 |
|
 |
|
 |
 |
|
 |
 |
 |
 |
 |
 |
 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ARTICLE |
Division of Emergency Medicine, Children's Hospital, Boston, Massachusetts; Department of Pediatrics, Harvard Medical School, Boston, Massachusetts
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
METHODS. We conducted a medical chart review of children 
19 years of age with fever and erythroderma who presented to an urban pediatric emergency department over 60 months. Historical, clinical, and laboratory data were abstracted from the medical chart.
RESULTS. Fifty-six patients with fever and erythroderma were studied. Eighteen percent of patients presented with hypotension. Thirty-three percent of the remaining patients who were normotensive on arrival developed shock. Fifty-two percent of patients with hypotension required vasopressor support. The most important predictors of developing hypotension after presentation were: age 
3 years, ill appearance, vomiting, glucose 110 mg/dL, calcium 8.6 mg/dL, platelets 300000/µL, elevated creatinine, polymorphonuclear leukocytes 80%, and presence of a focal infection. Among all patients studied, 4 variables were determined to be the most important predictors of developing toxic shock syndrome: age 3 years, ill appearance, elevated creatinine, and hypotension on arrival.
CONCLUSIONS. Overall, 45% of patients with fever and erythroderma developed shock, including 33% of those who were normotensive on presentation. Older age, presence of vomiting, identification of a focal bacterial source, as well as specific laboratory parameters can be used to help predict which patients are likely to have hemodynamic deterioration. Given the high rate of clinical decompensation, all of the patients with fever and erythroderma should be hospitalized, closely monitored, and managed aggressively.
Key Words: erythroderma • fever • toxic shock syndrome • Staphylococcus aureus • Streptococcus pyogenes
Abbreviations: TSS—toxic shock syndrome • CDC—Centers for Disease Control and Prevention • ED—emergency department • CI—confidence interval • IQR—interquartile range • s/p—status post • BUN—blood urea nitrogen • ALT—alanine aminotransferase • PMN—polymorphonuclear leukocyte
Erythroderma is an uncommon finding among febrile patients and may be a manifestation of a variety of illnesses, including infections, toxin-mediated disease, and drug-related reactions. The evaluation of fever in children has been well studied, but patients with fever and erythroderma are a special circumstance and challenge for the clinician. Some children who present with fever and erythroderma have rapid clinical deterioration or progress to toxic shock syndrome (TSS). At presentation, it is difficult to determine which patients will develop shock. No previous studies have investigated whether historical examination, physical examination, or laboratory findings can help predict clinical deterioration or the progression to TSS. The Centers for Disease Control and Prevention (CDC) definitions for staphylococcal and streptococcal TSS are listed in Tables 1 and 2.1
|
|
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
19 years of age with fever and erythroderma who presented to an urban pediatric emergency department (ED) between January 2001 and December 2005. Potential cases were identified through an electronic text search tool; physician medical charts of the ED visit were searched for the specific words or phrases. The text search methodology allows for wild-card characters and "fuzzy" logic to identify words or parts of words near other terms. The following terms were searched with variations using wild cards and logical statements: erythroderma, diffuse erythema, confluent erythema, toxic and shock (or hypotension), toxin and shock (or hypotension), staphylococcal and shock (or hypotension), and streptococcal and shock (or hypotension). As an example of the search strategy, a search for "eryth%" and "shock" (or "hypotens%") would identify all of the charts with the words erythematous, erythema, erythroderma, erythrocytes, and erythromycin, as well as the word shock, hypotension, or hypotensive (even if the chart reads "no signs of shock" or "never hypotensive"). Patients were included if they met the following criteria: fever of 38°C by history or triage temperature and erythroderma defined as a diffuse skin erythema or "sunburn-like erythema" that was not evanescent or patchy. Rashes that were described as scarlatiniform, "sandpaper," maculopapular, or not diffuse (limited to 1 body part or region) were not considered an erythroderma for study purposes. Patients were further excluded if they were immunocompromised or were transferred to the ED from an inpatient unit at another facility. All of the cases were reviewed by both authors for entry into study.
Data Collection: ED Visit
Historical, clinical, and laboratory data were abstracted from the medical chart and entered into a database form with data validation rules. Historical data included the following: date of birth, date of visit, gender, 14 presenting symptoms (fever, measured temperature, headache, dizziness, altered mental status, sore throat, neck stiffness, myalgias, chest pain, abdominal pain, vomiting, diarrhea, gastrointestinal bleeding, and decreased urine output), duration of illness (based on fever), menses, tampon use, new medication in last 14 days (except antipyretics), and any regular medications. Physical examination signs at presentation included: triage vital signs, general appearance, mental status, mucosal hyperemia (conjunctival, pharyngeal, or vaginal), signs of focal infection, and presence of petechiae or peeling/desquamation.
For study purposes, specific a priori definitions were developed for examination elements: general appearance was coded as "well" (as defined by a record with terms such as: "well appearing," "alert," "happy," "nonill appearing," "playful," and "interactive"), "not well" (defined as "fussy," "lethargic," "sleepy but arousable," "irritable," "ill, but non toxic," "sick," and "uncomfortable"), and "toxic" (defined as "limp," "cyanotic," "mottled," "moribund," "toxic," "septic," "poorly perfused," "apneic," "gray," and "severe distress"); the most severe category was chosen if terms from multiple categories existed. Mental status was coded as "normal" (defined by terms such as "alert," "interactive," and "appropriate"), "lethargic" ("difficult to arouse," "poor tone," and "somnolent"), or obtunded ("decreased responsiveness," "unresponsive," "confused," and "inappropriate"). Signs of focal infections (abscess, adenitis, pharyngitis, pneumonia, purulent vaginal or cervical discharge, tracheitis, cellulitis, septic arthritis, osteomyelitis, and urinary tract infection/pyelonephritis) were also abstracted from the record.
Historical elements were coded as present, absent, or missing (if not recorded) on the basis of the medical chart. Physical examination findings were also coded as present or absent; when a sign was not mentioned, the finding was coded as absent.
Laboratory results from the ED evaluation were collected and studied if collected within 2 hours of the initial evaluation and before therapy with fluids. Chemistries, results of complete blood count, liver function tests, coagulation studies, inflammatory markers (erythrocyte sedimentation rate and C-reactive protein), urinalysis (collected within 8 hours from triage), and any body fluid analysis (cerebrospinal fluid, joint fluid, and pleural fluid) were recorded.
Data Collection: Clinical Management and Follow-up
The ED record and the inpatient medical chart (when admitted) were reviewed for therapeutic interventions and course. Clinical data included: length of stay, length of ICU stay, development of hypotension (5th percentile systolic arterial pressure for age),2 fluid resuscitation (requirement of 60 mL/kg of crystalloid within a 6-hour period), administration and duration of vasopressors, antibiotic use, and any laboratory data including results of bacterial cultures or serologies. Antibiotics that were considered to be "antistaphylococcal" included cephalosporins (first generation), clindamycin, dicloxacillin, nafcillin, oxacillin, and vancomycin. The medical chart was also reviewed for any outpatient follow-up visits at the hospital related to the index encounter, any morbidity as defined by decrease in neurologic function or other prolonged organ dysfunction, and survival.
Outcome Measures
Major outcomes included (1) the development of hypotension defined as either a systolic blood pressure 90 mmHg if 16 years of age or a systolic blood pressure 5th percentile if <16 years of age,2 (2) the administration of vasopressors, (3) the development of shock defined as either the administration of isotonic fluid boluses or colloid 60 mL/kg within a 6-hour period or hypotension (described as above), and (4) the development of CDC-defined staphylococcal or streptococcal TSS (confirmed or probable).1 The CDC definitions for TSS are listed in Tables 1 and 2.
Analysis
For patients who were not hypotensive on arrival to the ED, we analyzed historical examination, physical examination, and laboratory parameters as potential predictors for development of hypotension, shock, and shock requiring vasopressors. For all of the patients, including those hypotensive on arrival, we evaluated clinical data for predicting the development of staphylococcal or streptococcal TSS. Historical data elements with >20% missing data were not analyzed. A dichotomous variable for elevated creatinine was classified by age-based ranges.3
Statistical analysis was performed using SPSS 14.0 (SPSS Inc, Chicago, IL). 
2 analysis was used to compare categorical data, and independent samples t tests were used to compare mean values of continuous variables (normally distributed data). Mann Whitney U test was used to compare nonparametric data. A P value of .05 was considered statistically significant. Odds ratios with 95% confidence intervals (CIs) were calculated for categorical comparison, and mean differences with 95% CIs were calculated for continuous variable comparison. CIs for proportions and likelihood ratios were calculated with Stata 6 (Stata Corp, College Station, TX).
Recursive partitioning analysis (CART 5, Salford Systems, Palo Alto, CA) was used for determining the best cutpoint for continuous variables and to develop a multivariate prediction model for both the development of hypotension and the development of TSS. Recursive partitioning analysis develops a classification tree where a parent node is divided into child nodes using available candidate predictors. At each decision point, the variable that best divides the parent node toward homogenous nodes (on the basis of binary outcome variable) is used. At the end of the analysis, all of the patients are divided into "high-risk" and "low-risk" child nodes for the outcome variable. We entered our a priori predictors (gastrointestinal symptoms, ill appearance, mucosal hyperemia, acidosis, and thrombocytopenia), as well as any predictors with P = .2 on univariate analysis. A "penalty" was assigned to variables for missing data such that the program preferentially uses variables with complete data. In addition, we adjusted the "cost" of misclassification; a misclassification of a patient that eventually became hypotensive into a "low-risk" (for hypotension) group was considered 50 times worse that misclassifying a patient as "high-risk" for hypotension when in fact they never became hypotensive. Cost of misclassification was not adjusted for development of TSS. The study was approved by the institutional review board, and data collection was compliant with the Health Insurance Portability and Accountability Act of 1996.
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
2 search strategies. Four patients with fever and erythroderma were excluded: 2 patients were immunocompromised (both oncologic diagnoses), 1 patient had a prolonged inpatient admission before the ED visit, and 1 patient had incomplete documentation of the ED visit. The balance of 56 cases was studied. We estimated the frequency of fever and erythroderma to be 
0.024% per year (60 cases of 256606 visits in 5 years) or 1 patient per 4000 visits in our academic pediatric ED. The median age of the study patients was 10.7 years (interquartile range [IQR]: 2.8–14.8 years) with 50% of the patients being male. Forty-six patients presented normotensive, and 10 patients (18%) were hypotensive on arrival. Table 3 shows the presenting symptoms and signs of patients with fever and erythroderma. Forty-eight percent of patients presented with vomiting, and 42% of patients presented with duration of fever <24 hours. Forty-three percent of patients reported using a new medication (diphenhydramine, hydroxyzine, promethazine, cough suppressant, and oral antibiotics) during the 14 days before presentation, and 11% reported use of a medication chronically. Three patients presented after a surgical procedure (5 days status post [s/p] incision and drainage of pilonidal cyst, 8 days s/p herniorrhaphy and orchiopexy, and 5 days s/p circumcision). Thirty-two percent (9 of 28) of the female patients presented with active menstruation, and 8 of 9 were currently using tampons.
|
39°C. On physical examination, 45% of patients were described as ill appearing, and all of the patients had a normal mental status. Twenty-one percent of patients presented with conjunctivitis, and 29% had oropharyngeal hyperemia. Fourteen percent of patients had petechiae in addition to erythroderma. Fifty-two percent of patients had an identifiable focal source of infection including cellulitis without abscess (15), abscess (9), group A streptococcal pharyngitis (2), adenitis (1), osteomyelitis (1), and pneumonia (1). In comparing those who were not hypotensive on arrival with those hypotensive on arrival, mean age and ill appearance were less for the patients not hypotensive on arrival (8.4 vs 13.7 years [P = .008] and 35% vs 89% [P = .007], respectively). There was no significant difference in mean temperatures, heart rates, and respiratory rates. In addition, there was no significant difference in physical examination findings, such as mental status, mucosal hyperemia, petechiae, or focal bacterial infections. The initial laboratory findings of study patients are listed in Table 4. Overall there was not a significant difference in laboratory values for those with and without hypotension on arrival except for the mean number of polymorphonuclear neutrophils (P = .046) and proportion of patients with elevated creatinine (P = .024).
|
60 mL/kg over a 6-hour period) for a total of 15 (33%) of 46 who developed clinical shock. The median time between presentation and hypotension was 8 hours (IQR: 2.8–12.5 hours). Of the 23 patients who were hypotensive at any point during their hospitalization, 12 (52%) required vasopressor support. The median length of vasopressor infusion was 15 hours (IQR: 9–63 hours). As mentioned previously, 13 (28%) of 46 patients who were normotensive on arrival developed hypotension, and 7 of these patients required vasopressors.
|
The 5 patients who were discharged from the ED had discharge diagnoses of abscess, viral pharyngitis, viral syndrome, staphylococcal scalded skin syndrome, and serum sickness. The patient who was discharged with the diagnosis of serum sickness returned to the ED the next day with fluid-responsive shock; final diagnosis for this patient was serum sickness or viral sepsis. One patient was noted to have follow-up in the medical chart with no mention of subsequent illness, and 3 patients had no follow-up recorded in the medical chart. Only 1 of the 56 patients suffered significant morbidity (loss of tissue with grafting secondary to tissue necrosis), and there were no deaths reported.
Predictors
Development of Hypotension After Arrival
The clinical characteristics of patients who remained normotensive and those who developed hypotension after arrival to the ED are compared in Table 5. Patients were more likely to develop hypotension if they were older (P = .017), presented with vomiting (P = .015), or were ill appearing (P = .036). In addition, those patients who developed hypotension were more likely to have hematuria (red blood cells in the urine; P = .045), higher mean blood (serum) urea nitrogen (BUN; P = .01), higher mean glucose (P = .015), higher percentage of peripheral neutrophils (P = .003), elevated creatinine (P = .042), and a lower mean calcium (P = .034).
|
Development of TSS
Patients who were ultimately diagnosed with CDC-defined TSS were compared with those who never met TSS criteria. Patients who developed TSS had higher mean ages (years: 13.4 [3.3] vs 7.7[5.9]; P = .001), higher mean BUN (mg/dL: 20 [14] vs 13 [4]; P = .007), and higher mean alanine aminotransferase (ALT) (U/L: 67 [44] vs 36 [40]; P = .045); they also had lower mean calcium (mg/dL: 8.4 [0.7] vs 9.3 [0.7]; P = .004), elevated creatinine (7 of 16 [44%] vs 3 of 31 [10%]; P = .020), and were more likely to be ill appearing on arrival (11 of 15 [73%] vs 13 of 40 [33%]; P = .013). The following laboratory values were infrequently obtained at presentation and, therefore, not studied: prothrombin time, partial thromboplastin time, erythrocyte sedimentation rate, C-reactive protein, cell count, and culture from sterile fluid, including cerebrospinal fluid, joint fluid, and pleural fluid.
Recursive Partitioning Analysis
Thirteen variables were entered as potential predictors on the basis of a priori hypotheses and results of univariate analysis. Nine variables were determined to be the most important predictors of developing hypotension in patients initially normotensive: age 3 years, ill appearance, vomiting, glucose 110 mg/dL, calcium 8.6 mg/dL, platelets 300000/uL, elevated creatinine, polymorphonuclear leukocytes (PMNs) 80%, and presence of a focal infection. Table 6 shows these variables with associated positive and negative likelihood ratios and with cut points for continuous variables. In all of the patients studied, 4 variables were determined to be the most important predictors of developing TSS: age 3 years, ill appearance, elevated creatinine, and hypotension on arrival. Decision trees were created using recursive partitioning for the development of hypotension and development of TSS. The models and their prediction success are shown in Figs 2 and 3.
|
|
|
Sixteen patients (29%) met the CDC definition for either staphylococcal or streptococcal TSS (confirmed or probable); the frequency of CDC-defined TSS was 0.0064% (16 cases of 256606 visits over 5 years). However, many patients were considered to have possible toxin-mediated disease or possible TSS as their final diagnosis. One patients did not meet CDC-defined TSS because of a maximum temperature <38.9°C; 2 patients did not meet CDC-defined TSS because they had only 2 of 3 organ systems involved; and 2 patients never had a recorded systolic blood pressure below the 5th percentile despite having clinical shock. Only 2 bacterial isolates were sent for detection of toxin production; the only positive test for TSS toxin occurred in a patient with fever, erythroderma, and hypotension who never met CDC definition of TSS secondary to lack of multisystem organ involvement. Serum samples from 5 patients were sent for toxin antibody, and all of the samples were reported as negative. Among patients with fever and erythroderma who did not have a focal bacterial infection or progression to TSS, other final diagnoses included: drug hypersensitivity (2 patients), staphylococcal scalded skin syndrome, serum sickness, viral syndrome, superinfected eczema, and superinfected varicella.
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
TSS was first named by Todd et al5 in 1978. They described a series of pediatric cases of high fever, ill appearance, headache, conjunctival hyperemia, confusion, poor perfusion, and rash that were associated with vomiting, diarrhea, decreased urine output, hepatic abnormalities, renal abnormalities, shock, and a staphylococcal toxin.
The incidence of TSS greatly increased in 1979 and 1980 in young healthy women.6 This increase led to many retrospective case-control studies that described an association between TSS and prolonged tampon use during menstruation.6–12 Many recommendations were made on the basis of the literature, and an aggressive media campaign ensued leading to a decrease in menstrual TSS cases.13
Nonmenstrual TSS can be seen in a wide variety of clinical scenarios, including focal lesions of soft tissue, bone infections, and lung infections. Cases have been reported in burn victims and after specific infections, such as influenza and varicella.14–16 Other cases have been reported to be postsurgical, postpartum, and postabortion.13
In 1980, the first CDC case definition for TSS was published.9 The case definition has since been revised, and a separate case definition of streptococcal TSS exists.1,16,17 Although there are strict CDC definitions for TSS, the clinical presentation, laboratory abnormalities, and course of illness have been noted to be highly variable.18
Medical literature on TSS in children is limited. According to the 17-year (1979–1996) CDC surveillance update, 50 cases of TSS were reported in children <5 years of age with more than half of the cases occurring in children <2 years of age. Of the 50 cases, 61.7% were associated with a nonsurgical skin lesion, and the overall case/fatality ratio was 4%.13
Wiesenthal and Todd19 described the clinical and laboratory findings of 7 children (4 boys and 3 girls) <10 years of age with CDC-defined TSS during the period of March 1977 to August 1981. All of the patients had fever, erythroderma, and mucous membrane hyperemia on admission. Four patients were hypotensive on admission, and 3 developed hypotension during the hospital course. One patient died from shock, another patient developed extensive tissue necrosis requiring grafting, and finally 1 patient met the definition for Kawasaki disease and developed left-sided heart failure, coronary aneurysms, and pulmonary edema. The 3 girls appeared to have a milder illness than the boys, and S aureus grew from cultures of 4 patients. They concluded that the clinical and laboratory findings (electrolyte, renal, and hepatocellular) were similar to adult patients; however, the pediatric cases had more pulmonary dysfunction.19
Other descriptive case reports of children with TSS have included small numbers of patients.20–22 In 2000, Laupland et al14 conducted a population-based prospective surveillance study describing children with invasive group A streptococcus disease. They reported that 7% of children with invasive group A streptococcal disease developed streptococcal TSS. The case fatality rate for streptococcal TSS was 56%. In 2005, Chiang et al21 reported 4 cases of streptococcal TSS in children without skin and soft tissue infections; pharyngitis was identified in 3 of their 4 cases.
Summary of Findings
In our population, we found that 28% of patients who presented with fever and erythroderma developed hypotension after arrival to the ED. Overall, 45% of patients with fever and erythroderma developed shock, including 33% of those who were normotensive on presentation. Of those who developed hypotension, >50% of patients required vasopressors. Despite the development of hypotension, use of vasopressors, and requirement of critical care, long-term morbidity was uncommon. Fifty-two percent of patients with fever and erythroderma had a focal source of infection at presentation and, of those, half of the patients had a positive bacterial culture. Only 5% of the patients had bacteremia despite the large number of patients with sepsis physiology consistent with the role of toxin and inflammatory mediators rather than overwhelming bacterial infection.
Patients who were older and who presented with vomiting, ill appearance, elevated BUN, elevated creatinine, elevated glucose, decreased calcium, elevated percentage of polymorphonuclear leukocytes, and hematuria were more likely to develop hypotension after arrival. Exploratory analysis with recursive partitioning identified 9 variables as the most important predictors of developing hypotension: age 3 years, ill appearance, vomiting, glucose 110 mg/dL, calcium 8.6 mg/dL, platelets 300000/µL, elevated creatinine, PMNs 80%, and presence of a focal infection.
Only 29% of patients met the CDC definition for either staphylococcal or streptococcal TSS. Patients who were older and who presented with ill appearance, elevated BUN, elevated creatinine, elevated ALT, and decreased calcium were more likely to develop TSS. Recursive partitioning identified 4 variables as the most important predictors of developing TSS: age 3 years, ill appearance, elevated creatinine, and hypotension on arrival. The relatively low percentage of patients who met the CDC definition likely reflects a wider spectrum of disease not captured by the restrictive nature of the CDC definition. It is possible that the CDC definition for TSS may be conservative for clinical purposes.
Limitations
The most significant limitation is the retrospective design that limits the ability to abstract accurate, complete clinical information from the medical chart. In addition, we may have incomplete case identification through our text search tool; case identification was redundant for the various search strategies and, therefore, should have minimized the number of missing patients. Clinical findings, such as general appearance and mental status, are often subjective and difficult to determine retrospectively even in well-documented charts. We used a priori definitions to assist in abstracting data of this type; however, limitations in accuracy exist. We attempted to capture those patients with true erythroderma by using strict criteria; however, we relied on appropriate documentation to make the determination. Historical elements not documented in the medical chart were considered to be missing, whereas physical examination findings not documented were presumed to be negative; the decision was felt to be best for a retrospective review but carries the biases of inaccurate and incomplete documentation. We also studied management of these patients that is not based on any specific guideline or standard of care; for instance, the use of vasopressors was studied and felt to be objectively recorded, but the threshold to use or discontinue vasopressors is not universal among physicians. Our outcomes are also not entirely complete with regard to long-term follow-up on patients. Three of the 5 patients discharged from the ED initially did not have long-term follow-up.
Although performing repeated univariate analyses, our results are subject to a type I statistical error; an adjusted level of significance could have been used, but with such a small study population, we retained a significance for P .05 to broadly identify potential predictors for hypotension and TSS. Although this is the first study of children with fever and erythroderma that attempts to predict outcome, the sample size limits the conclusions about predictive strength of any of the clinical variables. Using a relatively small number of patients to build a multivariate model, recursive partitioning likely "overfit" the data, especially with adjusted cost of misclassification. Thus, our models should be viewed as exploratory analysis with the need for larger derivation sets and, ideally, prospective study.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
FOOTNOTES |
|---|
Address correspondence to Robyn L. Byer, MD, Children's Hospital, 300 Longwood Ave, Boston, MA 02115. E-mail: robyn.byer{at}childrens.harvard.edu
This work was presented in part at the annual meeting of the Pediatric Academic Societies; May 2, 2006; San Francisco, CA.
The authors have indicated they have no financial relationships relevant to this article to disclose.
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
||
|
||
Home | My Pediatrics | Journal Information | Current Issue | Past Issues | Subscriptions & Services | Contact Us Click here for faster international access © 2006 American Academy of Pediatrics. All rights reserved. |
||
|
||
