Association of Area Deprivation Index With Severity of COVID-19 in Pediatric Patients: A Multisite Study of Racially and Ethnically Diverse Children

Abstract

Area deprivation index (ADI) is associated with the risk of severe COVID-19 in adults. However, this association has not been established in children. Information on ADI, demographics, clinical features, disease severity, and outcomes was analyzed for 3434 children with COVID-19. A multivariate logistic regression revealed that non-Hispanic Asians, extremes of weight, and higher ADI were associated with severe disease.

Socioeconomic factors drive a host of health-related outcomes and socioeconomic deprivation can set children on a trajectory of health inequities if experienced early in life.¹ The area deprivation index (ADI) is a validated metric that ranks neighborhoods by socioeconomic disadvantage.² The ADI is a factor-based composite measure of socioeconomic deprivation constructed from 17 variables in the domains of income, education, employment, and housing quality from the American Community Survey 5-year estimates aggregated to US Census block groups, proxies for neighborhoods. Recent studies in adults have reported neighborhood-level deprivation as a predictor of coronavirus disease 2019 (COVID-19) risk and outcomes.^3,4

Children represent approximately 17% of all reported cases of COVID-19.⁵ and a fraction develops severe disease, including multisystem inflammatory syndrome in children (MIS-C).⁶ Previous studies have highlighted the role of demographic and physical characteristics such as age, race, ethnicity and body mass index for COVID-19 disease severity and outcomes in children.^3,7,8 However, studies that adjusted for the deprivation level found comparable COVID-19 severity across various races and suggested socioeconomic status as a mediator of the association between race and hospitalization.^3,7 For example, Savorgnan et al in a study of a large cohort of children in Texas, found an increase in odds of severe MIS-C by 21% for each decile increase in ADI.⁹ However, this study was limited due to being conducted in a single state and included only those COVID-19 patients with a diagnosis of MIS-C. Hence the study results are not generalizable to other populations and the entire range of severe COVID-19 infections.

Our study objectives were to determine the relationship of ADI with severity of COVID-19 in a large cohort of children evaluated in geographically distinct areas of the USA and to identify social and demographic risk factors correlated with severe illness.

METHODS

Study Design and Sites

This was a prospective study of a sample of children who were evaluated for symptoms related to SARS-CoV-2 infection between March 29, 2021, and December 31, 2021, at 3 sites (A. Children’s Hospital of Michigan, B. UPMC Children’s Hospital of Pittsburgh, and C. Texas Children’s Hospital). Sites A and B are tertiary pediatric centers located in the Midwest and Northeast regions of the United States, respectively, while site C is a quaternary pediatric hospital located in the Southwest region of the United States. The study included children of age <18 years with SARS-CoV-2 infection.

Study Definitions and Measures

SARS-CoV-2 infection: Defined by a positive reverse transcription polymerase chain reaction (RT-PCR) test, serology or antigen test. The patients were categorized as having “severe” disease if they required noninvasive respiratory support including high flow oxygen >6 L, bilevel positive airway pressure (BiPAP), continuous positive airway pressure, or mechanical ventilation, and/or inotropic/vasoactive support, extracorporeal membrane oxygenation (ECMO) or if the infection resulted in death. Otherwise, the patients were categorized as having “mild” disease. The outcome variable is disease severity, and the explanatory variables are the ADI score which ranges from 1 to 100 at the national level, with higher values indicating higher deprivation, sex, age, race/ethnicity and weight status.

Study Data

Information on demographic data, final diagnosis and treatment given were entered into a database by trained research assistants. Clinical data were obtained from the electronic medical records of the participating sites. The reference distributions for race/ethnicity, sex, age and weight were obtained from the CDC database as the population-adjusted averages of Texas, Michigan and Pennsylvania.¹⁰ Using home addresses, the 9-digit postal codes were obtained through a secure application programming interface with the US Post Office. Subsequently, the 9-digit postal codes were linked to ADI scores using the 2020 version of the Neighborhood Atlas database.² The postal codes and all personal identifiers were then removed from the working database to avoid sharing PHI data. In the combined multisite database, subjects were fully de-identified. The study was reviewed and approved by the Institutional Review Boards of all sites.

Statistical Analysis

Continuous variables were discretized by their median and interquartile range (IQR). Categorical variables were mapped to frequencies and proportions and compared by the Fisher exact test and the Binomial test. The association between multiple covariates and severity of disease was established using a multivariate logistic regression. All statistical analyses were conducted using the statsmodels v0.13.2 package in the Python 3.8.3 environment. A P value of <0.05 was considered statistically significant and all tests were 2-sided.

RESULTS

Data on a total of 3434 children were included in the final analysis. The median age of the cohort was 8.6 years (IQR: 2.0–14.4 years), 52% were recorded as male. The median ADI was 64 (IQR: 43–81). Of the total, two-thirds of the cohort (67%) had an ADI classification > US national median. See Figure 1 for a graphical comparison of the ADI and weight distributions of the cohort and the US general population. The demographics and ADI classification of this cohort as compared with the reference population of children are provided in Table 1. Our sample was significantly different from the reference 3-state population in age, race, weight and ADI distributions. There was a larger proportion of younger children (<6 years old), a smaller proportion of children with healthy weight, and a larger proportion of children lived in areas with a higher ADI score in our cohort compared with the reference population.

TABLE 1. - Comparison Between the Entire COVID-19 Pediatric Cohort and Reference Population in the United States, and Also Between “Mild” and “Severe” Subgroups

Variables	Entire Cohort	Reference, %	Binomial Test, P	Mild	Severe	Fisher’s Exact Test, P
Total	3434 (100%)	n/a	n/a	2926 (85%)	508 (15%)	n/a
Sex
Female	1664 (48%)	49	0.573	1434 (49%)	230 (45%)	0.124
Male	1770 (52%)	51	0.573	1492 (51%)	278 (55%)	0.124
Age
<6 y	1411 (41%)	31	<0.001	1219 (42%)	192 (38%)	0.107
6–12 y	764 (22%)	32	<0.001	669 (23%)	95 (19%)	0.038
>12 y	1259 (37%)	37	0.448	1038 (35%)	221 (44%)	<0.001
Weight
Underweight	323 (9%)	5	<0.001	243 (8%)	80 (16%)	<0.001
Healthy weight	1878 (55%)	80	<0.001	1662 (57%)	216 (43%)	<0.001
Overweight	407 (12%)	10	<0.001	363 (12%)	44 (9%)	0.017
Obese	826 (24%)	5	<0.001	658 (22%)	168 (33%)	<0.001
ADI
Low (≤ 50th)	1144 (33%)	50	<0.001	995 (34%)	149 (29%)	0.041
High (> 50th)	2290 (67%)	50	<0.001	1931 (66%)	359 (71%)	0.041
Race/ethnicity
Hispanic	1572 (46%)	35	<0.001	1354 (46%)	218 (43%)	0.162
Non-Hispanic White	823 (24%)	46	<0.001	706 (24%)	117 (23%)	0.613
Non-Hispanic Black	817 (24%)	13	<0.001	682 (23%)	135 (27%)	0.114
Non-Hispanic Asian	97 (3%)	5	<0.001	77 (3%)	20 (4%)	0.110
Other/Unknown	125 (4%)	n/a	n/a	107 (4%)	18 (4%)	1.000

The reference in race/ethnicity, sex, age and weight distributions were obtained from the population-adjusted average of the study sites’ states (Texas, Michigan and Pennsylvania).

n/a indicates not applicable.

FIGURE 1.: Comparison between the COVID19 pediatric cohort and the general population. Histograms of the cohort ADI (A) and age- and sex-adjusted weight (B) compared with the US national distributions as reference.

Five hundred eight children (15%) had severe disease. Based on univariable analysis, we found that children of age >12 years (P < 0.001), with underweight and obese status (P < 0.001), or high ADI score (P = 0.041) had a higher proportion of severe cases. A multivariable logistic regression revealed that non-Hispanic Asians in reference to non-Hispanic White (odds ratio [OR] = 1.77, P = 0.039), both underweight (OR = 2.51, P < 0.001) and obese (OR = 1.84, P < 0.001) in reference to healthy weight, ages >12 years in reference to 6–12 years (OR = 1.45, P = 0.006) and higher ADI (OR = 1.06 per each 10-point increase, P = 0.008) were independently associated with severe disease while also controlling for child sex. See details in Table 2.

TABLE 2. - Multivariable Logistic Regression of COVID-19 Severity (N = 3434)

Covariables	Coef.	OR (95% CI)	P
Intercept	−2.65	n/a	<0.001
Sex
Male (ref: female)	0.18	1.20 (0.99–1.45)	0.068
Race/ethnicity
Hispanic (ref: Non-Hispanic White)	−0.13	0.88 (0.68–1.13)	0.300
Non-Hispanic Black (ref: Non-Hispanic White)	0.09	1.09 (0.82–1.44)	0.542
Non-Hispanic Asian (ref: Non-Hispanic White)	0.57	1.77 (1.03–3.03)	0.039
Other/unknown (ref: Non-Hispanic White)	−0.04	0.96 (0.56–1.66)	0.890
Age
0–6 y (ref: 6–12 y)	0.14	1.15 (0.88–1.51)	0.311
12–18 y (ref: 6–12 y)	0.37	1.45 (1.12–1.89)	0.006
Weight
Underweight (ref: healthy weight)	0.92	2.51 (1.87–3.35)	<0.001
Overweight (ref: healthy weight)	−0.09	0.91 (0.64–1.29)	0.603
Obese (ref: healthy weight)	0.61	1.84 (1.46–2.33)	<0.001
ADI
Ten-point increase in ADI score	0.06	1.06 (1.01–1.11)	0.008

CI indicates confidence interval; OR, odds ratio.

DISCUSSION

Our study highlights the impact of race and ethnicity, body habitus and social economic factors on severity of COVID-19 in children living in the Southwest, Midwest and Northeast regions of the United States. Overall, the study cohort of COVID-19 pediatric patients was found to be heavily skewed toward high disadvantage. Moreover, Non-Hispanic Asians, underweight and obese children and those living in areas with even higher deprivation index were significantly more likely to have severe disease.

Our findings are aligned with other related studies confirming that social constructs (socioeconomic status and race/ethnicity) have an impact on prevalence, hospitalization and death rates in COVID-19.^3,4,8 In fact, our cohort contains a larger proportion of Hispanic and non-Hispanic Black subjects, smaller proportion of subjects with healthy weight, and higher ADI scores compared with the reference population (Table 1). Disease severity, ADI, race/ethnicity and weight status are mutually associated among the subjects in our cohort. Hence, to quantify the effect of ADI on disease severity, it is important to control for race/ethnicity and weight as covariables in the logistic regression analysis. Savorgnan et al studied 206 children with MIS-C in Texas and determined that Non-Hispanic Black race and high ADI to be associated with severe MIS-C.⁹ While our study had similar results about the influence of socioeconomic factors on severe disease, we found only non-Hispanic Asian race (in reference to non-Hispanic White) to be associated with severe disease. One of the reasons for this noted discrepancy could be the small number of non-Hispanic Asians in our cohort which may have impacted our results.

It is well known that overweight and obese status is a significant risk factor for the incidence and severity of COVID-19 in adult patients. See Wu et al¹¹ and references therein. However, an additional interesting finding in our study is the association of underweight status with severe COVID-19 in pediatric patients. Previous studies in adults have reported this association.^11,12 Stridsman et al in their study of a cohort of adults with chronic obstructive pulmonary disease from a Swedish registry found underweight status among others to be a predictor of severe COVID-19.¹² Wu et al found that in adults, underweight was associated with a 2.85-fold higher risk of death secondary to COVID-19 after adjusting for potential confounders.¹¹ Although the exact reasons for this association are unclear, we propose that poor nutritional status could result in impaired immune response thus making the host more vulnerable to severe disease.

Our study was limited to 3 large health centers with their satellites across 3 states only and hence may not be generalizable. While ADI is an acceptable indicator of social status, it does not provide granular information for each individual/family regarding social risks and determinants of health. Hence, it is possible that there are several other family-level factors that may exist which could not be captured by the ADI. Another limitation due to the nature of this retrospective study is the lack of a complete record of comorbidities and risk factors for the cohort of patients.

CONCLUSIONS

Race/ethnicity, underweight and obesity and high ADI levels are predictors of severe disease in pediatric COVID-19. Further large-scale studies are required to confirm our findings so that populations and social determinants of health can be targeted for appropriate interventions.