Background: The contribution of 'environment' has been investigated across diverse and multiple domains related to health. However, in the context of large-scale genomic studies the focus has been on obtaining individual-level endophenotypes with environment left for future decomposition. Geo-social research has indicated that environment-level variables can be reduced, and these composites can then be used with other variables as intuitive, precise representations of environment in research.
Method: Using a large community sample (N = 9498) from the Philadelphia area, participant addresses were linked to 2010 census and crime data. These were then factor analyzed (exploratory factor analysis; EFA) to arrive at social and criminal dimensions of participants' environments. These were used to calculate environment-level scores, which were merged with individual-level variables. We estimated an exploratory multilevel structural equation model (MSEM) exploring associations among environment- and individual-level variables in diverse communities.
Results: The EFAs revealed that census data was best represented by two factors, one socioeconomic status and one household/language. Crime data was best represented by a single crime factor. The MSEM variables had good fit (e.g. comparative fit index = 0.98), and revealed that environment had the largest association with neurocognitive performance (β = 0.41, p < 0.0005), followed by parent education (β = 0.23, p < 0.0005).
Conclusions: Environment-level variables can be combined to create factor scores or composites for use in larger statistical models. Our results are consistent with literature indicating that individual-level socio-demographic characteristics (e.g. race and gender) and aspects of familial social capital (e.g. parental education) have statistical relationships with neurocognitive performance.
Keywords: Census; Philadelphia Neurodevelopmental Cohort; factor analysis; geo-coding; neurocognition; socioeconomic status.