Abstract
This manuscript describes the development of the Census of Social Institutions (CSI), a reliable direct observation parcel-level built environment measure. The CSI was used to measure all non-residential parcels (n = 10,842) in 21 one-mile-radius neighborhoods centered around census block groups of varying income and ethnicity in a large metropolitan area. One year test-retest and inter-rater intra-class correlations showed high reliability for major use type and detail code observations. The CSI accurately captured the presence of about 9,500 uses, including 828 multiple major use and 431 mixed major use parcels that would have been missed in standard commercial databases. CSI data can be utilized to determine the health impacts of environmental settings.