This cohort study examines data for children with health insurance to gauge the association between a diagnosis of autism spectrum disorder and neighborhood disadvantage at birth and determine any potential effect modification by maternal and child demographic characteristics.
Key Points
Question
Is neighborhood disadvantage associated with a higher likelihood of the diagnosis of autism spectrum disorder (ASD) in a population with health insurance?
Findings
In a cohort study that included 318 372 children with health insurance, neighborhood disadvantage at birth was associated with a higher likelihood of ASD diagnoses, independent of maternal education.
Meaning
Providing resources for early intervention and family support in communities with a higher likelihood of ASD is important, while maintaining investment for universal screening.
Abstract
Importance
Family socioeconomic status has been associated with autism spectrum disorder (ASD) diagnoses. Less is known regarding the role of neighborhood disadvantage in the United States, particularly when children have similar access to health insurance.
Objective
To evaluate the association between neighborhood disadvantage and the diagnosis of ASD and potential effect modification by maternal and child demographic characteristics.
Design, Setting, and Participants
This cohort study examined a retrospective birth cohort from Kaiser Permanente Southern California (KPSC), an integrated health care system. Children born in 2001 to 2014 at KPSC were followed up through KPSC membership records. Electronic medical records were used to obtain an ASD diagnosis up to December 31, 2019, or the last follow-up. Data were analyzed from February 2022 to September 2023.
Exposure
Socioeconomic disadvantage at the neighborhood level, an index derived from 7 US census tract characteristics using principal component analysis.
Main Outcomes and Measures
Clinical ASD diagnosis based on electronic medical records. Associations between neighborhood disadvantage and ASD diagnosis were determined by hazard ratios (HRs) from Cox regression models adjusted for birth year, child sex, maternal age at delivery, parity, severe prepregnancy health conditions, maternal race and ethnicity, and maternal education. Effect modification by maternal race and ethnicity, maternal education, and child sex was assessed.
Results
Among 318 372 mothers with singleton deliveries during the study period, 6357 children had ASD diagnoses during follow-up; their median age at diagnosis was 3.53 years (IQR, 2.57-5.34 years). Neighborhood disadvantage was associated with a higher likelihood of ASD diagnosis (HR, 1.07; 95% CI, 1.02-1.11, per IQR = 2.70 increase). Children of mothers from minoritized racial and ethnic groups (African American or Black, Asian or Pacific Islander, Hispanic or Latinx groups) had increased likelihood of ASD diagnosis compared with children of White mothers. There was an interaction between maternal race and ethnicity and neighborhood disadvantage (difference in log-likelihood = 21.88; P < .001 for interaction under χ24); neighborhood disadvantage was only associated with ASD among children of White mothers (HR, 1.17; 95% CI, 1.09-1.26, per IQR = 2.00 increase). Maternal education and child sex did not significantly modify the neighborhood-ASD association.
Conclusions and Relevance
In this study, children residing in more disadvantaged neighborhoods at birth had higher likelihood of ASD diagnosis among a population with health insurance. Future research is warranted to investigate the mechanisms behind the neighborhood-related disparities in ASD diagnosis, alongside efforts to provide resources for early intervention and family support in communities with a higher likelihood of ASD.
Introduction
Autism spectrum disorder (ASD) affects 2.3% of US children1 and is associated with hardships2 and economic burdens3 for children and families. While family-level socioeconomic status (SES) has been associated with ASD,4,5 less is known about neighborhood disadvantage, which relates to health independent of family SES6,7 and is associated with environmental and contextual factors that may affect maternal health during pregnancy and child development. Only 3 studies considered individual ASD diagnoses with family- and neighborhood-level SES simultaneously, with mixed findings.8,9,10 Higher SES was associated with lower ASD likelihood in countries with free and universal health care.11 While the opposite pattern has been found in the United States,10 it reversed after 2010, with lower SES related to higher ASD diagnosis rates more recently.12 This may be due to changes in access to developmental screening, as inequities in screening may mask underlying disparities.
Consequently, studies are needed to determine if neighborhood disadvantage is associated with increased ASD diagnoses when reducing inequities in access to screening and diagnostic services and if such disparities vary based on child and maternal sociodemographic characteristics. To address these questions, we used a birth cohort in Kaiser Permanente Southern California (KPSC) hospitals for whom integrated insurance and care ensures greater access to, and reduced disparities in, diagnostic screening for ASD.
Methods
Study Population
This retrospective cohort included mothers with singleton deliveries at KPSC hospitals from 2001 to 2014 (eMethods and eFigure 1 in Supplement 1). Characteristics and covariates for mothers and children were extracted from electronic medical records (eMethods in Supplement 1). ASD diagnosis was ascertained based on at least 2 clinical diagnoses using codes from the International Classification of Diseases, Ninth Revision, and International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (eMethods and eTable 1 in Supplement 1).13 Maternal race and ethnicity based on self-report is included in the electronic medical record and was recorded as African American or Black, Asian or Pacific Islander, Hispanic or Latinx, White, and other or multiple. Children were followed up from age 1 year until the first ASD diagnosis, the last date of KPSC membership, death, or December 31, 2019, whichever occurred first.
The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines. Institutional review boards at KPSC and the University of Southern California approved this study with a waiver of individual participant consent.
Neighborhood Disadvantage
Neighborhood was defined as the US census tract of the residential address at the child’s birth. Data were extracted from the 2000 US decennial census for births from 2001 to 2004 and from American Community Survey 5-year estimates (2005-2009 and 2010-2014) for corresponding birth years. Neighborhood disadvantage was defined as the first principal component of 7 census tract indicators14: poverty, unemployment, female-headed households with children, public assistance, less than a high school education, bachelor’s degree or greater, and professional occupation (eMethods, eFigure 2, and eTable 2 in Supplement 1).
Statistical Analysis
We used Cox proportional hazard regression models with census tracts modeled as random intercepts adjusting for covariates. Neighborhood disadvantage was treated as a continuous variable, and the linear association was estimated as a hazard ratio (HR) scaled to its IQR. Neighborhood disadvantage was also categorized into quintiles to examine nonlinearity. Multiplicative interaction tests were used to examine effect modification by maternal race and ethnicity, maternal education, and child sex in fully adjusted models. The global interaction was tested using the likelihood ratio test comparing 2 models, 1 with and 1 without the interaction terms. If the global interaction was significant, stratification analysis was conducted, with neighborhood disadvantage scaled to a stratum-specific IQR. Two-sided statistical tests were used (α = .05) and 95% confidence intervals were reported. Sensitivity analyses are described in the eMethods in Supplement 1. Analyses were performed in R version 4.2 (R Foundation) during the time period February 2022 to September 2023.
Results
Among 318 372 mothers with singleton deliveries during the study period, there were 6357 children (2.00%) diagnosed with ASD at a median age of 3.53 years (IQR, 2.57-5.34 years). Males (n = 5164) were 4.3 times more likely than females (n = 1193) to have ASD diagnoses (Table 1).
Table 1. Characteristics of Children With and Without ASD Diagnosis.
| Characteristic | Children, No. (%) | ||
|---|---|---|---|
| Overall (n = 318 372) | With ASD (n = 6357) | Without ASD (n = 312 015) | |
| Sex | |||
| Male | 162 986 (51.2) | 5164 (81.2) | 157 822 (50.6) |
| Female | 155 386 (48.8) | 1193 (18.8) | 154 193 (49.4) |
| Maternal age at delivery, median (IQR), y | 30.4 (26.3 to 34.3) | 31.3 (27.5 to 35.2) | 30.4 (26.2 to 34.3) |
| Parity | |||
| 0 | 111 854 (35.1) | 2656 (41.8) | 109 198 (35.0) |
| 1 | 104 428 (32.8) | 2042 (32.1) | 102 386 (32.8) |
| ≥2 | 84 091 (26.4) | 1265 (19.9) | 82 826 (26.5) |
| Unknown | 17 999 (5.7) | 394 (6.2) | 17 605 (5.6) |
| Maternal education | |||
| High school or lower | 112 045 (35.2) | 1911 (30.1) | 110 134 (35.3) |
| Some college | 94 418 (29.7) | 2036 (32.0) | 92 382 (29.6) |
| College graduate or higher | 108 871 (34.2) | 2362 (37.2) | 106 509 (34.1) |
| Unknown | 3038 (1.0) | 48 (0.8) | 2990 (1.0) |
| Race and ethnicity | |||
| African American or Black | 29 752 (9.4) | 645 (10.1) | 29 107 (9.3) |
| Asian or Pacific Islander | 39 854 (12.5) | 929 (14.6) | 38 925 (12.5) |
| Hispanic or Latinx | 161 319 (50.7) | 3181 (50.0) | 158 138 (50.7) |
| White | 80 909 (25.4) | 1445 (22.7) | 79 464 (25.5) |
| Othera | 6538 (2.1) | 157 (2.5) | 6381 (2.0) |
| History of maternal comorbidityb | 46 647 (14.7) | 1150 (18.1) | 45 497 (14.6) |
| Year of birth | |||
| 2001-2004 | 78 236 (24.6) | 1439 (22.6) | 76 797 (24.6) |
| 2005-2009 | 111 064 (34.9) | 2003 (31.5) | 109 061 (35.0) |
| 2010-2014 | 129 072 (40.5) | 2915 (45.9) | 126 157 (40.4) |
| Neighborhood SES measures, median (IQR)c | |||
| Povertyd | 9.5 (4.7 to 17.3) | 10.0 (4.8 to 17.8) | 9.5 (4.7 to 17.3) |
| Unemploymentd | 8.8 (6.2 to 11.9) | 8.9 (6.2 to 11.9) | 8.8 (6.2 to 11.9) |
| Female-headed householdsd | 17.5 (11.7 to 24.6) | 17.8 (12.1 to 25.3) | 17.5 (11.7 to 24.6) |
| Public assistanced | 3.3 (1.6 to 6.3) | 3.4 (1.7 to 6.4) | 3.3 (1.6 to 6.3) |
| Less than high schoold | 21.2 (10.6 to 35.6) | 21.6 (10.3 to 35.9) | 21.2 (10.6 to 35.6) |
| Bachelor and postgraduated | 19.3 (10.5 to 32.5) | 19.7 (10.8 to 33.5) | 19.3 (10.5 to 32.5) |
| Professional occupationd | 28.9 (19.1 to 40.8) | 28.6 (19.1 to 41.5) | 28.9 (19.1 to 40.8) |
| Disadvantage index | 0.07 (−1.22 to 1.48) | 0.12 (−1.25 to 1.54) | 0.07 (−1.22 to 1.48) |
Abbreviations: ASD, autism spectrum disorder; SES, socioeconomic status.
The category other race and ethnicity includes patients who identified as other or multiple races.
≥1 Diagnoses of heart, lung, kidney, or liver disease or cancer.
Based on census tract at birth address.
Percentage of each neighborhood SES census tract indicator, for example, percentage of families below the poverty line across all census tracts.
The neighborhood disadvantage index explained 66.1% of the variance in the neighborhood SES indicators. The index exhibited variability by maternal race and ethnicity, maternal education, and child sex (eFigure 3 in Supplement 1).
Higher neighborhood disadvantage at birth was associated with higher likelihood of ASD diagnosis (HR, 1.09; 95% CI, 1.05-1.13) per IQR increase in disadvantage score, adjusted for birth year, maternal age, parity, maternal comorbidity, and medical center (Table 2, model 1; crude associations in eTable 3 in Supplement 1). While further adjustment for maternal race and ethnicity, maternal education, and child sex attenuated the point estimate, the association was still significant (HR, 1.07; 95% CI, 1.02-1.11, per IQR = 2.70 increase) (Table 2, model 2 and Figure, A). Residence in neighborhoods in the fourth and fifth highest quintiles of disadvantage was associated with higher ASD likelihood (Table 2 and eFigure 4A in Supplement 1). Similar associations were found for each separate neighborhood indicator and alternative neighborhood composites (eMethods and eTable 4 in Supplement 1).
Table 2. Associations Between Neighborhood Disadvantage and ASD Diagnosis in Models Adjusting for Varying Sets of Covariates.
| Hazard ratio (95% CI) | ||
|---|---|---|
| Model 1a | Model 2b | |
| Disadvantage index (per IQR = 2.70) | 1.09 (1.05-1.13) | 1.07 (1.02-1.11) |
| Disadvantage index quintile | ||
| First | 1 [Reference] | 1 [Reference] |
| Second | 0.98 (0.91-1.06) | 0.97 (0.89-1.05) |
| Third | 1.03 (0.95-1.12) | 1.01 (0.93-1.09) |
| Fourth | 1.17 (1.08-1.27) | 1.14 (1.05-1.24) |
| Fifth | 1.16 (1.07-1.26) | 1.12 (1.02-1.22) |
Abbreviation: ASD, autism spectrum disorder.
Model 1 adjusted for birth year, maternal age, parity, history of maternal comorbidity, and medical center. Census tracts are modeled as random intercepts.
Model 2 adjusted for covariates in model 1 plus maternal race and ethnicity, maternal education, and child sex at birth. Census tracts are modeled as random intercepts.
Figure. Associations Between Neighborhood Disadvantage, Maternal Race and Ethnicity, and Autism Spectrum Disorder (ASD) Diagnosis.
A, Main association between neighborhood disadvantage (per IQR = 2.70 increase) and ASD in the entire study population from the fully adjusted model 2. B, Main association between maternal race and ethnicity and ASD in the entire study population from the fully adjusted model 2. C, Association between neighborhood disadvantage and ASD in analyses fully stratified by maternal race and ethnicity scaled to the stratum-specific IQR, adjusted for birth year, maternal age, parity, history of maternal comorbidity, medical center, maternal education, and child sex at birth. IQRs for African American or Black, Asian or Pacific Islander, Hispanic or Latinx, White, and other race and ethnicity were 2.71, 2.25, 2.45, 2.00, and 2.36, respectively. Similar results were found when the coefficient for each race and ethnicity group was extracted from the interaction model (eTable 8 in Supplement 1). The category other race and ethnicity includes patients who identified as other or multiple races. Census tracts were modeled as random intercepts. The interpretation of the coefficient was an effect corresponding to the range of exposure of the middle 50% of the population.
Children of mothers from minoritized racial and ethnic groups had higher likelihood of ASD diagnosis than children of White mothers (Figure, B and eTable 5 in Supplement 1). A significant interaction was observed between maternal race and ethnicity and neighborhood disadvantage (difference in log-likelihood = 21.88; P < .001 for interaction under χ24). Neighborhood disadvantage was significantly associated with ASD diagnosis only for children of White mothers (HR, 1.17; 95% CI, 1.09-1.26, per IQR = 2.00 increase in White race) (Figure, C). Similar patterns were observed for alternative neighborhood SES measures (eTable 6 in Supplement 1) and a dichotomized indicator of high disadvantage at quintiles 4 and 5 (eFigure 4C in Supplement 1).
There was no significant effect modification by maternal education (difference in log-likelihood = 7.00; P = .07 for interaction under χ23) or child sex (difference in log-likelihood = 2.76; P = .10 for interaction under χ21). Sensitivity analyses produced similar results (eResults and eTables 7 and 8 in Supplement 1).
Discussion
This study found that neighborhood disadvantage was associated with higher likelihood of ASD diagnosis, independent of maternal education, in a large birth cohort with access to health insurance. This result is consistent with studies in Sweden, where screening and health care are universally accessible.8 Thus, when disparities in access to screening and diagnostic services are reduced, neighborhood disadvantage is associated with higher likelihood of ASD diagnosis.
While children of mothers from minoritized racial and ethnic groups had higher likelihood of ASD diagnosis, the association between neighborhood disadvantage and ASD diagnosis was only observed for children of White mothers, a pattern similar to the stronger associations between neighborhood disadvantage and preterm birth and low birthweight in White populations.7 The mechanisms behind these patterns are unknown and merit future research. However, they may be due to systemic racism, discrimination, and their impact on maternal health during pregnancy.15 Thus, it is possible that families from minoritized racial and ethnic groups may not experience similar health improvements with increasing neighborhood SES as White populations because of the discrimination and related stressors and barriers experienced in more affluent neighborhoods.15
Strengths and Limitations
The study has several strengths, including the use of an integrated system with standard diagnostic procedures, a large study population, and ample statistical power. In addition, results were robust to sensitivity analyses and the use of alternative neighborhood measures.
With respect to limitations, a lack of information on fathers and socioeconomic indices such as income may have led to unmeasured family-level confounding. Second, although the KPSC membership covers both commercial insurance and California's Medicaid health care program, results need replication in other health service settings. Third, even within an integrated health system, access to care may vary by factors such as co-payments. Additionally, the prevalence of ASD may not be accurately estimated based on clinical diagnoses in electronic medical records.
Conclusions
In this retrospective cohort study, neighborhood disadvantage was associated with higher likelihood of ASD diagnosis when differences in access to care were reduced by providing similar health insurance. However, the benefits of reduced neighborhood disadvantage were only observed in White populations while children of mothers from minoritized racial and ethnic groups had higher likelihood of ASD diagnosis. Future research is needed to identify the mechanisms underlying these complex associations. Investment in early intervention and family support for communities and populations with higher diagnosis rates, along with universal access to screening and diagnosis, is warranted.
eMethods. Supplemental Methodological Details
eResults. Sensitivity Analyses
eTable 1. ICD codes and diagnostic description for ASD ascertainment
eTable 2. Factor loadings of the primary neighborhood disadvantage index and indices of neighborhood disadvantage and advantage
eTable 3. Crude associations between neighborhood disadvantage index and ASD
eTable 4. Associations between neighborhood disadvantage components, neighborhood deprivation and advantage indices, and ASD in models adjusting for varying sets of covariates
eTable 5. Multivariable-adjusted associations between race and ethnicity, maternal education, and ASD
eTable 6. Stratified analyses: Associations between neighborhood characteristics and ASD by maternal race and ethnicity
eTable 7. Sensitivity analyses: Adjusted associations between neighborhood disadvantage index and ASD
eTable 8. Associations between neighborhood disadvantage index and ASD by race and ethnicity based on the interaction model
eFigure 1. Derivation of study sample.
eFigure 2. The distribution of neighborhood SES measures from the three census sources among all census tracts in Southern California
eFigure 3. The distribution of neighborhood disadvantage index by maternal and children demographic characteristics
eFigure 4. The main and interactive associations between dichotomized neighborhood disadvantage (quintiles 4-5 vs quintiles 1-3), maternal race and ethnicity and ASD
eReferences
Data sharing statement
References
- 1.Maenner MJ, Shaw KA, Bakian AV, et al. Prevalence and characteristics of autism spectrum disorder among children aged 8 years: Autism and Developmental Disabilities Monitoring Network, 11 sites, United States, 2018. MMWR Surveill Summ. 2021;70(11):1-16. doi: 10.15585/mmwr.ss7011a1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Kuhlthau K, Orlich F, Hall TA, et al. Health-related quality of life in children with autism spectrum disorders: results from the Autism Treatment Network. J Autism Dev Disord. 2010;40(6):721-729. doi: 10.1007/s10803-009-0921-2 [DOI] [PubMed] [Google Scholar]
- 3.Buescher AV, Cidav Z, Knapp M, Mandell DS. Costs of autism spectrum disorders in the United Kingdom and the United States. JAMA Pediatr. 2014;168(8):721-728. doi: 10.1001/jamapediatrics.2014.210 [DOI] [PubMed] [Google Scholar]
- 4.Bhasin TK, Schendel D. Sociodemographic risk factors for autism in a US metropolitan area. J Autism Dev Disord. 2007;37(4):667-677. doi: 10.1007/s10803-006-0194-y [DOI] [PubMed] [Google Scholar]
- 5.Durkin MS, Maenner MJ, Meaney FJ, et al. Socioeconomic inequality in the prevalence of autism spectrum disorder: evidence from a U.S. cross-sectional study. PLoS One. 2010;5(7):e11551. doi: 10.1371/journal.pone.0011551 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Minh A, Muhajarine N, Janus M, Brownell M, Guhn M. A review of neighborhood effects and early child development: how, where, and for whom, do neighborhoods matter? Health Place. 2017;46:155-174. doi: 10.1016/j.healthplace.2017.04.012 [DOI] [PubMed] [Google Scholar]
- 7.Ncube CN, Enquobahrie DA, Albert SM, Herrick AL, Burke JG. Association of neighborhood context with offspring risk of preterm birth and low birthweight: a systematic review and meta-analysis of population-based studies. Soc Sci Med. 2016;153:156-164. doi: 10.1016/j.socscimed.2016.02.014 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Li X, Sjöstedt C, Sundquist K, Zöller B, Sundquist J. Neighborhood deprivation and childhood autism: a nationwide study from Sweden. J Psychiatr Res. 2014;53:187-192. doi: 10.1016/j.jpsychires.2014.02.011 [DOI] [PubMed] [Google Scholar]
- 9.Kelly B, Williams S, Collins S, et al. The association between socioeconomic status and autism diagnosis in the United Kingdom for children aged 5-8 years of age: findings from the Born in Bradford cohort. Autism. 2019;23(1):131-140. doi: 10.1177/1362361317733182 [DOI] [PubMed] [Google Scholar]
- 10.King MD, Bearman PS. Socioeconomic status and the increased prevalence of autism in California. Am Sociol Rev. 2011;76(2):320-346. doi: 10.1177/0003122411399389 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Rai D, Lewis G, Lundberg M, et al. Parental socioeconomic status and risk of offspring autism spectrum disorders in a Swedish population-based study. J Am Acad Child Adolesc Psychiatry. 2012;51(5):467-476.e6. doi: 10.1016/j.jaac.2012.02.012 [DOI] [PubMed] [Google Scholar]
- 12.Winter AS, Fountain C, Cheslack-Postava K, Bearman PS. The social patterning of autism diagnoses reversed in California between 1992 and 2018. Proc Natl Acad Sci U S A. 2020;117(48):30295-30302. doi: 10.1073/pnas.2015762117 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Coleman KJ, Lutsky MA, Yau V, et al. Validation of autism spectrum disorder diagnoses in large healthcare systems with electronic medical records. J Autism Dev Disord. 2015;45(7):1989-1996. doi: 10.1007/s10803-015-2358-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Levy BL, Owens A, Sampson RJ. The varying effects of neighborhood disadvantage on college graduation: moderating and mediating mechanisms. Sociol Educ. 2019;92(3):269-292. doi: 10.1177/0038040719850146 [DOI] [Google Scholar]
- 15.Colen CG, Ramey DM, Cooksey EC, Williams DR. Racial disparities in health among nonpoor African Americans and Hispanics: the role of acute and chronic discrimination. Soc Sci Med. 2018;199:167-180. doi: 10.1016/j.socscimed.2017.04.051 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eMethods. Supplemental Methodological Details
eResults. Sensitivity Analyses
eTable 1. ICD codes and diagnostic description for ASD ascertainment
eTable 2. Factor loadings of the primary neighborhood disadvantage index and indices of neighborhood disadvantage and advantage
eTable 3. Crude associations between neighborhood disadvantage index and ASD
eTable 4. Associations between neighborhood disadvantage components, neighborhood deprivation and advantage indices, and ASD in models adjusting for varying sets of covariates
eTable 5. Multivariable-adjusted associations between race and ethnicity, maternal education, and ASD
eTable 6. Stratified analyses: Associations between neighborhood characteristics and ASD by maternal race and ethnicity
eTable 7. Sensitivity analyses: Adjusted associations between neighborhood disadvantage index and ASD
eTable 8. Associations between neighborhood disadvantage index and ASD by race and ethnicity based on the interaction model
eFigure 1. Derivation of study sample.
eFigure 2. The distribution of neighborhood SES measures from the three census sources among all census tracts in Southern California
eFigure 3. The distribution of neighborhood disadvantage index by maternal and children demographic characteristics
eFigure 4. The main and interactive associations between dichotomized neighborhood disadvantage (quintiles 4-5 vs quintiles 1-3), maternal race and ethnicity and ASD
eReferences
Data sharing statement