Google Earth Validation of Environmental Measures of Risk and Health Inequality

To improve research on the aspects of neighborhood environments that are amenable to structural and environmental interventions, reliable, valid and convenient assessment methods of malleable features of neighborhood environments are needed. This study examines the reliability and validity of a neighborhood virtual audit tool, The Virtual Neighborhood Inventory for Environmental Typology (V-NIfETy).  The V-NIfETy uses remote sensing techniques, namely the Street View feature in Google Earth, as a desk-based assessment of the neighborhood environment. A total of 802 block faces in each of Baltimore’s 242 residential neighborhood statistical areas were assessed using the V-NIfETy. Environmental characteristics measured via the V-NIfETy were validated against data obtained from a systematic social observational assessment based on day- and night-time assessments from the Neighborhood Inventory for Environmental Typology (NIfETy). The NIfETy includes data across 7 domains, namely (a) physical layout; (b) types of dwellings; (c) adult activity; (d) youth activity; (e) physical order and disorder; (f) social order and disorder; and (g) violence and ATOD indicators. The virtual data were also correlated with indices from US Census, namely community disadvantage index to establish predictive validity. While the daytime assessment is relatively inexpensive, easy to administer and is strongly correlated with youth-reported exposure to violence, alcohol, tobacco and other drugs the extent to which virtual measures of neighborhood environment relate to these outcomes is unknown. The virtual assessments are significantly cheaper to administer and offer great promise for use in remote locations, and areas of excess risk.  Limitations of the virtual assessment include the lack of data on the social aspects of neighborhood disorder and inconsistencies in the timing and resolution of data. Intra-observer reliability, criterion validity, and predictive validity were all in the fair to substantial range.   Implications of this method for future health equity research will be discussed.