Distinguishing Incomplete Kawasaki and Nonsevere Multisystem Inflammatory Syndrome in Children

With institutional review board approval, we performed a retrospective cohort study at a tertiary pediatric healthcare system in the southeastern United States comprised of 3 freestanding hospitals with >600 beds. In the nonsevere MIS-C cohort, we reclassified MIS-C cases based on the 2023 definition from March 2020 to February 2021.²  In the iKD cohort, we identified children who met the American Heart Association criteria⁶  for iKD and received intravenous immunoglobulin (IVIG) from January 2018 to January 1, 2019. We excluded those who received ICU care or had coronary abnormalities on echocardiogram during their acute illness and follow up from both cohorts.

We excluded children with prepandemic iKD with coronary abnormalities as they had no diagnostic uncertainty. We excluded children with MIS-C with coronary abnormalities because it would be difficult to know whether they had MIS-C or KD.⁷  We focused on children without ICU care because 40% to 50% of patients with MIS-C do not require ICU care,^8–11  resembling the prepandemic iKD cohort.

The MIS-C likelihood score developed by Godfred-Cato et al was applied to both cohorts.⁵  It includes 7 criteria, with corresponding point values. Thrombocytopenia (platelet < 150 × 10³/µL) with 2 points, abdominal pain, headache, pericardial effusion, and C-reactive protein (CRP) >10 mg/dL each worth 1 point, rash with −1 point, and mucocutaneous lesions with −2 points. The summed score ranges from −3 to 6. Each score correlates to a probability ratio, pMIS-C, and the higher the score with its corresponding ratio, the more likely MIS-C is the diagnosis. A score of 0 implies diagnostic ambiguity. A score of <0 suggested KD, whereas a score of >0 suggested MIS-C.

Utilizing the 2023 report for MIS-C as a guide,¹  data from children were extracted into an electronic database (Research Electronic Data Capture, Vanderbilt University). Epidemiologic data were collected. Clinical features including fever duration before treatment, presence of the five cardinal KD symptoms, and other organ system involvements associated with MIS-C were recorded.^5,10,12–14  Cardiac dysfunction (ejection fraction <55%) was noted. Troponin level was not compared because it was not collected for iKD. Laboratory results on admission and the highest and lowest white counts before treatment were collected.

Statistical analysis including descriptive statistics and univariable analysis was conducted. Comparisons were conducted by the Wilcoxon Rank Sum test for continuous variables, χ-square test for categorical variables, or Fisher exact test for categorical variables with small cell size (n ≤ 5). A P value < .05 was significant.

We included 68 children in the nonsevere MIS-C cohort and 28 in the iKD cohort (Supplemental Fig 1). Twenty-four (86%) of the iKD cohort met organ system and inflammatory criteria in the 2023 MIS-C definition and 44 (65%) in the nonsevere MIS-C cohort had ≥2 KD features.

The median age (years) and interquartile ranges for nonsevere MIS-C and iKD cohorts were 8 (5–10) and 4 (2–7.25; P < .001), respectively (Table 1). Median fever duration (days) before treatment of nonsevere MIS-C and iKD cohorts were 6 (5–7) and 8 (5–10; P < .001), respectively. Of the cardinal KD symptoms, oral mucosal changes and rash were significantly likely to be present in iKD cohort (64% vs 35%, P = .009, and 82% vs 56%, P = .015, respectively). Cardiac dysfunction was more likely to present in nonsevere MIS-C cohort (P = .034) (Table 1).

The iKD cohort was more likely to present with leukocytosis before treatment (39% vs 4%; P < .001), whereas the nonsevere MIS-C cohort was more likely to have lymphopenia (43% vs 18%; P = .021) or thrombocytopenia (40% vs 0%, P < .001). The iKD cohort was more likely to have pyuria from documented urinalyses (61% vs 20%, P < .001). The levels of CRP were comparable between cohorts (P = .162) (Table 2). All patients in the nonsevere MIS-C cohort tested positive for coronavirus disease 2019 polymerase chain reaction or serology.

The median score was higher in nonsevere MIS-C cohort (2 vs −1.5, P < .001) (Table 3). The score correctly predicted the diagnosis as MIS-C for 50 (74%) children in nonsevere MIS-C cohort and KD for 20 (71%) children in iKD cohort. However, 9 (13%) and 4 (14%) children in nonsevere MIS-C and iKD cohorts, respectively, had a score of 0, 50% chance of having either diagnosis. Furthermore, 9 (13%) of the nonsevere MIS-C cohort and 4 (14%) of the iKD cohort had scores that predicted diagnoses that were incongruent with the surveillance diagnoses, for example, the score predicted MIS-C, whereas iKD was the documented diagnosis. For those 8 patients in the iKD cohort with 0 or positive score, 6 of them had a CRP >10mg/dL.

The clinical overlap of MIS-C and KD has been well described, with 30% to 40% of children with MIS-C diagnosed based on the 2020 definition³  meeting KD criteria and the majority being iKD.^12,13,15  We found a remarkable overlap in clinical and laboratory parameters between noncritically ill patients treated for iKD before the pandemic and MIS-C based on the 2023 definition.²  Eighty-six percent of the iKD cohort could have potentially met the 2023 definition of MIS-C had they had a SARS-CoV-2 infection or exposure and 65% of the nonsevere MIS-C cohort had ≥2 KD features. Although it is possible some with nonsevere MIS-C would have been treated as having iKD prepandemic because of phenotypical similarities, key differences were children with nonsevere MIS-C were older, more likely to have cardiac dysfunction, thrombocytopenia, and lymphopenia, and less likely to have pyuria.

The clinical overlap between iKD and nonsevere MIS-C is important to recognize, as the seroprevalence of SARS CoV-2 antibody is increasing over time, especially after the ο wave and subsequent variants.⁴  As nonsevere MIS-C can be treated with corticosteroid monotherapy with good short-term outcomes^16,17  and as IVIG is a critical first-line therapy for children with KD, differentiating these 2 conditions has clinical relevance.

Using the MIS-C likelihood scale developed earlier,⁵  >25% of the patients had a score that was either inconclusive or suggested alternative classification. In these situations, lymphopenia and thrombocytopenia may be more suggestive of MIS-C. Clinicians will also need to consider the epidemiologic trend of SARS-CoV-2 infection in their community and the prevalence of MIS-C. More recent studies found the prevalence of MIS-C is decreasing with an increasing proportion of cases not requiring ICU stays.^11,18 

Strengths of the study include the standardized data collection, specific exclusion criteria, application of the new MIS-C definition, and first documented application of the MIS-C likelihood score to those with iKD without coronary involvement. The standardized data collection allows for easy comparison between cohorts, whereas the exclusion criteria helped identify nonsevere presentations, resembling current clinical scenarios.^1,9,18  There are several limitations, including the retrospective design; clinical symptoms may not have been consistently collected. Subjects included in the iKD cohort had 2 to 3 KD symptoms on exam unless they were infants. We did not include patients treated but with <2 symptoms, even if they had other KD symptoms in history. We might be undercounting the incidence of iKD. However, we reduced the bias that may present with the decision to treat for iKD between clinicians.

It is unclear if nonsevere MIS-C was more prevalent in older children because of characteristics of the condition or if older children were more likely to be exposed and, hence, more likely to develop MIS-C.

There was considerable overlap between children with nonsevere MIS-C and iKD prepandemic with 86% of iKD cohort meeting the organ and inflammatory criteria of the 2023 MIS-C definition. The MIS-C likelihood score predicted concordance in diagnosis for <75% children from either cohort. Our study demonstrated that findings from a basic blood count and urinalysis were more useful differentiating features than the presence of organ system involvements. Therefore, clinicians may focus on the differences on complete blood count and urinalysis if there are diagnostic challenges between the two presentations.