 |
|
 |
|
 |
 |
|
 |
 |
 |
 |
 |
 |
 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ARTICLE |
a Primary Children's Medical Center, Salt Lake City, Utah
b New England Research Institutes, Watertown, Massachusetts
c Medical University of South Carolina, Charleston, South Carolina
d University of Toronto, Hospital for Sick Children, Toronto, Ontario, Canada
e Columbia University Medical Center, New York, New York
f Children's Hospital Boston and Harvard Medical School, Boston, Massachusetts
g Childrens Hospital Los Angeles and University of Southern California, California
h Duke University Medical Center, Durham, North Carolina
i Children's Hospital of Philadelphia, Philadelphia, Pennsylvania
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
METHODS. We collected sociodemographic and clinical data at presentation for all patients who were treated for presumed Kawasaki disease at 8 centers (7 in the United States, 1 in Canada). Delayed diagnosis was evaluated by total number of illness days to diagnosis and by the percentage of patients who were treated after day 10 of illness. Independent predictors of delayed diagnosis were identified by using multivariate linear and logistic regression.
RESULTS. Of the 589 patients who received intravenous immunoglobulin, 27 were treated before screening for the trial and excluded; 562 patients formed the cohort for analysis. Kawasaki disease was diagnosed at 7.9 ± 3.9 days, 92 (16%) cases after day 10. Centers were similar with respect to patient age and gender. Centers differed in the patient percentage with incomplete Kawasaki disease; clinical criteria of cervical adenopathy, oral changes, and conjunctivitis; and distance of residence from the center. Independent predictors of greater number of illness days at diagnosis included center, age of <6 months, incomplete Kawasaki disease, and greater distance from the center. Independent predictors of diagnosis after day 10 were age of <6 months, incomplete Kawasaki disease, and greater distance). Socioeconomic variables had no association with delayed diagnosis.
CONCLUSIONS. Even after adjustment for patient factors, illness duration at diagnosis varies by center. These findings underscore the need to maintain a high index of suspicion of Kawasaki disease in the infant who is younger than 6 months and has prolonged fever even with incomplete criteria. Outreach educational programs may be useful in promoting earlier recognition and treatment of Kawasaki disease.
Key Words: Kawasaki disease • delayed diagnosis • Pediatric Heart Network
Abbreviations: KD—Kawasaki disease • IVIg—intravenous immunoglobin • PHN—Pediatric Heart Network • OR—odds ratio
Kawasaki disease (KD) is an acute febrile vasculitis that is associated with coronary artery aneurysms in 
20% of untreated children.1,2 Treatment with intravenous immunoglobulin (IVIg), if administered during the first 10 days of illness, reduces the prevalence of coronary aneurysms approximately fivefold.3,4 When treatment is delayed beyond day 10 of illness, the incidence of coronary artery aneurysms increases nearly threefold over earlier treatment.5 Indeed, treatment on or before day 7 of illness is ideal.6,7
Timely diagnosis of KD is often a challenge. Despite almost 4 decades of research, the cause of KD remains unknown, so the diagnosis depends on nonspecific clinical signs rather than a definitive laboratory test. The epidemiologic case definition includes fever and at least 4 of 5 principal clinical criteria: changes in the extremities, polymorphous exanthema, bilateral conjunctival injection, changes in the lips and oral cavity, and cervical adenopathy.1,7 These findings may be transient, and the constellation of principal clinical criteria that are present on physical examination may vary over time. Furthermore, coronary artery aneurysms have been documented in incomplete cases with <4 principal clinical criteria.8
The purposes of this study were to describe the prevalence of and risk factors for delayed diagnosis of KD. Our study population comprised all children who were treated with IVIg during a 2-year period at clinical centers that participate in the National Heart, Lung, and Blood Institute Pediatric Heart Network (PHN) corticosteroid trial for KD treatment.9
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Patient data were deidentified by using study identification numbers, and the study protocol and screening forms for the trial were reviewed and approved by the institutional review board at each of the 8 clinical centers and the data-coordinating center. An independent data and safety monitoring board monitored the conduct of the trial.
Data Collection and Definitions
Demographic data that were collected on the screening form included age at presentation, gender, race and ethnicity, and zip or postal code. Clinical data included number of illness days from onset of fever to diagnosis of KD, previous treatment with IVIg, and the principal clinical criteria that had been or were present by the time of KD diagnosis.
We assessed delay in diagnosis of KD in 2 ways: (1) total number of illness days from onset of fever to diagnosis of KD; and (2) the proportion of patients whose illness was diagnosed after day 10. By convention, the first day of fever was considered day 1 of illness.
The distance from the patient's residence to the center was derived from zip or postal code. Socioeconomic status was estimated by using 2000 census data (www.census.gov/Press-Release/www/2002/sumfile3.html) and was available by US zip code (thus not calculated for Canadian patients). Three measures of socioeconomic status were analyzed: the percentage of the population with a high school degree or higher, median family income, and the percentage of households under the poverty level. Each US patient was assigned the socioeconomic status that corresponded to his or her zip code.
Statistical Analyses
The primary outcome measures for this study were the number of days from the onset of fever to diagnosis of KD as a continuous variable and whether diagnosis occurred after day 10 of illness. We compared center differences in outcomes and proportion with incomplete KD using analysis of variance and Fisher's exact test, respectively. Univariate associations between number of fever days and potential predictors were identified using linear regression. Univariate associations between the proportions of patients whose illness was diagnosed after day 10 were identified using logistic regression. The Tukey multiple-comparisons method was used to compare mean number of illness days between pairs of racial subgroups. Generalized additive modeling was used to examine nonlinearities in age and distance with respect to outcome and to suggest appropriate transformations. Univariate predictors significant at the P < .20 level were then evaluated in multivariate modeling. Multivariate linear and logistic regressions were used to identify independent predictors of the number of illness days until diagnosis and the proportion diagnosed after day 10 of illness, respectively. Least-squares means ± SEs are used to report covariate-adjusted number of illness days. Analyses were conducted by using SAS 9.1 (SAS Institute, Inc, Cary, NC).
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
4 criteria (9.5 ± 4.7 vs 7.3 ± 3.3 days, respectively; P < .001). Similarly, a higher percentage of patients with incomplete KD received the diagnosis after day 10 of illness (29% vs 12%; P < .001). The total number of illness days to diagnosis was associated with the presence versus absence of individual clinical criteria: extremity change (7.7 ± 3.8 vs 8.6 ± 4.0 days; P = .02), rash (7.8 ± 3.8 vs 9.0 ± 4.4 days; P = .01), bilateral conjunctival injection (7.6 ± 3.4 vs 10.7 ± 5.8 days; P < .001), oral change (7.6 ± 3.4 vs 10.2 ± 5.7 days; P < .001), and cervical lymphadenopathy (7.5 ± 3.4 vs 8.3 ± 4.2 days; P = .02). Similarly, diagnosis after day 10 of illness was more common when specific clinical criteria were absent (Table 1).
Younger age was also significantly associated with delayed diagnosis; nonparametric modeling demonstrated that diagnosis was most likely to be delayed in infants who were younger than 6 months. Infants who were younger than 6 months (n = 37), compared with older children (n = 525), received the diagnosis after a significantly greater number of illness days (10.0 ± 5.4 vs 7.8 ± 3.7 days; P < .001). Similarly, a significantly greater proportion of infants who were younger than 6 months received the diagnosis after day 10 of illness (41% vs 15%; P < .001). Age had a curvilinear relationship with diagnosis after day 10 (Fig 1), with adolescents also at risk for delayed diagnosis; however, statistical power was limited in this age range, and a term for additional risk in this age group was not retained in modeling.
|
In univariate analysis, race was significantly associated with the total number of illness days to diagnosis (P = .04) because white patients had a longer time to diagnosis compared with Asian patients (mean: 8.3 vs 7.1 day, pairwise adjusted P = .01). The trend was similar for the proportion of patients who received the diagnosis after day 10 of illness, with 20% of white patients having late diagnosis and 10%, 12%, and 11% of Asian, black/African American, and "other" patients, respectively, having a late diagnosis (P = .04). Gender had no association with either the total number of illness days to diagnosis (P = .58) or the percentage of patients who received the diagnosis after day 10 of illness (P = .42).
Among US cases, socioeconomic status was not related to either of the outcome measures used to define delayed diagnosis. The mean for median family income of the patients who received a diagnosis early (on or before day 10) versus late (after day 10) of illness was $59700 ± $27200 vs $58200 ± $21500, respectively (P = .66). The mean percentage of families who were estimated to be below poverty level was 10 ± 10% vs 9 ± 10% for the patients who received the diagnosis early versus late (P = .41). The mean percentage of families who were estimated to have a high school degree or higher was 80 ± 15% vs 82 ± 14% for the patients who received the diagnosis early versus late (P = .50).
We considered whether center differences were related to the likelihood of delayed diagnosis. Both the total number of days from onset of fever to diagnosis (center mean number of illness days range: 6.4–9.7 days; P < .01) and the proportion of total patients who received the diagnosis after day 10 of illness (range: 4% to 30%; P = .02) varied significantly by center. Centers also varied significantly in the proportion of treated patients who had <4 principal clinical criteria (range: 13% to 47%; P = .004).
Multivariate Analyses
The results of multivariate regression analyses are summarized in Tables 2 and 3. Centers differed in their proportion of patients with incomplete KD and with specific principal clinical criteria; however, clinical center remained an independent risk factor for an increased number of illness days from onset of fever to diagnosis of KD even when adjusting for these factors (Fig 2). Specifically, in models that adjusted for the presence versus absence of incomplete KD (ie, <4 vs 4 principal criteria), independent risk factors for greater number of illness days until diagnosis included age < 6 months (P = .01), distance from center (P < .03), < 4 principal criteria (P < .001), and clinical center (P = .03). When the multivariable model included individual principal criteria rather than an indicator variable for incomplete KD, independent predictors included age < 6 months (P < .01), clinical center (P < .05), absence of bilateral conjunctival injection (P < .001), and absence of oral changes (P < .001). Greater distance from center was marginally associated with delays in diagnosis in this model (P = .06). We assessed whether the impact of distance and incomplete KD on number of illness days depended on age (<6 vs 6 months) and found no interaction.
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
We used 2 outcomes to assess delays in the diagnosis of KD: the number of days from onset of illness to diagnosis (illness days) and diagnosis on or before day 10 versus after day 10 of illness. We found that predictors of delayed diagnosis included age < 6 months, fulfilling <4 principal clinical criteria (ie, incomplete KD), greater distance of residence from the clinical center, and specific clinical center (illness days only). The presence of conjunctival injection or of oral changes was protective against a late diagnosis of KD.
Previous studies have explored the consequences and predictors of delayed diagnosis of KD. A retrospective chart review of patients who had KD and were treated during a 6-year period in Denver revealed that patients who were treated after day 10 of illness were 2.8 times more likely to have coronary artery aneurysms than those who were treated earlier.5 Indeed, because vascular injury is evident as early as 1 week after the onset of fever,6 the most recent American Heart Association/American Academy of Pediatrics guidelines for KD recommend treatment by day 7 if possible.7 Treatment within this period depends on both timely medical consultation by the parents and early recognition of KD by the clinician who is evaluating the child.
A variety of factors may have a negative impact on the timely diagnosis of KD. Anderson et al5 retrospectively explored the reasons for diagnosis after day 10 of illness during an outbreak in Colorado. In contrast to our prospective study in which 16% of children received the diagnosis after day 10 of illness, they found that 24% of their KD patients received the diagnosis after day 10. In their study, late diagnosis was not related to age, gender, time to the first medical visit, number of medical visits, physician specialty, number of antibiotics received, white blood count, or sedimentation rate but was related to dispersion of symptoms over a longer period. In fact, patients in their study eventually exhibited the typical features of KD. Delays in diagnosis were also associated with significantly more days of fever, rash, conjunctival injection, or oral changes in their study. Because data were obtained only at 1 point in time (screening), we were unable to assess whether the incomplete cases in our study eventually fulfilled the recommended American Heart Association/American Academy of Pediatrics criteria for KD. Their latter finding was not supported by our study, however, in which the presence of conjunctival injection and oral changes actually lessened the likelihood of delayed diagnosis.
Incomplete cases have also been implicated as a major factor in late diagnosis. A review11–13 reported that children with fewer than the recommended 4 criteria account for 15% to 20% of patients who are treated for KD. The percentage of incomplete cases was higher in this study (28%) and may represent the referral pattern for KD at the participating centers.
Younger age was previously reported5,14–16 as a risk factor for late diagnosis and coronary abnormalities. A recent study from Taiwan15 reported that infants who were < 6 months received the diagnosis an average of 2 days later than older children, and 50% received the diagnosis after day 10 compared with 22% of children who were 6 months. Despite these reports, many physicians continue to have a low index of suspicion for KD in the febrile infant. In a 2004 survey, 57% of 132 pediatricians and 26% of 345 pediatric infectious disease physicians reported that they did not consider the diagnosis of KD for a child who was younger than 6 months or older than 8 years.17 The failure even to consider KD may account for our finding that infants in this age group received the diagnosis after significantly more illness days and had 3.5 times the odds for receiving the diagnosis after day 10 of illness. Our data suggest a U-shaped relationship with age and support previous reports of delayed diagnosis in older children as well,18,19 but because of the small number of patients in the older age group, there was insufficient power to show significance.
Practice variation among centers may also account for delays in diagnosis. Clinical centers differed significantly in the proportion of incomplete cases that were treated with IVIg as well as in the number of illness days to diagnosis and the proportion of patients who were treated after day 10. We found that the specific clinical center at which a child was treated was an independent predictor of greater illness days at diagnosis even after adjusting for age < 6 months, clinical presentation, and distance. In contrast, center differences for diagnosis after day 10 of illness were explained at least in part by differences in the specific principal criteria (eg, absence of conjunctival injection or extremity changes) met at the centers.
This study should be viewed in light of its limitations. First, because there is no diagnostic laboratory test for KD, the child may be treated even without diagnostic certainty, especially for incomplete cases. The design of this study did not permit us to determine the criteria used for diagnosis or to explore the extent to which time to diagnosis of KD was affected by differing thresholds among physicians and centers for treatment of incomplete KD. Second, the patient's zip code was used as a proxy for socioeconomic status. Although it is a common practice for health researchers to use this approach, the use of such aggregate proxy can introduce statistical bias.20 Third, the data set did not allow us to determine the particular characteristics of each center that were associated with a higher percentage of patients with delayed diagnosis. Specific factors that were evaluated in previous studies, such as the medical specialty of the provider, number of antibiotics received, number of follow-up visits, and access-to-care issues, were not collected from our screened cohort. Finally, in patients who were screened but not enrolled in the randomized trial, data were not collected on coronary artery abnormalities or laboratory test results, further limiting our ability to explore risk factors for and consequences of delayed diagnosis.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
ACKNOWLEDGMENTS |
|---|
We thank the following individuals for contributions to enrollment of patients, data collection, or study coordination: Children's Hospital Boston: Annette L. Baker, MSN, PNP, David R. Fulton, MD, Amy L. Woodward, MD, Fatma Dedeoglu, MD, Bryce Beinstadt, MD, PhD, Susan Kim, MD, and Ellen McGrath, RN; Children's Hospital of Los Angeles: Wilbert Mason, MD, and Vanessa Guerrero, RN; Children's Hospital of Philadelphia: Alexa Hogarty, MD, Steven Paridon, MD, Jack Rychik, MD, and Meg Harkins, RN; Columbia University Medical Center: Lisa Imundo MD, Deborah Levy MD, Daphne Hsu, MD, and Seema Mital, MD; Duke University Medical Center: Page Anderson, MD, Ann Marie Nawrocki, RN, and Daphne Harrington; Hospital for Sick Children: Timothy Bradley, MD, Rae Yeung, MD, and Elizabeth Radojewski, RN; Medical University of South Carolina: Girish Shirali, MD, Mark Scheurer, MD, and Amy Jones, RN; Primary Children's Medical Center: Richard V. Williams, MD, Deborah U. Frank, MD, PhD, Lloyd Y. Tani, MD, Linda M. Lambert, MSN, FNP, and Abigail L. Smart, RN; data-coordinating center (New England Research Institutes): Patti Nash, Dianne Gallagher, and Paul Mitchell; C-reactive protein core laboratory: Steven D. Douglas, MD, Donald E. Campbell, PhD, Joseph M. McMann, MS, MT (ASCP)SI, Patricia Wuthnow, MT(ASCP), CLS(NCA), Nichole Gilper, MT (ASCP), CLS(NCA), and Marie Wilson, MT(ASCP)SI; echocardiographic core laboratory: David Cabral and Edward Marcus; National Heart, Lung, and Blood Institute: Gail D. Pearson, MD, ScD, Victoria Pemberton, RNC, BS, CCRC, Mario Stylianou, PhD, and Judith Massicot-Fisher, PhD; and the PHN Chair, Lynn Mahony, MD.
We also thank members of the Data and Safety Monitoring Board: John D. Kugler, MD (Chair), David J. Driscoll, MD, Sally A. Hunsberger, PhD, Catherine L. Webb, MD, Lawrence S. Wissow, MD, MPH, Kathryn D. Davis, PhD, and James S. Tweddell, MD.
|
FOOTNOTES |
|---|
Address correspondence to L. LuAnn Minich, MD, 100 N Medical Dr, Salt Lake City, UT 84113. E-mail: l.luann.minich{at}ihc.com
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 © 2007 American Academy of Pediatrics. All rights reserved. |
||
|
||
