Objective: To describe and quantify the presentations of Kawasaki disease (KD) in a children’s hospital over 10 years to assess the Harada score in a US population.
Methods: A retrospective chart review from 2001 to 2011 of children discharged from Cleveland Clinic with the diagnosis of KD. Demographic and clinical data were collected and Harada scores were derived to evaluate efficacy in predicting risk for coronary artery aneurysms (CAAs).
Results: A total of 105 children met diagnostic criteria for KD, and 97 of 105 had long-term follow-up. Full criteria for KD were found in 67 of 105 (64%); 38 had incomplete presentations. CAA developed in 10 children, 5 during follow-up despite treatment with intravenous immunoglobulin (IVIG.) Children with incomplete presentations had a higher risk of developing CAA (20% vs 5%, P = .03) and a delayed diagnosis (median days from fever to diagnosis 8.0 vs 5.0 days, P < .001). Of children who developed CAA, 9 of 10 had a positive Harada score (sensitivity of 90%). All children who developed CAA after IVIG were in the high-risk group, but 1 child with an incomplete presentation who had a CAA at presentation was missed by the score. Overall, the negative predictive value was 98%.
Conclusions: As in Japanese studies, a positive Harada score in a US population could be used to identify a high-risk population for CAA development. All children who developed CAA after treatment with IVIG would have been assigned to a high-risk category. Though not specific enough to select initial therapy, the score might be useful in identifying high-risk children for evaluation of new therapies and more frequent follow-up.
Kawasaki disease (KD), also known as mucocutaneous lymph node syndrome, is an acute systemic inflammatory illness, a multisystem vasculitis of young children with variable presentations.1–4 The most dangerous manifestations are not part of the diagnostic criteria and include myocarditis, congestive heart failure (CHF), and coronary artery aneurysms (CAAs). Prompt treatment with IVIG has been shown to treat all manifestations of KD and to significantly decrease the risk for development of coronary artery aneurysms. Treatment within 10 days of onset of fever decreases the risk for development of CAA from 20% to as low as 5%.5
Methods to predict which children are at greatest risk for CAA have been sought to determine prognosis and select children for more rigorous follow-up. No definitive method has been found. The Harada score, developed to identify children at high risk for development of CAA at a time when not all children with KD were treated with IVIG in Japan, was an effective screening tool.6 It was developed by retrospectively correlating routine laboratory and clinical findings during the early stage of illness with development of CAA. The score includes 7 items: white blood cell count (WBC), platelet count, C-reactive protein (CRP), hematocrit, albumin, age, and gender (Table 1). Children with 4 of the 7 criteria are considered at high risk for CAA and would be treated with IVIG.
All children in the United States with KD are given IVIG, so an effective score would not be useful for selecting children to treat with IVIG. However, an effective score could be used to select children for evaluation of additional therapies (eg, steroids) to prevent CAA that occur despite treatment with IVIG. Because the presentations and outcomes of KD differ in various genetic populations, this score must be evaluated in a US population before it could be used clinically. We therefore sought to evaluate the Harada score in a well-characterized group of children with KD treated at Cleveland Clinic Children’s.
This study was a retrospective chart review of patients admitted to Cleveland Clinic Children’s from 2001 to 2011 whose final diagnosis was KD. Because it was a retrospective chart review of our patients, the informed consent requirement was waived by our institutional review board. Patients were identified by using International Classification of Diseases, Ninth Revision codes for “Kawasaki Disease,” “Kawasaki’s Disease,” and “Mucocutaneous Lymph Node Syndrome.” Charts were reviewed to assess whether each patient met the complete diagnostic criteria for KD and whether uncommon clinical findings were present.1 Variables collected for each patient included age; gender; signs and symptoms; diagnostic tests at the time of presentation including WBC, hemoglobin, platelet count, serum albumin, alanine aminotransferase, aspartate aminotransferase, CRP, erythrocyte sedimentation rate, and urinalysis; echocardiogram results; time to diagnosis from onset of fever; time to treatment from onset of fever; need for retreatment; and presence of CAA on presentation or at follow-up. Clinical manifestations were also collected, including uncommon manifestations of KD. A Harada score was calculated for each patient at the time of diagnosis to evaluate predictive value for CAA development.
Each echocardiogram read as abnormal at the time of diagnosis was reviewed by a cardiologist, and normal variants were removed. Abnormal echocardiogram results were defined according to the American Heart Association.1 Findings of enhanced echobrightness of the soft tissue surrounding the coronary artery lumen and diffuse dilatation were verified independently by 2 pediatric cardiologists.
The Harada score was considered positive if it was ≥4. In cases where a subject’s missing Harada score data would not change whether the score was positive, the subject was included for analysis. Study data were collected and managed using Research Electronic Data Capture electronic data capture tools hosted at Cleveland Clinic.7 The subject cohort was described using medians with interquartile range, means with standard deviations, and counts and percentages with 95% confidence intervals (CIs). Patients with complete KD were compared with those with incomplete KD, and patients with typical presentations of KD were compared with patients with uncommon clinical features, using χ 2 or Fisher’s exact tests for categorical variables and t tests or Wilcoxon rank sum test for continuous variables. Measures of diagnostic accuracy for the Harada score in the prediction of CAA were calculated, including sensitivity, specificity, and positive and negative predictive value, and their exact 95% CIs were estimated. All analyses were performed on a complete case basis, and all tests were 2-tailed and performed at a significance level of .05.
A total of 106 patients were identified with a diagnosis of KD over the 10-year period. All had admission data and an initial echocardiogram. One patient was coded as having KD but did not meet criteria and on follow-up was thought to not have had KD and was removed from the study. Median age at time of diagnosis for the 105 children was 2.8 years (range 2 months–14 years). Seventy-one patients (68%) were male. Of the 92 patients with available ethnicity information, 53 were white (58%), 26 were African American (28%), 9 were Asian (10%), and 4 Hispanic (4%). There were 97 cases with follow-up echocardiograms at 4 weeks after presentation that could be used for evaluation of the Harada score (Table 2).
Signs and Symptoms
Most children (67 of 105) presented with complete criteria for KD. Children with incomplete KD (38 of 105) presented somewhat later, median 8 days compared with 5 days of fever (P < .001), and were more likely to develop CAA (P = .033). There were 10 children who developed CAA; 5 had CAA at presentation, and 5 developed CAA at follow-up within 4 weeks. Of these 10 patients with CAA, 70% had incomplete presentations. Patients with incomplete KD had diagnosis delayed past 10 days after the onset of fever 24% of the time (9 of 38), compared with <5% of the time in those with complete KD (3 of 64), an odds ratio of 6.3 (95% CI, 1.6–25.1, P = .008) (Table 2).
Uncommon clinical findings were present in 14 of 105 (13%) patients, including CHF (3), myocarditis (2), shock (1), hydrops of the gall bladder (4), jaundice (1), testicular swelling (2), and pleural effusion (5). The presence of these uncommon clinical findings did not delay time to diagnosis or development of CAA. The children with CHF, myocarditis, and shock each responded promptly to treatment with IVIG. None of our children with uncommon cardiac manifestations had CAA on presentation, and 5 of 6 responded quickly to IVIG with rapid reversal of signs and symptoms. One developed a CAA on follow-up.
A multivariate analysis of risk factors for development of CAA despite treatment was attempted, but there were too few children to evaluate important factors such as race, so this was not possible.
Harada Score and CAA
In our cohort, 93 of the 105 children had adequate data to calculate a Harada score at presentation. The Harada score identified 48 of 93 to be at high risk for CAA (score of ≥4). One child did not have adequate follow-up, leaving 47 of 48 for long-term follow-up for development of CAA. Of the high-risk children with adequate follow-up, 19% (9 of 47) had or developed a CAA, giving a positive predictive value of 19% (exact 95% CI, 9%–33%). Of the children who developed CAA, 9 of 10 patients had a positive Harada score, a sensitivity of 90% (exact 95% CI, 56%–100%), and a specificity of 51% (exact 95% CI, 39%–62%). The negative predictive value of a low Harada score was 98% (exact 95% CI, 89%–100%). The score was most accurate in children with complete KD (Table 3). The 1 child missed by the score was a 10-year-old girl who presented with a CAA and 10 days of fever.
The etiology of KD remains unknown, complicating diagnosis and treatment of the disease. Although IVIG has changed prognosis significantly, KD continues to be associated with the development of CAA in a small percentage of children, 10% in our series, with CAA developing despite treatment in half of the children in whom CAA developed. The Harada score was developed at a time when cost and availability made it impossible to treat all children with KD in Japan with IVIG. The score was used to select high-risk children diagnosed with KD who would receive IVIG. In the United States, all children diagnosed with KD are treated, and this score was never used. Evaluating children for risk of CCA would be useful, however, to select a high-risk population to evaluate new therapies efficiently. Low-risk children are adequately treated just with IVIG, but high-risk children might benefit from additional therapies. Genetic variations appear to explain the variations in disease incidence, severity, and presentations described in different populations,8 making it necessary to evaluate the usefulness of any clinical scoring system in a particular population before it can be used.
In addition to the Harada score, a variety of markers have been evaluated to try to predict children at highest risk for development of CAA. Some scores have tried to predict which children will not respond to treatment with IVIG (a risk factor for the development of CAA). These models have included baseline neutrophil and band counts, hemoglobin concentration, platelet count, and temperature on the day after infusion of IVIG9; hemoglobin, WBC, and serum albumin,10,11 as well as CRP, lactate dehydrogenase, and total bilirubin at presentation12; age, WBC, percentage neutrophils, platelet count, aspartate aminotransferase, sodium, and CRP11; albumin alone13; and recently brain natriuretic peptide.10,14 These models have each been described in small groups of genetically different populations. None are completely accurate in predicting development of CAA. The common pattern appears to be that children with evidence for the most severe inflammation in any particular population are at highest risk for CAA and for lack of response to IVIG. A small but significant percentage of children triaged to high risk actually go on to develop CAA. All these scores have low specificity.
The Harada score has been used extensively in Japan with sensitivities adequate to select a group of high-risk children to receive therapy to prevent CAA but with low specificity. When applied to our patients at the time of diagnosis of KD, the score separated our population into high- and low-risk groups for development of CCA. The high-risk group included 9 of 10 children who developed CAA; 19% of children in the high-risk group went on to develop CAA, similar to results in the early Japanese literature. One child in the low-risk group had a CAA on presentation. The score had a negative predictive value of 98%. Although this score would not be used to triage children who do not need IVIG, its original indication, it might be useful in selecting patients at highest risk to evaluate new therapies to try to decrease the incidence of CAA beyond treatment with IVIG.
Although our study was retrospective, the database was robust and follow-up sufficient to enable verification of the Harada scale in a US population. The 10% rate for CAA in our series is on the high side, as is our 20% retreatment rate. We also had a significant number of children with cardiac manifestations at presentation, a risk factor in some series for CAA. Our population may be skewed toward children with cardiac presentations of KD because Cleveland Clinic Children’s Hospital’s cardiac service is a tertiary referral center for children with heart disease. Our population of children diagnosed with KD may also be more uniform than that seen at some centers because we have a multidisciplinary approach to this disease. The children are on our hospitalist service, but every child with possible KD is evaluated by the infectious diseases and cardiology departments. This approach may help prevent children with other diagnoses from being included as having KD.
This study suggests that the Harada score was effective in a US population in selecting a high-risk group of children with KD who might benefit from evaluation for additional therapy beyond the standard IVIG. The Harada score is worthy of additional evaluation by other US groups to confirm our findings.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: No external funding.
POTENTIAL CONFLICT OF INTEREST: The authors have no potential conflicts of interest to disclose.
- coronary artery aneurysm
- congestive heart failure
- 95% confidence interval
- intravenous immunoglobulin
- Kawasaki disease
- white blood cell count
- Newburger JW,
- Takahashi M,
- Gerber MA,
- et al
- Durongpisitkul K,
- Gururaj VJ,
- Park JM,
- Martin CF
- Copyright © 2014 by the American Academy of Pediatrics