Impact of race and socioeconomic deprivation on clinical outcomes and healthcare utilization in pediatric multiple sclerosis

Abstract

Background

Health disparities in adult-onset multiple sclerosis have been identified in the Black/African American (AA) population. A higher relapse rate has been suggested in Black/AA patients with pediatric-onset MS (POMS), but little work explores healthcare utilization and social determinants of health (SDOH).

Objective

To evaluate racial, ethnic, and socioeconomic disparities in POMS outcomes.

Methods

Retrospective chart review identified 31 eligible patients diagnosed with POMS at Children's of Alabama between 2013 and 2023. Demographics, outcomes, and healthcare utilization over 2 years from diagnosis were collected. Patient addresses were connected to SDOH measures from the US Census. Bivariate analysis was performed using Fisher's Exact Test, Wilcoxin Test, and 2-sided t-test.

Results

Black/AA children had a higher Expanded Disability Status Scale (EDSS) at first presentation (p = 0.0276) and were more likely to initiate fingolimod vs. glatiramer acetate (p = 0.0464). Living further from Children's of Alabama was associated with a higher most recent EDSS (p = 0.0301) and fewer neurology appointments (p = 0.0167). Families living in more socioeconomically deprived census tracts had significantly more hospital admissions.

Conclusion

Black/AA POMS patients had a more severe initial presentation and were started on higher efficacy medication. We identified disparities in EDSS and healthcare utilization based on SDOH data linked to a child's home address.

Introduction

Pediatric-onset multiple sclerosis (POMS) constitutes 3–10% of multiple sclerosis (MS) cases and follows a more inflammatory disease course than that of adult-onset MS (AOMS).^1,2 Children experience a higher lesion burden at diagnosis and on surveillance imaging as well as a higher relapse rate than their adult counterparts. ³ In the United States, the prevalence of MS in Black/African American (AA) patients is increasing, suggesting the disease may have been historically underrecognized in this group. ⁴ In AOMS, higher disease severity and disability accumulation have been described in US Black and Hispanic/Latinx individuals as compared to White individuals.^5,6 A series of 46 POMS patients found that Black children had a higher annualized relapse rate (ARR) compared to White children. ⁷ In a single-center cohort of 42 patients, Black children scored lower on language and complex attention testing than their White counterparts, independent of socioeconomic status. ⁸ As race is a social construct, these healthcare disparities cannot be presumed to be secondary to an undefined genetic predisposition alone.

Structural racism, in which societal infrastructure reinforces inequitable distribution of resources such as housing, income, education, criminal justice, environmental hazards, and healthcare, is a known driver of health disparities. ⁹ Historically marginalized populations face disadvantages in these and other social determinants of health, which are the conditions in which people are born, grow up, learn, and work. ¹⁰ In a large case-control study, lower household income, lower educational attainment, and being Black or Hispanic/Latinx were all independently associated with lower cognitive scores in AOMS patients. ¹¹ Higher educational attainment and premorbid income were associated with lower disability in a study through the Swedish MS Registry. ¹² Inequities in healthcare access and healthcare utilization can result in diagnostic delays and poorer clinical outcomes. Black and Hispanic/Latinx patients with neurologic conditions including MS were found less likely to have subspecialty care, and Black patients had higher Emergency Room utilization.^13,14 Black patients were found to have lower home and community-based service utilization than White patients (specifically in equipment, nursing, and case management). ¹⁵ In a large southwestern medical center, after adjusting for SDOH factors, race and ethnicity were no longer significantly associated with disability level. ¹⁶ Such work demonstrates the increasing focus on investigating the impact of SDOH in AOMS, but no significant work has been done in POMS.^6,17,18

Individual SDOH have been found to correlate with neighborhood level factors collected by the US Census based on address. ¹⁹ These neighborhood-based measures known as geomarkers are defined as “any objective, contextual, or geographic measure that influences or predicts the incidence of outcome or disease” and serve as a surrogate for daily living conditions that impact health outcomes. ²⁰ Such factors include neighborhood poverty, crime, transportation availability, pollution, population crowding and access to healthy food and healthcare. ²⁰ In an AOMS cohort, greater neighborhood disadvantage was associated with higher disability and lower quality of life. ²¹

The present study seeks to evaluate disparities with individual SDOH, as well as, assess the relationship between SDOH factors, healthcare utilization, clinical outcomes, and race and ethnicity.

Materials and Methods

Retrospective chart review was performed under a human research and subject protocol approved by the University of Alabama at Birmingham institutional review board. Informed consent was waived because this was a database and chart review without direct patient contact. Inclusion criteria were diagnosis with POMS based on 2017 McDonald criteria at first presentation to Children's of Alabama between 2013 and 2023. Radiologically isolated syndrome and clinically isolated syndrome with later development of MS were excluded so as to focus on time from symptom onset to diagnosis.

Demographics of sex assigned at birth, age at presentation, self-reported race and ethnicity, insurance status (public/uninsured vs. private) were collected. Clinical characteristics and outcomes, along with health utilization metrics, were extracted. Characteristics at onset were collected including days from symptom onset to diagnosis of POMS, presentation of optic neuritis, presentation of transverse myelitis, cerebral localization, brainstem presentation, presence of spinal cord lesions, inability to ambulate without assistance, and initial DMT. Subsequent clinical outcomes over 2 years from diagnosis were gathered, including annualized relapse rate (ARR), time to relapse, new lesion(s) detected on MRI, and requirement to switch DMT. EDSS was collected from chart documentation at time of onset, at 2 years, and most recent follow up after 2 years, with duration of follow up. Healthcare utilization measures included cumulative encounters over 2 years from diagnosis of hospital admissions, neurology appointments, and MRI scans. Patient addresses were geocoded with HIPAA-compliant DeGAUSS software and connected to SDOH measures from the US Census. These included proportion of population in census tract (receiving assisted income, obtaining high school education, without health insurance, living in poverty), median income of population, and proportion of houses in census tract that are vacant. Deprivation index, which incorporates these measures, was extracted. ²² Distance from Children's of Alabama was also determined via DeGAUSS software.

Statistical analysis

Descriptive statistics were calculated for all variables: n (%) for categorical variables, mean ± standard deviation (SD) or median (med) and (minimum, maximum) for continuous variables. Medians for continuous variables compared using Wilcoxon test with a Steel-Dwass all pairs post-hoc comparison as applicable, proportions for categorical variables were compared using a Fisher's Exact test. Nonparametric Spearman's rho was utilized for correlation between geocoded values and SDOH metrics. Unadjusted means comparisons for geocoded metrics were compared with a two-sided t-test. Covariate adjusted ANOVA/linear regression, with Tukey's Honestly Significant Difference, was used to assess relationship of race and healthcare utilization (predictors) with geocoded outcomes. Due to small sample size of Hispanic and/or Latino persons, ethnicity was not considered as a covariate (n = 3). Other than Steel-Dwass for Wilcoxon or Tukey's Honestly Significant Difference test for the ANOVA, no adjustments were made for multiple testing, and missing data were not imputed; values of p < 0.05 were considered meaningful. All analyses were conducted JMP Pro 16 (Cary, NC).

Results

There were 33 patients identified, 16 Black/AA, 15 White, and 3 patients of Hispanic Ethnicity. Of the Hispanic patients: two were female and one was male, with a mean age of onset of 14.7 (SD 1.5, range 13–16) years of age. One female reported White race; the two other patients declined to self-report race and are otherwise not included in the analysis (n = 2 excluded). The Hispanic male was Spanish-speaking; all other patients in our cohort were English-speaking. This resulted in 31 patients included for analysis.

Relationship of race with clinical characteristics and outcomes

There were no significant differences based on race in sex assigned at birth, age of onset, or duration of follow-up (Table 1). There was a difference in the initial DMT started by race (p = 0.0464). Even when excluding the single patient on interferon therapy, the difference persists (p = 0.0467). Black/AA children were more likely to start on fingolimod, whereas White children more likely to start on glatiramer acetate. There was no difference in the proportion of patients requiring to switch DMT during follow-up by race (Table 1).

Table 1.

Participant demographics (N = 31).

Characteristic	All	Black/AA (16, 51.6%)	White (15, 48.4%)	p-value
Gender, n(%): Female	16 (51.6)	11 (68.8)	9 (60.0)	0.7160*
Age (years): Mean (SD)	14.2 (1.8)	13.9 (1.8)	14.6 (1.8)
Med (min, max)	14 (11, 17)	14 (11, 17)	15 (11, 17)	0.2148**
Follow-up (years): Med (min, max)	3.2 (2, 8.5)	3.8 (2.1, 8.5)	2.8 (2, 7.3)	0.1291**
DMT Initial, n(%)
Glatiramer acetate	9 (29.0)	2 (12.5)	7 (46.7)	0.0464*
Fingolimod	10 (32.3)	8 (50.0)	2 (13.3)
IFN	1 (3.2)	0 (0)	1 (6.7)
Rituximab	11 (35.5)	6 (37.5)	5 (33.3)
DMT Switch, n(%): Yes	16 (51.2)	7 (43.8)	9 (60.0)	0.4795*

Abbreviations: DMT: disease modifying therapy; IFN: interferon; DMT switch: requirement to switch DMT; Med: median; SD: standard deviation.

*Fisher's Exact Test, **Wilcoxon Test.

Type of demyelinating syndrome at onset did not vary based on race, nor did the presence of spinal cord lesions on MRI imaging (Table 2). Similar ARR and time to relapse were seen in Black/AA and White patients.

Table 2.

Disease characteristics by on race (N = 31).

Characteristic	All	Black/AA (16, 51.6%)	White (15, 48.4%)	p-value
Present at Onset, n(%): Optic Nerve	6 (19.4)	5 (31.3)	1 (6.7)	0.1719*
Transverse Myelitis	7 (22.6)	3 (18.8)	4 (26.7)	0.6851*
Cerebral Syndrome	1 (3.2)	1 (6.3)	0 (0)	>0.9999*
Brainstem	13 (41.9)	9 (56.3)	4 (26.7)	0.1489*
Spinal lesion	22 (73.3)	11 (67.8)	11 (78.6)	0.6687*
Unable to walk at motor nadir	3 (9.7)	2 (12.5)	1 (6.7)	>0.9999*
Symptom to Dx (days): Med (min, max)	84 (3, 738)	71 (6, 738)	84 (3, 367)	0.4063**
EDSS 1st: Med (min, max)	2.5 (0, 6)	3 (1, 6)	2 (0, 5)	0.0276**
EDSS 2nd: Med (min, max)	1.5 (0, 3.5)	1.5 (0, 3.5)	2 (0, 3)	0.5033**
EDSS Last: Med (min, max)	1.5 (0, 4.5)	0.5 (0. 4.5)	1.5 (0, 3)	0.5837**
EDSS 2nd-1st: Med (min, max)	−3 (-5, 1)	−2 (-4.5, 0.5)	0 (-5, 1)	0.0226**
EDSS Last-1st: Med (min, max)	−2 (-5, 2)	−2 (-3.5, 2)	−0.5 (-5, 1)	0.0201**
MRI Progression, n(%): Yes	18 (58.1)	8 (50.0)	10 (66.7)	0.4725*
Relapse, n(%): Yes	18 (58.1)	9 (56.3)	9 (60.0)	>0.9999*
Follow Up (years): Med (min, max)	3.2 (2, 8.5)	3.8 (2.1, 8.5)	2.8 (2, 7.3)	0.1291**
ARR: Med (min, max)	0.5 (0, 3.5)	0.5 (0, 3.5)	0.5 (0, 3.5)	0.5734**
Time to Relapse (days, n = 18): Med (min, max)	124 (30, 1062)	210 (90, 365)	120 (30, 1062)	0.2360**

Abbreviations: spinal lesion: presence of spinal lesion(s) on initial MRI; Symptom to Dx: time in days from symptom onset to diagnosis; EDSS 1^st: EDSS at onset; EDSS 2^nd: EDSS at 2 years from diagnosis; EDSS last: EDSS at most recent follow-up; EDSS 2^nd-1^st: change in EDSS from onset to 2 years from diagnosis; EDSS last-1^st: change in EDSS from onset to most recent follow-up; MRI Progression: new lesions on subsequent MRIs, relapse: clinical relapse within 2 years from diagnosis, follow-up: years from diagnosis to most recent follow-up, ARR: annualized relapse rate for 2 years from diagnosis, time to relapse: time from onset attack to first relapse.

*Fisher's Exact Test, **Wilcoxon Test.

Black/AA patients had a median EDSS 1-point higher than White patients at time of first EDSS (p = 0.0276, Wilcoxon). However, by time of 2-year and most recent EDSS, there were no differences seen by race (p = 0.5033, 0.5837, respectively). Given that Black/AA patients had a higher initial EDSS but similar EDSS at end of follow up, it is not surprising that Black/AA patients also exhibited a greater improvement in EDSS from initial to 2-year EDSS and from initial to most recent EDSS (p = 0.0266 and 0.0201 Wilcoxon, respectively) (Figure 1).

Figure 1.

EDSS over time by race. Solid line Black/AA, Dashed line White; unadjusted mean (95% CI) EDSS.

The three patients of Hispanic ethnicity had a mean initial EDSS of 2.5 (range 2-3), 2-year EDSS of 1 (range 0–2), and most recent EDSS of 1 (range 0–2). The mean number of days from symptom onset to diagnosis was 387.3 (standard deviation 447.5, range 124-904).

Hospital utilization metrics within the first two years following diagnosis (median number of inpatient days, number of admissions, number of MRI scans, number of neurology appointments) did not significantly differ based on race.

Characterization of social determinants of health by race

Neighborhood-level indices of SDOH were generated based on patient address. Language and insurance were also collected from patient records.

All geocoded metrics were different by race, excepting fraction of neighborhood population without health insurance (Table 3). Black/AA patients in our cohort lived in neighborhoods which had a higher fraction of the population on assisted income, lower level of high school education, lower median income, higher proportion of population without health insurance, higher proportion in poverty, higher proportion of vacant housing, and a higher deprivation index. White patients lived significantly further away from Children's of Alabama compared to Black/AA patients (54.0 vs. 28.6 miles). Insurance status also varied based by race (p = 0.0091). Black/AA patients in our cohort were more likely to have public insurance compared to White patients (62.5% vs. 13.3%).

Table 3.

Geocoded metrics by race (N = 31).

Metric	All	Black/AA (16, 51.6%)	White (15, 48.4%)	p-value
Assisted Income: (Mean, SD)	0.16 (0.13)	0.23 (0.14)	0.09 (0.05)	0.0010*
High School Education: (Mean, SD)	0.85 (0.08)	0.82 (0.07)	0.88 (0.08)	0.0337*
Median Income: (Mean, SD)	52579.8 (19674.0)	41878.5 (16731.5)	63994.5 (16093.9)	0.0008*
No Health Insurance: (Mean, SD)	0.10 (0.04)	0.11 (0.01)	0.09 (0.04)	0.0555*
Poverty: (Mean, SD)	0.17 (0.13)	0.23 (0.13)	0.11 (0.08)	0.0027*
Vacant Housing: (Mean, SD)	0.17 (0.09)	0.23 (0.08)	0.11 (0.07)	0.0002*
Deprivation Index: (Mean, SD)	0.40 (0.12)	0.47 (0.12)	0.32 (0.08)	0.0004*
Distance (miles): (Mean, SD)	40.9 (32.1)	28.6 (21.2)	54.0 (37.1)	0.0248*

Abbreviations: SD: standard deviation. Metrics - Proportion of population [receiving assisted income, obtaining high school education, without health insurance, living in poverty], median income of population, proportion of houses that are vacant, distance from Children's of Alabama; *two-sided t-test.

Relationship of social determinants of health with outcomes and healthcare utilization

Living a further distance from Children's of Alabama was associated with a significant increase in most recent EDSS (p = 0.0301; OR 1.05 [95% CI 1.004, 1.09]). For every additional mile further away from the hospital, most recent EDSS was 5% more likely to be 1 or higher compared to 0. Patients living closer to Children's of Alabama had significantly more improvement in EDSS from initial to most recent visit (p = 0.0058) (Figure 2). There was no interaction between race and distance from Children's Hospital when predicting Last EDSS or change in EDSS score. However, there was a significant interaction between race and first EDSS score when predicting a change in EDSS score (p = 0.0457). Whites with a higher starting EDSS (2.5 or greater), saw larger improvements compared to Black patients with same starting EDSS, regardless of distance from Children's Hospital. There were no other differences found in clinical outcomes (time from symptom onset to diagnosis, EDSS measures, ARR and time to relapse, MRI progression) based on SDOH factors.

Figure 2.

Change in EDSS by distance from children's hospital and starting EDSS Score. Change in EDSS from initial to most recent follow-up. Dashed line starting EDSS 0–2, Solid line starting EDSS 2.5–6. Least Squares Means from covariate adjusted model with 95% CI on predicted LS Means.

The interplay of SDOH with healthcare utilization was also investigated. No differences were found based on SDOH metrics in the number of MRIs during the first two years after POMS diagnosis. The number of hospital admissions during this period was trichotomized (0, 1, 2-3) and found to have differences based on SDOH variables (Table 4). Families with 2-3 admissions were found to live in census tracts with lower proportion of completed high school education, lower median income, higher proportion in poverty, and a higher deprivation index as compared to families with 1 admission and with 0 admission (Table 4). There was no interaction between race and healthcare utilization measures in models predicting geocoded metrics (all p > 0.15, results not shown).

Table 4.

Socioeconomic deprivation by number of hospital admissions (N = 29).

Hospital admissions Med (min, max)	None (n = 5)	1 Admission (n = 18)	2-3 Admissions (n = 6)	p-value*
Assisted Income	0.10 (0.07, 0.22)	0.10 (0.02, 0.53)	0.18 (0.12, 0.37)	0.1373
High School	0.86 (0.82, 0.95)	0.87 (0.74, 0.98)	0.77 (0.63, 0.82)	0.0043
Median Income	55734.00 (48938, 92974)	53320.50 (15651, 82713)	32306.50 (26553, 43991)	0.0124
No Health Insurance	0.09 (0.04, 0.15)	0.09 (0.02, 0.18)	0.12 (0.11, 0.14)	0.1864
Poverty	0.11 (0.06, 0.22)	0.12 (0.03, 0.54)	0.25 (0.15, 0.38)	0.0446
Vacant Housing	0.13 (0.04, 0.37)	0.14 (0.06, 0.34)	0.20 (0.07, 0.29)	0.6078
Deprivation Index	0.39 (0.23, 0.45)	0.35 (0.23, 0.72)	0.49 (0.42, 0.63)	0.0265
Distance (miles)	29.26 (14.86, 84.32)	29.07 (1.66, 154.7)	23.33 (2.18, 74.0)	0.4524

Abbreviations: SD: standard deviation; Med: median; CI: confidence interval. Metrics - Proportion of population [receiving assisted income, obtaining high school education, without health insurance, living in poverty], median income of population, proportion of houses that are vacant, distance from Children's of Alabama; *Wilcoxon.

The number of neurology appointments during this period were dichotomized (4–7, n = 12 and 8 or more, n = 13). Those families with 4–7 appointments lived further median distance from Children's of Alabama than those with 8 or more appointments (42.1 miles vs. 24.4 miles, p = 0.0167, Wilcoxon). Even with an outlier of 154.7 miles removed from the group having 4–7 appointments, this difference persists (p = 0.0298). No other SDOH metrics demonstrated a relationship with the number of neurology appointments.

Discussion

Here we present a retrospective evaluation for differences in diagnostic delays, disease characteristics, outcome measures, and healthcare utilization in POMS based on self-reported race/ethnicity and social determinants of health metrics that are neighborhood-based or individual-based. To our knowledge, this is the first significant work to evaluate SDOH and healthcare utilization in POMS.

EDSS was found to be significantly higher at presentation in Black/AA vs. White patients, which likely contributed to the finding that Black/AA patients were more likely to be started on higher efficacy DMT than White patients. Higher efficacy treatment may contribute towards the greater decrease in EDSS that was observed in Black/AA patients, resulting in equivalent EDSS based on race at 2 year and most recent follow-up. These findings suggest that worsening disability accumulation over time in Black/AA patients that has been inconsistently found in AOMS studies could be mitigated by higher efficacy treatment and that further work is warranted in this area in POMS. Utilization of high efficacy treatment at diagnosis, rather than an escalation approach, is thought to reduce long-term disability and has become increasingly favored over time in POMS.^23,24 This approach is supported by our results. A more severe presentation at onset could suggest a genetic basis and/or could suggest a delay to reach diagnosis due to systemic racism or SDOH. As race is a social construct and there is significant genetic admixture in those self-reporting Black/AA race, further work is needed to determine the biological relationship of ancestry with clinical outcomes. One study has found that African ancestry by HLA may correlate with higher disability by the Multiple Sclerosis Severity Score. ²⁵ No differences were found based on SDOH factors in relation to time to reach diagnosis in our small study. The necessity of dissecting diagnostic delays is particularly relevant in children, where multiple sclerosis is less likely to be considered by non-specialists given its rarity.

Implicit bias also plays a negative role for Black/AA patients in reaching a diagnosis in a disease that was historically thought of as a “Caucasian disease”, with more recent prevalence data debunking this belief. ⁴ With this small sample of patients, median time from symptom onset to diagnosis was evaluated with no significant differences identified based on race. However, the mean time to diagnosis revealed potential differences that warrant further exploration in larger cohorts (90 days in Caucasian, 166 days in Black/AA, and 387 days in the 3 Hispanic patients). When attempting to identify barriers towards obtaining timely diagnoses, further consideration of these outliers is critical. The two patients with a significantly prolonged diagnostic delay (at least twice as long as all other patients) were both of minority racial/ethnic groups. Of note, the Hispanic male who took 904 days to receive his diagnosis was the only non-English speaking patient in our cohort, suggesting that language barriers may contribute to diagnostic delays. This patient had private insurance. These findings align with identified diagnostic delays in Hispanic/Latinx AOMS patients as compared to White AOMS patients. ²⁶ Contributory themes for our two patients gleaned from chart review included failed subspecialty referrals and mistrust of the medical system.

Black/AA children in our study were at risk for health disparities based on socioeconomic deprivation derived from one's home address as well as being more likely to have public insurance. Differences in 2-year and long-term clinical outcomes were not identified based on race or most SDOH measures. This suggests that geomarker derived risks may have been mitigated by appropriate clinical care and treatment, though our study is not designed to confirm these possibilities.

Healthcare access is a consideration of primary importance in optimizing outcomes in POMS. Barriers may include suboptimal insurance coverage, financial considerations with expensive DMTs, transportation, and distance from POMS subspecialty care. Importantly, further distance from Children's of Alabama correlated with both decreased number of neurology appointments as well as worsening EDSS over time. This further emphasizes the need for more pediatric MS specialists to promote the best quality of care in geographically underserved areas where general child neurologists are managing POMS.

Additional layers of complexity applying to the POMS as compared to AOMS population result in barriers in progressing this work. One factor is the relative rarity of the disease. Limited pediatric series suggesting differential racial disease characteristics and outcomes do not have the power to interrogate socioeconomic factors that intimately link with race. ⁷ In children, individual SDOH factors such as quality of schooling, are intimately linked to parental factors such as parental education/health literacy, employment, and economic stability. Children with MS are also diagnosed at a time where brain maturation and myelination is ongoing and is decreased in POMS patients versus healthy controls. ²⁷ The impact of adverse childhood experiences (ACEs) on the maturating POMS brain needs to be assessed. ACEs are traumatic events (such as abuse, neglect, household dysfunction, discrimination, financial insecurity) in childhood that are more prevalent in historically marginalized populations. ACEs trigger chronic stress and associate with decreased orbitofrontal cortex volumes and increase the risk/severity of chronic disease, to possibly include MS.^28–31

This study has a number of limitations. First and foremost, our sample size is small but is consistent with recruitment from a single specialized pediatric center (one of 11 centers in the United States that constitute the Pediatric Network of MS Centers). This results in lower power in pursuing covariate analysis to delineate what differences may be more attributable to race vs. socioeconomic deprivation. Specifically, there were only 2 White patients with public insurance, limiting association of insurance status with healthcare utilization or SDOH factors. Similarly, deeper associations with type of DMT could also not be investigated. Ancestry is also not able to be investigated in this retrospective work. Neighborhood-based indices derived from the US Census give insight into socioeconomically deprived areas that are at risk for health disparities, but do not provide information on an individual, household level. ¹⁰ Our population also had few patients of Hispanic ethnicity due to small sample size and thus were not amenable to statistical analysis. For similar reasons, language as a SDOH could not be analyzed.

Given the lack of prior SDOH work in POMS, our study provides value in identifying areas of focus in a framework for future research. Further prospective, multi-institutional work is needed to dissect the interplay of genetics, social determinants of health, and racism on health outcomes in POMS so that interventions to optimize care in this population can be developed.