Sociodemographic Disparities in Severity of Multisystem Inflammatory Syndrome in Children: The National Cross-Sectional MUltisystem Inflammatory Syndrome In Children (MUSIC) Study

Abstract

Objective

To assess sociodemographic disparities associated with multisystem inflammatory syndrome (MIS-C) severity and adverse outcomes.

Study design

This long-term outcomes after the MUltisystem Inflammatory Syndrome In Children (MUSIC) observational study included persons <21 years old with MIS-C hospitalized at 1 of 32 large US pediatric MUSIC medical centers. The primary outcome was a composite greater illness severity (eg, inotropic medications, intubations), and secondary outcomes were days from symptom onset to hospital admission and hospital length of stay (LOS). Predictor variables included patient distance to the hospital, neighborhood Social Deprivation Index, race and ethnicity, and non-primary English language.

Results

Among 1115 patients with MIS-C, median age was 9 years (IQR 5.6, 12.7), 39.2% were female, 28.1% were non-Hispanic Black, 27.8% were Hispanic, and 47.3% had public insurance. On multivariable analysis, more severe illness was seen among Hispanic patients (OR 1.5; 95% CI 1.1–2.2); non-Hispanic Black race (OR 1.7; 95% CI 1.2–2.4), and ages 13–21 years (OR 3.7; 95% CI 2.4–5.7). Longer time from symptom onset to hospital admission was associated with farther distance from hospital (P = .006). Risk factors for longer hospital LOS were adolescent age (≥13 years; P = .002), and non-Hispanic Black race (P = .018).

Conclusions

Disparities in MIS-C severity outcomes and hospital LOS were associated with older age and being Black and/or Hispanic. Given the ongoing and newer strains of COVID-19, it is important to understand disease severity and disparities in outcomes due to MIS-C. (J Pediatr 2025;285:114670).

SARS-CoV-2, which causes COVID-19, was first noted in Wuhan City in December 2019, igniting a worldwide pandemic that reached the US in early 2020. COVID-19 was initially noted to be generally mild or asymptomatic in children, causing few pediatric hospitalizations and minimal mortality compared with the illness in adults.^1–3 However, by the spring of 2020, reports of a postinfectious complication of COVID-19, now called the multisystem inflammatory syndrome in children (MIS-C), were described first in Europe^4–6 and then in North America.^7–9 Although a recent national database study demonstrated that greater numbers of COVID-19 infection in Black and Hispanic individuals were driving disparities in MIS-C,¹⁰ a separate national database study demonstrated that racial/ethnic disparities were worse for those with MIS-C compared with COVID-19, particularly for Black and male patients, as well as those with lower socioeconomic status (SES).¹¹ Given that the majority of these national database studies had significant study limitations, including reliance on accurate International Classification of Diseases coding of MIS-C and lack of disease-specific data (limited details on access to care and severity of disease on presentation), we sought to understand potential racial and ethnic disparities in MIS-C on a more granular level.

The long-term outcomes after the MUltisystem Inflammatory Syndrome In Children (MUSIC) study was conducted by the National Heart, Lung, and Blood Institute-funded Pediatric Heart Network. This study was developed to define details surrounding the population affected, as well as long-term cardiovascular and other organ system health status in one of the largest samples of children and adolescents with MIS-C to date.¹² This longitudinal multicenter study on MIS-C required harmonization of data elements across 32 participating North American sites and provided a unique framework to analyze the relationships of sociodemographic and socioeconomic factors with MIS-C outcomes and facilitate collaboration across existing registries and center projects. It provided a unique opportunity to obtain granular data at the national level and allowed for assessment of MIS-C outcomes across varying geographic, racial and ethnic, and socioeconomic pediatric populations.

In this secondary analysis of the MUSIC study database, we sought to assess sociodemographic disparities associated with MIS-C disease severity and adverse outcomes, including hospital length of stay (LOS) and delay in hospital presentation for care.

Methods

The design of the MUSIC study has previously been described.¹² In brief, this 32-center ambidirectional observational study included persons <21 years old who were hospitalized for MIS-C between June 2020 and January 2022. In this article, we excluded participants from the single Canadian study site because of differences in insurance and access to care between the US and Canada.

Our primary outcome variable was a composite of severe illness comprising >1 of the following: (1) use of vasopressors, including “shock requiring vasopressors” or “vasoactive infusions”; (2) myocarditis or cardiac dysfunction diagnosed during hospital stay (defined by an elevated troponin or left ventricular ejection fraction <55%, respectively); (3) use of mechanical ventilator support via endotracheal tube or tracheostomy; or (4) receipt of mechanical support with extracorporeal membrane oxygenation. Our secondary outcome variables included (1) days from onset of fever or symptoms to earliest admission (to any hospital); and (2) hospital LOS, defined as time from first admission (to any hospital) to hospital discharge (from the MUSIC site).

Our primary predictor variables included race and ethnicity, primary parent/caregiver language, neighborhood Social Deprivation Index (SDI; greater score is worse), geographic region, and distance of patient residence to the presenting hospital. Race and ethnicity were defined as Hispanic, Non-Hispanic Black, Non-Hispanic White, and Non-Hispanic other (Asian, multiple races, American Indian/Alaskan Native, Native Hawaiian/Pacific Islander, other, as well as unknown race and ethnicity). Information on race and ethnicity was collected from medical records or reports by parents, caregivers, or patients. Primary parent/guardian language spoken at home was defined as English vs non-English. We analyzed neighborhood SES using the SDI derived from ZIP code, because census tract data were limited. Geographic region included the 4 categories of Northeast, Midwest, South, and West, as defined by the US Census Bureau.¹³ Distance of patient residence to the presenting hospital was determined by the ZIP code of the patient and that of the medical center, and distance between ZIP codes was calculated as the geodetic distance in miles between 2 ZIP code locations, using the centroid of each ZIP code in the calculation. We also performed a subanalysis of individual hospital sites to determine associations with the primary outcome variables.

Covariates in our models included (1) age at hospital admission (or at first MIS-C symptoms, if missing); (2) sex; (3) insurance type, defined as private, public (governmental), or other/unknown; (4) presence or absence of an asthma diagnosis before MIS-C hospitalization; (5) COVID-19 era (Alpha-hospital admission June 2020 to June 2021 vs Delta-hospital admission July 2021 to December 2021); and (6) obesity, as defined by body mass index (BMI), calculated from height and weight measured at admission or, if not available, during the same hospitalization, and classified as normal or underweight (BMI z score ≤1.0), overweight (BMI z score >1.0 and ≤2.0), or obese (BMI z score >2.0).

The data are summarized descriptively, with all the continuous variables presented as mean SD or median along with IQR. Categorical variables are reported as frequencies and percentages. We used univariable analysis to assess the influence of each predictor on every outcome, employing linear and/or logistic regression models on the basis of the type of outcome. Predictors that showed a P value of less than .20 in the bivariate models qualified for inclusion in the multivariable models. To assess the combined impact of predictors, multivariable models were developed, employing backward selection method at the .05 level.

In the case of collinear predictors (eg, different categorizations of race), a decision on which to use was made on the basis of univariable results across outcomes. The following were chosen: race categories (instead of non-Hispanic White vs all others), SES as a continuous variable (rather than tertiles), and categorical age groups (rather than continuous). Data were analyzed using SAS EG ,version 8.3.

Results

Among 1132 patients with MIS-C included in the study, median age was 9 years (IQR 5.6, 12.7), 39.2% were female, 28.1% were non-Hispanic Black, 27.8% were Hispanic, 47.3% had public insurance, and median SDI was 54 (IQR 25.0, 82.5), with a greater SDI representing increased poverty (Table I). More patients were hospitalized at centers in Southern and Western regions than from those in Midwest and Northeast regions (Table I). Among 1110 participants with available data, 337 (29.7 %) were transferred to a MUSIC center from another hospital. Of 73 individuals living >100 miles from a MUSIC center, a greater proportion (67.1% vs 27.9%, P < .001) were transferred from another hospital, compared with those living ≤100 miles away, suggesting that distance from a MUSIC center could be considered a measure of access to a major medical center, rather than access to care in general.

Table I.

MUSIC MIS-C population, descriptive characteristics

Characteristics	Total (n = 1132)
Age at hospitalization, y
No.	1115
Mean (SD)	9.23 (4.57)
Median (Q1, Q3)	9.13 (5.60, 12.71)
Min, max	(0.14, 20.97)
Missing	17
Biological sex, No. (%)
Male	688 (60.8)
Female	444 (39.2)
Race and ethnicity, No. (%)
American Indian/Alaskan Native, non-Hispanic	1 (0.1)
Asian, non-Hispanic	27 (2.4)
Black, non-Hispanic	318 (28.1)
Hispanic or Latino	315 (27.8)
Multiple races, non-Hispanic	14 (1.2)
White, non-Hispanic	368 (32.5)
Other, non-Hispanic	11 (1.0)
Unknown or refused	78 (6.9)
Insurance information, No. (%)
Private	498 (44.0)
Self-pay	21 (1.9)
US government (eg, Medicaid)	536 (47.3)
Other governmental insurance outside the US	1 (0.1)
Dual coverage	16 (1.4)
Missing or unknown	60 (5.3)
Primary language spoken, No. (%)
English	997 (88.1)
Spanish	87 (7.7)
Both English and Spanish	39 (3.4)
Other	7 (0.6)
Unknown	2 (0.2)
Social Deprivation Index
No.	1124
Mean (SD)	53.05 (31.06)
Median (Q1,Q3)	54.00 (25.00, 82.50)
Min, max	(1.00, 100.00)
Missing	8
Geographic region, No. (%)
Northeast	195 (17.2)
Midwest	168 (14.8)
South	448 (39.6)
West	319 (28.2)
Unknown	2 (0.2)
Distance from medical center, miles
No.	1124
Mean (SD)	33.12 (51.08)
Median (Q1,Q3)	17.60 (8.90, 35.90)
Min, max	(0.00, 736.10*)
Missing	8
COVID-19 era, No. (%)
No.	1131
Alpha	841 (74%)
Delta	290 (26%)
BMI category, No. (%)
Normal/underweight	563 (49.7)
Overweight	268 (23.7)
Obese	209 (18.5)
Asthma, No. (%)
Yes	102 (9.0)
No	751 (66.3)
Unknown	279 (24.6)

^*The maximum distance was excluded as an outlier in modelling.

Univariate analysis selected candidate variables for further exploration in the multivariable analysis. The multivariable results indicated that age group, race and ethnicity, location, and distance from center were significantly associated with severe illness (Table II). Patients aged 6–12 years and 13–17 years showed significantly greater risk than those in the 0- to 5-year age group, with OR of 1.7 (95% CI 1.3–2.3) and 3.7 (95% CI 2.4–5.6), respectively. The odds of having severe illness were greater among Hispanic (OR 1.5; 95% CI 1.1–2.2) and non-Hispanic Black patients (OR 1.7; 95% CI 1.2–2.4) compared with non-Hispanic White patients. There was no effect of illness severity by COVID-19 era. Compared with patients residing in the Northeast, those in the Midwest were less likely to present with severe illness (OR 0.457; CI 0.29–0.73), whereas patients hospitalized in the South and West regions did not differ significantly from those in the Northeast. A subanalysis of illness severity by hospital site revealed that there were differences in severity by site, with 5 sites having statistically significantly lower severity (OR 0.29–0.48; P < .01), and 1 site having statistically significantly higher severity (OR 2.33; P < .01) compared with the largest site. However, individual sites were anonymized upon study analysis, and thus we had no further data on location of these sites. Finally, every additional 100 miles that participants lived from the MUSIC center was associated with 3.42 (95% CI 2.0–6.0) times greater odds of severe illness. The model C-statistic was 0.682.

Table II.

Final multivariable model of greater severity of illness

Predictors	OR	CI	P value
Age group, y			<.001
0–5	Reference
6–12	1.696	(1.25–2.31)
13–21	3.707	(2.43–5.65)
Unknown	>999.99	(0.00, >999.99)
Race and ethnicity			.029
Black, non-Hispanic	1.678	(1.16–2.43)
Hispanic	1.543	(1.06–2.24)
White, non-Hispanic	Reference
Other, non-Hispanic	0.97	(0.50–1.87)
Unknown	1.036	(0.59–1.82)
Geographic region			.001
Northeast	Reference
Midwest	0.457	(0.29–0.73)
South	0.754	(0.50–1.13)
West	1.066	(0.68–1.66)
COVID-19 era			.91
Delta	Reference
Alpha era	1.31	(0.96–1.81)
Distance from center (per 100 miles)	3.42	(1.95–6.00)	<.001

Observation used (No.) = 1093, R² = 05.091. c-statistic = 0.686.

In multivariable linear regression, independent predictors of longer time from symptom onset to presentation to the hospital were lower SDI and longer distance to the hospital. Lower SDI (greater SES) was associated with longer time from onset of symptoms to hospital admission (P = .009), as was longer distance from the medical center (0.0049 additional days per mile, equivalent to 1 additional day per 204 miles; P = .006). The symptom onset to presentation to the hospital was not associated with COVID-19 era or by hospital site. The model explained a very low percent of variance in this outcome (R² = 0.015, n = 1088).

The results of multivariable analysis for predictors of longer hospital LOS are shown in Table III. Independent risk factors included being in the 13- to 21-year age group compared with a reference group of those ages 0–5 years old (P = .001); race and ethnicity (P = .028), with longer LOS for those in the non-Hispanic Black and non-Hispanic Other groups compared with non-Hispanic White patients; other/unknown insurance (P < .001) compared with a reference of private insurance; and Midwest location compared with the reference group of those in the Northeast (P = .017) (Table III). An unknown asthma diagnosis was associated with a longer LOS (P = .012). However, these factors explained little variation in outcome (R² = 0.063).

Table III.

Final multivariable model of hospital LOS (days)

Predictors	Parameter estimate	SE	P value
Age group, y			.001
0–5	Reference
6–12	−0.64	0.46
13–21	1.1	0.54
Race and ethnicity			.028
Black, non-Hispanic	1.15	0.53
Hispanic	0.88	0.53
White, non-Hispanic	Reference
Other, non-Hispanic	2.68	0.95
Unknown	0.08	0.82
Insurance			<.001
Private	Reference
US government	0.6	0.44
Other/unknown	2.73	0.72
Location			.017
Northeast	Reference
Midwest	1.34	0.68
South	1.04	0.57
West	−0.4	0.59
Missing	−0.21	4.48
Asthma			.012
No	Reference
Yes	0.56	0.67
Missing	1.37	0.46

Observation used (No.) = 1097, R² = 0.063.

Discussion

In this large, ambidirectional study on MIS-C, we found disparities in the association between sociodemographic variables and MIS-C severity outcomes. Our study was a very large, cross-sectional study on MIS-C in which data were collected at individual MUSIC participating institutions to ensure granularity of data, allowing for the determination of severity of illness on presentation, time from symptom to first hospital presentation, as well as LOS. Independent risk factors for the composite outcome of severe MIS-C disease included Hispanic ethnicity and non-Hispanic Black race, as well as greater distance from home to the study center. Lower social deprivation (eg, better neighborhood SES) and longer distance from home to the study center were independent risk factors for longer time from illness onset to admission to a hospital. Finally, independent risk factors for longer hospital LOS included race and ethnicity, with worst outcomes in those who were non-Hispanic Black and non-Hispanic other, as well as having other/unknown insurance type. Notably, COVID-19 era did not appear to be associated with our identified outcomes.

Our findings build on previous literature showing the impact of race and ethnicity on MIS-C risk. Early in the pandemic, Hispanic and non-Hispanic Black children and young adults were identified as having disproportionately greater incidence of MIS-C.^14–17 The enhanced risk for MIS-C could not be solely attributed to greater rates of SARS-CoV-2 infection in these populations. Among children infected with SARS-CoV-2, Black and Hispanic/Latino children had adjusted incidence rate ratios for MIS-C that were 5.62 and 4.26 times greater than for White children.¹⁷ In another study of children and adolescents hospitalized with either with acute COVID-19 or MIS-C, Black-non-Hispanic (vs White non-Hispanic race and ethnicity) was a risk factor for MIS-C diagnosis. The role of neighborhood social determinants of health in the predilection of Black children for MIS-C has been a topic of some exploration. Using the National Inpatient Sample database, Ghimire et al found greater risks of hospitalization with MIS-C among Hispanic and Black children, compared with White children, in the lowest income quartile neighborhood but not the highest income quartile neighborhood.¹⁰ Other studies have noted LOS disparities for Black patients with MIS-C were exacerbated by socioeconomic factors, with a 1-day increase in LOS when moving from lowest to highest Social Vulnerability Index quartile.^11,18 Our current study showed a greater risk of severe disease in Hispanic and non-Hispanic Black patients, as well as longer hospital LOS for non-Hispanic Black and non-Hispanic other patients, a heterogeneous group that is challenging to characterize. Of note, although time from onset of fever/symptoms to hospital admission did not significantly differ among the hospital sites, there was a significant difference between sites in terms of greater severity of illness (some sites had greater prevalence of greater disease severity when compared with other sites). Thus, greater illness severity and longer LOS in these groups do not appear to be explained by greater social deprivation or longer time from symptom onset to presentation to a hospital, suggesting that other medical, environmental, or structural factors may be at play. It is likely that the LOS difference seen with Black patients is multifactorial, potentially including a lack of early recognition of MIS-C these patients, a more severe MIS-C presentation, or the intersectionality between existing poorer social determinants of health and MIS-C recovery and factors related to hospital discharge, including transportation, parental ability to remain in the hospital as the result of opportunity costs, etc. This has been seen in other studies as well, including one of the Kids’ Inpatient Database, that showed longer LOS for children of historically marginalized race and ethnicity with common pediatric inpatient diagnoses, which largely persisted from 2016 to 2019.¹⁹ There is no plausible biological explanation for these findings, and inequities in social needs, access to care, and quality of care likely contribute, as well as other factors, including systemic racism that result in poorer social determinants of health.^19,20

Previous studies have reported that the most common underlying condition in children with MIS-C is obesity, followed by asthma.²¹ In our study, however, neither obesity nor asthma was associated with more severe MIS-C disease presentation or longer LOS. A missing diagnosis of asthma was associated with longer LOS—it may be that some of those without a formal diagnosis listed carried an asthma diagnosis; however, these data were not available. There are existing data in children demonstrating that obesity is a risk factor for decreasing lung function after COVID-19 infection in children with asthma,²² but this association with asthma and obesity may not hold true for MIS-C, as had been noted in some studies.²³ It may be that although obesity is associated with worse morbidity and mortality outcomes in adults with COVID-19, that the underlying pathophysiology of MIS-C is not influenced by obesity or asthma and thus does not portend worse outcomes, as is seen in other studies.²⁴

Our study found that SDI, a measure of neighborhood socioeconomic status, was inversely associated with days from symptom onset to presentation to a hospital, ie, children from areas of less deprivation were hospitalized later in illness. The reasons for this surprising finding are unclear. We hypothesize that hospitalization after longer days of illness among patients with greater SES scores could be explained, in part, by their greater access to care at their local primary care physician, which may have been sought as a first line of assessment before referral for hospitalization. It is also possible that families with greater SES scores more precisely documented the first onset of fever. Finally, this finding could be the result of residual confounding by other unmeasured factors.

In our study, longer distance between a family’s residence and the study center was associated with greater likelihood of severe disease, as well as later presentation from fever onset to presentation to a hospital (local or study site). Across the study, nearly 30% of patients were transferred from local hospitals to a MUSIC center, whereas among the subgroup of patients living a further distance to the hospital, roughly 67% were transferred to MUSIC centers. These data suggest that there may be selection bias for patients transferred from a greater distance to have more severe disease requiring tertiary care at an academic pediatric institution.

It is unclear why patients in the Midwest region both had lower severity of illness as well as a longer LOS. It may be that Black and Hispanic patients were more concentrated in urban centers nearer to academic MUSIC-type institutions, and thus they did not have the same challenge with accessing care. It may be that the hospitals sites noted to have a lower severity of illness were concentrated in the Midwest. It may also be the case that, given the data on morbidity and mortality for adult minoritized patients with COVID-19, that a more conservative approach was taken with patients with MIS-C and may play a role in the longer LOS.²⁵

Limitations to this study should be noted. Race and ethnicity were unavailable in 6.9% of our subjects because of decline of families to self-identify these factors or to missing data. Participants in MUSIC study were seen at major medical centers, limiting generalizability to smaller or nonpediatric institutions. Our primary composite outcome for severe illness was created by consensus among investigators for use in this study. SDI was calculated on the basis of ZIP code because census county data were unavailable. Census tracts would have been preferable, because they provide more granularity than ZIP Codes, and ZIP codes exist only where US mail service is provided. Further, census tracts have a large and richer set of associated, more reliable demographic-economic data.²⁶ Although sociodemographic and environmental risk factors were identified, they explained a small amount of variation in outcomes, suggesting that other unidentified factors may be more predictive of these outcomes. Finally, our study design does not allow causal inference.

Recognition of associations of more severe MIS-C illness allow for exploration of next steps in assessing and ideally intervening on contributing factors at the population and institutional level. Potential interventions to improve treatment of MIS-C and other severe acute illnesses in future pandemics include improving community level of knowledge in their diagnosis, recognition of their presentations in diverse populations, and pathways for triage of patients with residence at distance from large cardiac centers, including via use of telemedicine.

Glossary

BMI

Body mass index

LOS

Length of stay

MIS-C

Multisystem inflammatory syndrome

MUSIC

Long-term outcomes after the Multisystem Inflammatory Syndrome In Children

SDI

Social Deprivation Index SES Socioeconomic status