What are the associations between neighborhood poverty, child cognitive performance, and brain structure after accounting for household-level poverty?
This cross-sectional study evaluates whether neighborhood poverty is associated with cognitive function and prefrontal and hippocampal brain structure in ways that are dissociable from household socioeconomic status.
The association between poverty and unfavorable cognitive outcomes is robust, but most research has focused on individual household socioeconomic status (SES). There is increasing evidence that neighborhood context explains unique variance not accounted for by household SES.
To evaluate whether neighborhood poverty (NP) is associated with cognitive function and prefrontal and hippocampal brain structure in ways that are dissociable from household SES.
This cross-sectional study used a baseline sample of the ongoing longitudinal Adolescent Brain Cognitive Development (ABCD) Study. The ABCD Study will follow participants for assessments each year for 10 years. Data were collected at 21 US sites, mostly within urban and suburban areas, between September 2019 and October 2018. School-based recruitment was used to create a participant sample reflecting the US population. Data analysis was conducted from March to June 2019.
NP and household SES were included as factors potentially associated with National Institutes of Health Toolbox Cognitive Battery subtests and hippocampal and prefrontal (dorsolateral prefrontal cortex [DLPFC], dorsomedial PFC [DMPFC], superior frontal gyrus [SFG]) volumes. Independent variables were first considered individually and then together in mixed-effects models with age, sex, and intracranial volume as covariates. Structural equation modeling (SEM) was used to assess shared variance in NP to brain structure and cognitive task associations. The tested hypotheses were formulated after data collection.
A total of 11 875 children aged 9 and 10 years (5678 [47.8%] girls) were analyzed. Greater NP was associated with lower scores across all cognitive domains (eg, total composite: β = −0.18; 95% CI, −0.21 to −0.15; P < .001) and with decreased brain volume in the DLPFC (eg, right DLPFC: β = −0.09; 95% CI, −0.12 to −0.07; P < .001), DMPFC (eg, right DMPC: β = −0.07; 95% CI, −0.09 to −0.05; P < .001), SFG (eg, right SFG: β = −0.05; 95% CI, −0.08 to −0.03; P < .001), and right hippocampus (β = −0.04; 95% CI, −0.06 to −0.01; P = .01), even when accounting for household income. Greater household income was associated with higher scores across all cognitive domains (eg, total composite: β = 0.30; 95% CI, 0.28 to 0.33; P < .001) and larger volume in all prefrontal and hippocampal brain regions (eg, right hippocampus: β = 0.04; 95% CI, 0.02 to 0.07; P < .001) even when accounting for NP. The SEM model was a good fit across all cognitive domains, with prefrontal regions being associated with NP relations to language (picture vocabulary: estimate [SE], –0.03 [0.01]; P < .001; oral reading: estimate [SE], –0.02 [0.01]; P < .001), episodic memory (picture sequence: estimate [SE], –0.02 [0.01]; P = .008), and working memory (dimensional card sort: estimate [SE], –0.02 [0.01]; P = .001; flanker inhibitory control: estimate [SE], –0.01 [0.01]; P = .01; list sorting: estimate [SE], –0.03 [0.01]; P < .001) and hippocampal regions being associated with NP associations with language (picture vocabulary: estimate [SE], –0.01 [0.004]; P < .001) and episodic memory (picture sequence: estimate [SE], –0.01 [0.004]; P < 0.001).