confirm cellulitis. Therefore, tools such as clinical algorithms, diagnostic decision support, and guidelines must be developed to address this important diagnostic area and other areas where diagnostic accuracy could be enhanced before invasive tests are used. These diagnostic decision support tools can be based on the predictable nature of the diagnostic errors (i.e., cellulitis is only rarely bilateral, and yet bilateral stasis dermatitis of the legs was misdiagnosed 4 times at one institution in this study). Improved diagnostic accuracy of cellulitis may alleviate healthcare system financial burdens. From 1999 to 2005 in the United States, cellulitis infection as a primary diagnosis increased >25 percent annually from a total of 22,451 hospitalizations in 1999 to 87,500 in 2005 . One recent study in the Netherlands revealed that although only 1 in 14 of patients with cellulitis was hospitalized, these patients contributed to more than 80 percent of the total associated costs . The average hospital length of stay (LOS) for complicated cellulitis was 5.3 days at an average cost of $13,000. The mean hospital LOS for uncomplicated cellulitis was 3.6 days with and an average cost of $8,000 . Economic consequences attributed to productivity and work days lost compound this impressive cost.
When misdiagnosis leads to inappropriate antibiotic prescription, healthcare costs can increase and inappropriate antibiotic use can further drive the emergence of antibiotic resistant pathogens. Cephalexin continues to be the preferred oral antibiotic for uncomplicated cellulitis  but may have an increased therapeutic failure rate compared to other oral antibiotics . Increased media attention focused on the rise in methicillin resistant Staphylococcus aureus (MRSA) prevalence among skin pathogens has contributed to an increase in reported skin infections, outpatient soft tissue cultures, and antibiotic prescription for MRSA [6, 14, 15].
There are very few studies examining the causes of diagnostic error. In this study, emergency room physicians may have been influenced by representative heuristics (mental matching resulting in narrower DDx), availability heuristics (accepting a wrong diagnosis due to ease of recall), or premature closure (closing the diagnostic possibilities too early so as to never consider the correct diagnosis). The admitting teams may have been strongly biased for similar reasons and also suffer from anchoring bias because they were accepting patients with a pre-existing diagnosis of cellulitis from the ED physicians. Our demonstration that the misdiagnosis of other skin conditions as cellulitis is a common occurrence, suggests different approaches to improving diagnostic accuracy should be studied. In addition to educational effort, clinical diagnostic decision support systems (CDDSS) tools may be promising. As demonstrated in this study, providing a robust DDx through the use of a visually-based CDDSS has the potential to direct providers to more accurate diagnoses. The results from this study will inform and improve the functionality of future iterations of the VCDDSS we tested. VCDDSS systems can reduce diagnostic error due to affective and cognitive influences, utilize de-biasing strategies such as developing insight and awareness and reduce reliance on memory by providing diagnostic alternatives . Systems that combine appropriateness criteria with established practice guidelines can help manage healthcare demand . Recent studies have demonstrated the tendencies for recall errors in senior residents  and also the ability of educational interventions to improve on known processes that lead to error . The utility of VCDDSS to improve diagnostic accuracy at the educational level and at the point-of-care is worthy of further investigation. Since complex dermatologic conditions requiring inpatient dermatology consultations are only correctly diagnosed by the primary team 24 percent of the time , we believe this approach to improved diagnostic accuracy will be useful for a wide range of clinical scenarios.
There are several limitations to our investigation. First, we considered consultant dermatologists and infectious disease physicians as the gold standard of diagnosis. The accuracy of these physicians has never been determined. However, potentially more accurate gold standards (e.g., skin biopsy) would introduce considerable risk to study patients and would not be acceptable from an ethical standpoint. Second, the patient population studied represents only a subset of patients who are evaluated by the ED for the treatment of skin and soft tissue infections since many patients with other non-infectious skin problems are not admitted to the hospital. Finally, bias may be introduced because the difference in time between the two evaluations (emergency medicine and specialist) may allow for variable access to additional clinical information (response to therapy, etc.). Also, the DDx evaluation of the admitting service is biased because another physician recently diagnosed the patient with cellulitis in the ED.
Concerning the diagnosis of cellulitis, there is considerable discordance between specialist attending diagnosis and primary care or emergency medicine physician diagnosis. Often, non-cellulitis cases are non-infectious and do not require hospitalization or the use of parenteral antibiotics. Considering the financial and societal consequences of mismanagement of cellulitis, a solution for diagnostic improvement is critical. This study highlights the potential for a VCDDSS, VisualDx™, to aid in generating an accurate DDx for a commonly misdiagnosed condition, cellulitis. Medical misdiagnosis and the need to improve diagnostic accuracy are not unique to cellulitis. In the context of healthcare information technology, emphasis should be placed on studies documenting the extent of these problems and evaluating interventions for improvement.
*Art Papier, MD is the Chief Medical Information Officer and Noah Craft MD, PhD, is a consultant to Logical Images, Inc.,
the makers of VisualDx.
Acknowledgement: This work was supported in part by GCRC Grant M01-RR00425 National Center for Research Resources. S.E. supported in part by an unrestricted research grant from Fallon Medica.
References1. Newman-Toker, D.E. and P.J. Pronovost, Diagnostic errors--the next frontier for patient safety. JAMA, 2009. 301(10): p. 1060-2. [PubMed]
2. To Err is Human: Building a Safer Health System. 2000, Institute of Medicine.
3. Mamede, S., et al., Effect of availability bias and reflective reasoning on diagnostic accuracy among internal medicine residents. Jama, 2010. 304(11): p. 1198-203. [PubMed]
4. Schwartz, A., et al., An educational intervention for contextualizing patient care and medical students' abilities to probe for contextual issues in simulated patients. JAMA, 2010. 304(11): p. 1191-7. [PubMed]
5. The DRG Handbook Comparative Clinical and Financial Benchmarks. 2006, Evanston,IL: Solucient.
6. Hersh, A.L., et al., National trends in ambulatory visits and antibiotic prescribing for skin and soft-tissue infections. Arch Intern Med, 2008. 168(14): p. 1585-91. [PubMed]
7. Norman, G., Dual processing and diagnostic errors. Adv Health Sci Educ Theory Pract, 2009. 14 Suppl 1: p. 37-49. [PubMed]
8. Hepburn, M.J., D.P. Dooley, and M.W. Ellis, Alternative diagnoses that often mimic cellulitis. Am Fam Physician, 2003. 67(12): p. 2471. [PubMed]
9. Bauer, J. and M. Maroon, Dermatology inpatient consultations: a retrospective study. J Am Acad Dermatol, 2010. 62(3): p. 518-9. [PubMed]
10. Klein, E., D.L. Smith, and R. Laxminarayan, Hospitalizations and deaths caused by methicillin-resistant Staphylococcus aureus, United States, 1999-2005. Emerg Infect Dis, 2007. 13(12): p. 1840-6. [PubMed]
11. Goettsch, W.G., J.N. Bouwes Bavinck, and R.M. Herings, Burden of illness of bacterial cellulitis and erysipelas of the leg in the Netherlands. J Eur Acad Dermatol Venereol, 2006. 20(7): p. 834-9. [PubMed]
12. Cunha, B.A., Cephalexin remains preferred oral antibiotic therapy for uncomplicated cellulitis. Am J Med, 2008. 121(11): p. e13; author reply e15-6. [PubMed]
13. Madaras-Kelly, K.J., R. Arbogast, and S. Jue, Increased therapeutic failure for cephalexin versus comparator antibiotics in the treatment of uncomplicated outpatient cellulitis. Pharmacotherapy, 2000. 20(2): p. 199-205. [PubMed]
14. Hahn, W., et al., The effect of media attention on concern for and medical management of methicillin-resistant Staphylococcus aureus: a multimethod study. J Public Health Manag Pract, 2009. 15(2): p. 150-9. [PubMed]
15. Moran, G.J., et al., Methicillin-resistant S. aureus infections among patients in the emergency department. N Engl J Med, 2006. 355(7): p. 666-74. [PubMed]
16. Croskerry, P. (2005) Diagnostic Failure: A Cognitive and Affective Appraoch. Advances in Patient Safety 2, 215-254.
17. Jacobs, C.M., Managing demand using clinical decision support tools. Healthc Financ Manage, 1997. 51(7): p. 41-2. [PubMed]
18. Davila, M., L.J. Christenson, and R.D. Sontheimer, Epidemiology and outcomes of dermatology in-patient consultations in a Midwestern U.S. university hospital. Dermatol Online J, 2010. 16(2): p. 12. [PubMed]
© 2011 Dermatology Online Journal