Area deprivation and pediatric cancer mortality: An analysis of 2000–2020 state cancer registry data

Abstract

Background:

Compared to adult cancers, less research has been conducted on health disparities in pediatric cancer outcomes. The effect of neighborhood deprivation, as measured by the Area Deprivation Index (ADI), on pediatric cancer survival is not well understood, especially among diverse populations.

Methods:

We conducted a population-based longitudinal study using data from the Iowa Cancer Registry and Louisiana Tumor Registry. The study included children diagnosed with cancer in the two states ages 0–19 years from 2000 to 2020. The primary exposure was ADI, and the primary outcome was cancer-specific mortality. Cox regression models with shared frailty to account for the geographic clustering of individuals were used to compute the association between ADI and mortality.

Results:

A total of 6,982 children were included between the two states: 2,939 in Iowa and 4,043 in Louisiana. In the adjusted analysis, higher deprivation was positively associated with poorer pediatric cancer survival. Compared to children in the least deprived neighborhoods, those in the most deprived neighborhoods had a 50% higher risk of cancer death (95% CI: 1.18–1.89) and a 3.15 times greater risk of early cancer death (95% CI: 1.39–7.16). Among children with extracranial solid tumors, those in the most deprived neighborhoods had a 54% higher risk of cancer death (95% CI: 1.04–2.27) compared to those in the least deprived neighborhoods.

Conclusions:

Higher ADI is associated with higher pediatric cancer mortality, highlighting the importance of social drivers of health on pediatric cancer outcomes.

Impact:

Neighborhood-level interventions will likely be needed to improve pediatric cancer outcomes.

Introduction

An estimated 14,910 children (aged 0–19 years) in the US were diagnosed with cancer in 2024 (1). The five-year survival for children with cancer has increased from 63% in the 1970s to over 85% in 2024 in large part due to improvements in treatment (2). However, differences in survival still persist by cancer type, sex, race and ethnicity, geographic location, and socioeconomic status (3, 4). There is also substantial financial burden associated with pediatric cancer treatment for children and their families (5). The average cost (including both direct and indirect costs) associated with childhood cancer is estimated to be $833,000 per family, with financial hardship often lasting more than one year after treatment ends (6, 7). Additionally, treatment for pediatric cancer generally occurs in specialized tertiary care centers, typically found in large urban areas, resulting in a significant travel burden for pediatric cancer patients and their families, particularly in rural states (8).

Compared to adult cancers, less research has been conducted on health disparities in pediatric cancer outcomes (3). In particular, gaps have been identified in the understanding of underlying causes of disparities in overall survival and delays in pediatric cancer treatment (3). A 2023 systematic review of studies on pediatric cancer survivorship care identified disparities by race, ethnicity, health insurance, and geographic area (9). Low socioeconomic status may impact outcomes due to opportunities to get referrals and second opinions, later stage at diagnosis, and poor treatment adherence (10, 11).

Neighborhood deprivation, also known as area deprivation, is a measure of socioeconomic disadvantage estimated at the neighborhood (i.e., census block group) level that is associated with various health outcomes even independent of individual socioeconomic status (12). Higher levels of neighborhood deprivation are linked to higher allostatic load (the cumulative burden of chronic stress) (13, 14), epigenetic age acceleration (15), and lower health-related quality of life (16). The Area Deprivation Index (ADI), a composite measure of 17 indicators related to income, education, employment, and housing based on American Community Survey (ACS) data, is a validated measure of neighborhood deprivation (17). Originally developed by the US Health Resources and Services Administration, ADI are updated, maintained, and freely available from the Neighborhood Atlas, making ADI an accessible measure of neighborhood deprivation for health research (12).

To date, few studies have examined the association between ADI and pediatric cancer survival in the United States (18–21). For example, higher levels of area deprivation have been associated with poor pediatric cancer survival in analyses using Texas Cancer Registry data (18, 19). Only one study has investigated more than one pediatric cancer type, and to our knowledge, no studies have used a multilevel modeling approach to account for correlations between geographical clusters of individuals within each study (20). Compared to previous studies on ADI and pediatric cancer survival, the populations of Iowa and Louisiana have a greater proportion of rural residents (36.0% and 26.8%, respectively) and limited availability of Children’s Oncology Group sites (Figure 1) (22), which may lead to a greater travel burden in these states. Additionally, compared to the overall US, Louisiana has a higher proportion of racial minorities (33% Black vs. 13.6% Black) and a greater proportion of people living in poverty (19.6% vs. 11.6%) (23). Further, both states have significant variability in levels of neighborhood deprivation (12). Thus, the objective of this study was to examine the association between ADI and pediatric cancer mortality using state cancer registry data for Iowa and Louisiana.

Figure 1.

Area Deprivation Index Map. This map displays the Area Deprivation Index by state with locations of Children’s Oncology Group centers

Materials and Methods

Study Population

This study is a population-based longitudinal study using data from the Iowa Cancer Registry and the Louisiana Tumor Registry. The Iowa Cancer Registry has been a part of the Surveillance, Epidemiology, and End Results Program (SEER) since 1973 and the Louisiana Tumor Registry has been a part of the SEER program since 2001 (24). The study was approved by the Institutional Review Boards at the University of Nebraska Medical Center, the University of Iowa, and Louisiana State University Health Sciences Center. Study participants included individuals with a first primary diagnosis of cancer in Iowa and Louisiana between 2000 and 2020. Individuals with non-malignant tumors or tumors in situ were excluded. Additionally, individuals diagnosed at autopsy only (i.e., death certificate only cases) were also excluded.

Variables

Survival time.

Vital status through December 31, 2020 was obtained from each registry and survival time (time from diagnosis to death or last follow-up/date of administrative censoring) was calculated in months. The primary outcome of interest was cancer-specific mortality (cases in which the cancer was considered the primary cause of death). Cause of death is ascertained through annual linkage to the National Death Index, as well as local death clearance, which includes death certificate review. Individuals that died from other causes were censored at the time of death.

ADI.

The primary exposure of interest was ADI. The two cancer registries geocoded patients’ residential addresses at the time of diagnosis at the census block group level. Data on ADI national rank was abstracted from the Neighborhood Atlas, Version 3, for corresponding census block groups (25). ADI in the Atlas is constructed from a five-year average of ACS data using previously developed methods (26, 27). For cases diagnosed from 2000 through 2010, ACS data from 2005–2009 was used; for cases diagnosed from 2011 through 2020, ACS data from 2013–2017 was used. ADI rank quartiles were treated as a categorical variable in the analyses (Q1: 1–25, Q2: 26–50, Q3: 51–75, Q4: 76–100).

Other variables.

Rurality at the census tract level was measured using 2010 Rural-Urban Commuting Area (RUCA) codes defined by the U.S. Department of Agriculture at the Census tract level as urban (RUCA codes 1–3) or rural (RUCA codes 4–10) (28). Race and ethnicity were recategorized as non-Hispanic White, non-Hispanic Black, non-Hispanic American Indian/Alaskan Native, non-Hispanic Asian or Pacific Islander, and Hispanic of any race. Cancer type was classified according to the International Classification of Childhood Cancer, Third edition (ICCC-3) codes. The cancer types were grouped as hematologic malignancies, CNS tumors, and other extracranial solid tumors. Additional variables considered were sex, age at primary diagnosis, and year of diagnosis.

Statistical Analysis

All analyses were conducted in R (version 4.4.0). To handle missing data on race/ethnicity (0.77%), rurality (0.13%), survival time (0.29%), ADI (10.31%), and cause of death (0.44%), multiple imputation was performed using the mice package in R. Kaplan-Meier curves were constructed to compare the survival functions by ADI quartiles, with a log-rank test used to test for differences in these survival functions. Descriptive statistics were computed and group differences tested using chi-square test and t-tests as appropriate. Per cancer registry rules, to preserve privacy, cell counts of five or less were not reported to preserve patient confidentiality. A directed acyclic graph (DAG) was used to identify which variables to include in the adjusted analysis to control for confounding (Supplemental Figure 1) (29). Multilevel Cox regression models with shared frailty (to account for the clustering of cases within states) were used to assess the relationship between ADI quartiles and cancer-specific mortality with random effects for state. Hazard ratios (HRs) and 95% confidence intervals (CIs) for these relationships were estimated. To examine if the effect of ADI on mortality was modified by race/ethnicity, rurality, and state, interaction terms were included in the multivariable models. Given the known differences in survival by cancer type (30), the decision was made a priori to stratify the models by the three main cancer groups. In stratified analyses, models were not converging due to the small numbers by race/ethnicity. To resolve these convergence issues, in the stratified analyses, a simplified race/ethnicity variable was used, combining non-Hispanic Asian/Pacific Islander, non-Hispanic American Indian/Alaskan Natives, and Hispanics into an “Other” category. The proportional hazards assumption was assessed with a goodness-of-fit test (31) and graphically with plots of the Schoenfeld residuals (32). If the assumption was not met, extended Cox regression model would be used to allow for the time-dependent covariate.

In a secondary analysis, we explored the effect of ADI on early mortality, defined as death within three months of diagnosis. In this analysis, all cases that survived longer than three months were artificially censored at the three-month time point and Cox regression models with shared frailty were used to assess the relationship between ADI and cancer-specific early mortality.

Sensitivity analyses.

Because this is a secondary analysis of cancer registry data, there is a limited number of potential confounder variables available in the dataset. To quantify the potential effect of unmeasured confounders on the association between ADI and mortality, E-values were calculated for the survival effect estimates (33, 34). The E-value quantifies the minimum strength of association an unmeasured confounder would need to have with both the exposure and outcome to negate the observed association (33).

Data availability

The data analyzed in this study were obtained from the Louisiana Tumor Registry and Iowa Cancer Registry and are available upon request from the cancer registries.

Results

A total of 6,982 children with primary cancer were included in the study from the original 7,085 children identified from the two states (Supplemental Figure 2). After combining the data for the two states, 89 tumors in situ were excluded. After exclusions, 2,939 were diagnosed in Iowa and 4,043 were diagnosed in Louisiana (Table 1). In both states, the median age at diagnosis was 10 years with an interquartile range of 3.50–16.50 years. The median year of diagnosis was 2011, with an interquartile range of 2006–2016. Of the cases diagnosed in Iowa, 85.40% were non-Hispanic White. There was more racial/ethnic diversity among the cases in Louisiana; 61.32% of cases were non-Hispanic White and 31.59% of cases were non-Hispanic Black. The distribution of cancer types was similar between the two states. Extracranial solid tumors were the most common, followed by hematologic malignancies and CNS tumors. In Iowa, the proportion of cases diagnosed in rural census tracts was high at 41.03%, compared to 15.71% of cases in Louisiana. The distribution of ADI quartiles differed between the two states (p<0.0001), although the most common quartile in both states was Q3.

Table 1.

Demographic characteristics of included study participants, children aged 0–19 diagnosed with cancer in Iowa and Louisiana from 2000–2020, n=6,982

	Total, N=6,982	Iowa, N= 2,939	Louisiana, N= 4,043	p-value
Age at diagnosis, Median (IQR)	10 (3.50–16.50)	10 (3.50–16.50)	10 (3.50–16.50)	0.67
Year of diagnosis, Median (IQR)	2011 (2006–2016)	2011 (2006–2016)	2011 (2006–2016)	0.50
Sex				0.47
Male	3,760 (53.85%)	1,598 (54.37%)	2,162 (53.48%)
Female	3,222 (46.15%)	1,341 (45.63%)	1,881 (46.52%)
Race/ethnicity				<0.0001
NH White	4,989 (71.46%)	2,510 (85.40%)	2,479 (61.32%)
NH Black	1,413 (20.24%)	136 (4.63%)	1,277 (31.59%)
NH AI/AN	16 (0.23%)	11 (0.37%)	5 (0.12%)
NH Asian/Pacific Islander	120 (1.72%)	48 (1.63%)	72 (1.78%)
Hispanic (any race)	444 (6.36%)	234 (7.96%)	210 (5.19%)
Cancer groups				0.08
Extracranial solid tumors	2,932 (41.99%)	1,274 (43.34%)	1,658 (41.01%)
Hematologic malignancies	2,778 (39.79%)	1,125 (38.28%)	1,653 (40.89%)
CNS tumors	1,272 (18.22%)	540 (18.37%)	732 (18.11%)
Rurality				<0.0001
Rural	1,841 (26.37%)	1,206 (41.03%)	635 (15.71%)
Urban	5,141 (73.63%)	1,733 (58.97%)	3,408 (84.29%)
Area Deprivation Index				<0.0001
Q1 (least deprived)	804 (11.52%)	119 (4.05%)	685 (16.94%)
Q2	1,737 (24.88%)	812 (27.63%)	925 (22.88%)
Q3	2,689 (38.51%)	1,188 (40.42%)	1,501 (37.13%)
Q4 (most deprived)	1,752 (25.09%)	820 (27.90%)	932 (23.05%)

Abbreviations: IQR, interquartile range; NH, non-Hispanic; AI/AN: American Indian/Alaskan Native; CNS, central nervous system

The median follow-up by ADI quartile ranged from 80.41 to 98.00 months (92.66 months overall), and did not statistically significantly differ by ADI (p=0.10) (Supplemental Table 1). Figure 2 shows the Kaplan-Meier curve of cancer-specific survival probabilities over time by ADI quartile. We observed a significant difference in survival by ADI (p<0.0001). Over the study period, survival was lowest in the most deprived quartile (Q4) and highest in the least deprived quartile (Q1). Over time, the survival probabilities for Q2 and Q3 were similar. At the median follow-up point, we also observed a significant difference in survival probability by ADI quartile (p=0.04).

Figure 2.

Survival Probability by ADI. Kaplan-Meier curve showing differences in survival probability over time by Area Deprivation Index (ADI) quartiles among children aged 0–19 diagnosed with cancer in Iowa and Louisiana from 2000–2020, n=6,982

Table 2 shows the unadjusted and adjusted hazard ratios for the association between ADI and mortality. In the unadjusted model, compared to those in the least deprived quartile the hazard of death was significantly higher for those in the two most deprived quartiles (HR=1.28, 95% CI: 1.03–1.60 and HR=1.63, 95% CI: 1.30–2.03), respectively. Based on our DAG, the full model adjusted for state, rurality, sex, race/ethnicity, cancer type, age at diagnosis, and year of diagnosis (the minimal sufficient adjustment set and other important predictors of survival). In the adjusted model, compared to those in the least deprived quartile, those in the most deprived quartile had a 50% increase in the hazard of death (95% CI: 1.18–1.89). There was no association between ADI and mortality for Q2 and Q3 compared to those in the least deprived neighborhoods (aHR=1.23, 95% CI: 0.97–1.56 and aHR=1.21, 95% CI: 0.87–1.51, respectively). In the adjusted analysis, in addition to ADI, sex, race/ethnicity, cancer type, and year of diagnosis were all associated with mortality (p<0.05).

Table 2.

Unadjusted and adjusted association between ADI and other covariates and cancer mortality among children aged 0–19 diagnosed with cancer in Iowa and Louisiana from 2000–2020, n=6,982

	Unadjusted		Adjusteda
	HR (95% CI)	p-value	aHR (95% CI)	p-value
Area Deprivation Index
Q1 (least deprived)	Reference		Reference
Q2	1.23 (0.98–1.56)	0.08	1.23 (0.97–1.56)	0.08
Q3	1.28 (1.03–1.60)	0.02	1.21 (0.97–1.51)	0.10
Q4 (most deprived)	1.63 (1.30–2.03)	<0.0001	1.50 (1.18–1.89)	0.0008
Rurality
Urban	Reference		Reference
Rural	1.01 (0.89–1.14)	0.91	0.99 (0.86–1.14)	0.92
Sex
Male	Reference		Reference
Female	0.79 (0.70–0.89)	<0.0001	0.78 (0.69–0.88)	<0.0001
Race/ethnicity
NH White	Reference		Reference
NH Black	1.72 (1.51–1.96)	<0.0001	1.67 (1.45–1.90)	<0.0001
NH AI/AN	1.96 (0.81–4.73)	0.14	1.92 (0.79–4.63)	0.15
NH Asian/Pacific Islander	1.53 (1.02–2.28)	0.04	1.66 (1.11–2.48)	0.01
Hispanic (any race)	0.92 (0.70–1.20)	0.53	0.96 (0.73–1.26)	0.78
Cancer groups
Extracranial solid tumors	Reference		Reference
Hematologic malignancies	0.73 (0.63–0.84)	<0.0001	0.72 (0.63–0.83)	<0.0001
CNS tumors	2.01 (1.76–2.30)	<0.0001	2.01 (1.76–2.31)	<0.0001
Age at diagnosis	1.00 (0.99–1.00)	0.32	1.00 (0.99–1.01)	0.81
Year of diagnosis	0.98 (0.97–0.99)	<0.0001	0.98 (0.97–0.99)	0.0002

^aAdjusted for all variables in the table

Abbreviations: NH, non-Hispanic; AI/AN: American Indian/Alaskan Native; CNS, central nervous system

Table 3 shows the results for the association between ADI and mortality stratified by cancer type. In the stratified analysis, the association between ADI and mortality was not statistically significant for children with hematologic malignancies (Q4 vs. Q1: aHR=1.47, 95% CI: 0.97–2.22) and CNS tumors (Q4 vs. Q1: aHR=1.37, 95% CI: 0.91–2.06). Among children with extracranial solid tumors, compared to those in the least deprived neighborhoods, those in the most deprived neighborhoods had a higher hazard of death (aHR=1.51, 95% CI: 1.02–2.23). There was no effect modification by race/ethnicity, rurality, and state on the association between ADI and mortality (p>0.05), and there were no significant differences in the stratified results (Supplemental Tables 2–4).

Table 3.

Association between ADI and mortality stratified by cancer type among children aged 0–19 diagnosed with cancer in Iowa and Louisiana from 2000–2020

	aHR (95% CI) a	p-value
Hematologic malignancies, n=2,778
ADI
Q1 (least deprived)	Reference
Q2	1.12 (0.73–1.72)	0.60
Q3	1.05 (0.70–1.58)	0.81
Q4 (most deprived)	1.47 (0.97–2.22)	0.07
CNS tumors, n=1,272
ADI
Q1 (least deprived)	Reference
Q2	1.19 (0.80–1.76)	0.40
Q3	1.13 (0.77–1.65)	0.53
Q4 (most deprived)	1.37 (0.91–2.06)	0.14
Extracranial solid tumors, n=2,932
ADI
Q1 (least deprived)	Reference
Q2	1.36 (0.92–2.02)	0.12
Q3	1.40 (0.97–2.03)	0.08
Q4 (most deprived)	1.51 (1.02–2.23)	0.04

^aAdjusted for: state, rurality, sex, race/ethnicity, age at diagnosis, and year of diagnosis

Table 4 displays the unadjusted hazard ratios for the association between ADI and early mortality (death within three months of diagnosis). In the unadjusted analysis, compared to those in the least deprived neighborhoods, the hazard of death within three months of diagnosis was 3.78 times higher for those in the most deprived neighborhoods (95% CI: 1.72–8.30). Similarly, in the adjusted analysis, the hazard of death within three months of diagnosis was higher among those in the most deprived neighborhoods compared to those in the least deprived neighborhoods (aHR=3.15, 95% CI: 1.39–7.16). In addition to ADI, race/ethnicity, cancer type, and age at diagnosis were statistically significantly associated with mortality. After stratifying by cancer type group, the association between ADI and early mortality was not statistically significant (Supplemental Table 5).

Table 4.

Unadjusted and adjusted association between ADI and other covariates and cancer early mortality among children aged 0–19 diagnosed with cancer in Iowa and Louisiana from 2000–2020, n=6,982

	Unadjusted		Adjusteda
	HR (95% CI)	p-value	aHR (95% CI)	p-value
Area Deprivation Index
Q1 (least deprived)	Reference		Reference
Q2	1.86 (0.81–4.27)	0.14	1.81 (0.79–4.15)	0.16
Q3	1.96 (0.88–4.36)	0.10	1.78 (0.79–3.99)	0.17
Q4 (most deprived)	3.78 (1.72–8.30)	0.001	3.15 (1.39–7.16)	0.01
Rurality
Urban	Reference		Reference
Rural	1.34 (0.94–1.92)	0.11	1.21 (0.81–1.79)	0.36
Sex
Male	Reference		Reference
Female	0.71 (0.50–1.01)	0.06	0.73 (0.51–1.03)	0.08
Race/ethnicity
NH White	Reference		Reference
NH Black	1.79 (1.23–2.60)	0.003	1.63 (1.09–2.43)	0.02
NH AI/AN	b		b
NH Asian/Pacific Islander	3.05 (1.33–7.00)	0.009	3.45 (1.50–7.97)	0.004
Hispanic (any race)	0.96 (0.45–2.08)	0.92	0.84 (0.38–1.83)	0.66
Cancer groups
Extracranial solid tumors	Reference		Reference
Hematologic malignancies	1.48 (0.98–2.23)	0.07	1.42 (0.94–2.15)	0.10
CNS tumors	2.71 (1.76–4.16)	<0.0001	2.46 (1.60–3.78)	<0.0001
Age at diagnosis	0.95 (0.93–0.98)	0.0003	0.95 (0.93–0.98)	0.001
Year of diagnosis	0.98 (0.95–1.00)	0.09	0.98 (0.95–1.01)	0.17

^aAdjusted for all variables in the table

^bSample size is too small to calculate

The E-values, which measure the magnitude of the association on the risk-ratio scale that an unmeasured confounder would need to have with both ADI and mortality to “explain-away” the observed association are displayed in Supplemental Table 6.(33, 34) An E-value quantitatively provides an estimate of how robust an estimate is to unmeasured, uncontrolled confounding; the larger an E-value is, more substantial unmeasured confounding would have to exist to negate the observed association. For the models for overall mortality, the E-value for the association comparing the most deprived neighborhoods to the least deprived neighborhoods was 2.37 (1.64 for the 95% CI). For the models estimating early mortality, the E-value for the association comparing the most deprived neighborhoods to the least deprived neighborhoods was 6.03 (2.26 for the 95% CI).

Discussion

In this study, high levels of neighborhood deprivation, as measured by the Area Deprivation Index, were associated with an increased risk of cancer death. Additionally, the association was even stronger for death within three months of diagnosis. Our study is the first to look at the association between ADI and early mortality, or death within three months of diagnosis. This relationship persists even after adjusting for individual-level demographic characteristics, which highlights the importance of social drivers of health in pediatric cancer outcomes.

Our results are in line with other studies that have looked at the association between ADI and pediatric cancer survival (18–21). Higher levels of ADI have been associated with poorer survival of acute lymphoblastic leukemia (18) and CNS tumors (19) in analyses using Texas Cancer Registry data. In a population-based study in Washington, children living in more disadvantaged neighborhoods had higher mortality rates, compared to those living in less disadvantaged areas (20). Additionally, in a hospital registry in Oklahoma, among children with acute lymphoblastic leukemia, higher levels of deprivation were associated with a 1.51–1.85 times higher hazard of mortality compared to children living in the least disadvantaged neighborhoods, although the overall sample size was small (N=275). Further, studies based in hospital systems may selectively include individuals already regularly involved in the healthcare system, which may have more favorable social determinants of health (21, 35). In contrast to the aforementioned studies of ALL and CNS tumors, our stratified results were only statistically significant for extracranial solid tumors. There is a high degree of heterogeneity in pediatric cancer, which can affect patterns of cancer sites, histologic types, age at diagnosis, clinical behaviors, and ultimately the type of treatment received, length of treatment, and treatment adherence (36, 37). The differences observed in our study may be due to sample size limitations when stratifying by cancer type.

Additionally, our results are in line with previous studies of other composite neighborhood-level measures of deprivation. In a study of SEER data, children living in more deprived areas (as measured by the Social Deprivation Index – a composite measure based on seven ACS indicators) had a 1.09 times higher hazard of death for metastatic cancers and 1.27 times higher hazard of death for non-metastatic cancers compared to those living in less deprived areas (38). The Yost Index, another composite measure of seven ACS indicators, was also associated with survival in children with central nervous system tumors in a study of SEER data and early mortality of acute myeloid leukemia in California (39, 40). Composite measures of social risk are able to capture the complex, multidimensional nature of social drivers, while removing the multicollinearity issues that may occur when including all of the individual characteristics in one model (41). ADI has the benefit of being a well-validated measure, allowing for comparison of studies in other populations and age groups (27, 42, 43). Compared to the previously mentioned composite measures, ADI captures a larger number of measures, including measures of housing quality and is calculated at a more granular geographic level (census block-group rather than county), which may account for the differences in effect size observed in our study (41, 44).

In addition to differences in the risk of mortality by neighborhood-level deprivation, we also observed racial/ethnic differences in survival. Consistent with previous studies, Non-Hispanic Black children in our cohort were at a higher risk of cancer death; however, in contrast to these studies, we did not observe that Hispanic children were at a higher risk of cancer mortality (45, 46). Additionally, variations in healthcare access, insurance coverage, and treatment adherence across geographic regions may contribute to differences in survival patterns. For example, in Texas, residence in a Hispanic enclave was associated with inferior survival of ALL (47). These disparities highlight the need for further studies on the complex interactions in race, ethnicity, and individual- and neighborhood-level disadvantage.

There is a growing movement to move beyond just describing disparities to developing solutions and systemic change (48). The mechanisms linking neighborhood deprivation and other individual and area-level measures of social drivers of health are still unclear; however, proposed mechanisms include access to health insurance, clinical trial accessibility, severity of disease at diagnosis, and treatment adherence (10, 11). There is also evidence that disparities in access to survivorship care after pediatric cancer exist by race/ethnicity, insurance access, employment, income, and geographic area, which may contribute to the deprivation-based disparities we observed in long-term survival (9). Interestingly, the association between ADI and survival was most pronounced within the first three months of diagnosis, suggesting that that the window for interventions may be fairly short. Interventions to reduce early mortality may include improving knowledge of variability signs and symptoms of childhood cancer by primary care providers to reduce time from presentation to diagnosis (49–52). Further, children with public insurance may have less access to primary care and may be more likely to experience delays in diagnosis and patients may face challenges with navigating the healthcare system (51, 53). Gupta et al., proposed a model for the mechanisms linking domains of socioeconomic status (including material resources, knowledge-related assets, and social standing) to childhood cancer outcomes (54). This proposed model may explain how the material resource components of ADI are associated with survival; further research is needed to examine how other factors including interpersonal and community-level resources impact pediatric cancer outcomes.

Our study is the largest study to date of ADI and pediatric cancer survival. Although our population-based study also improves upon previous studies in part because of its diverse population, there are several limitations. State cancer registry data adheres to high data quality standards, but registry data do not have information on potential confounders such as individual-level household income, healthcare access, and comorbidities, although the prevalence of comorbidities in this population is generally low. Based on the E-values presented in Supplemental Table 6, our results comparing the most deprived neighborhoods to the least deprived neighborhoods are fairly robust to unmeasured confounding, especially for the analysis for early mortality. Additionally, ADI in our study was based on individuals’ addresses at diagnosis only and does not reflect any potential moves; if any individuals moved closer to their treatment facility into an urban, less-deprived neighborhood, our results may underestimate the actual effect. Due to power limitations, we were not able to stratify our results by cancer type beyond the three large categories; future larger cohort studies should explore a more detailed evaluation of potential differences by cancer type. Finally, our data may only be generalizable to populations with similar demographics and geographic locations.

Conclusion

This study contributes to the growing literature that neighborhood-level deprivation is associated with pediatric cancer outcomes. Further research is needed to understand the underlying pathways and mechanisms linking neighborhood deprivation to survival, which could include health insurance and treatment access. Neighborhood-level interventions may be needed to improve pediatric cancer outcomes.