This research sought to describe the neighborhood environment of Raleigh, NC, a city of the new urban south, as part of a cohort study of risk factors for adverse pregnancy outcomes. The new south is rapidly growing and may experience neighborhood changes in resources and maintenance that may be important to capture through direct observation. Conducting a windshield tour of Raleigh, NC and surrounding suburbs was necessary because of the large geography and low population density. Although direct observation data were collected via driving, we found we were able to use a data collection instrument previously used in Baltimore, MD to capture neighborhood attributes.
The second objective of this paper was to analyze race-stratified neighborhood attributes, indicating that, within the PIN sample, low-income non-Hispanic white and non-Hispanic black women live in qualitatively distinct neighborhoods. We found that non-Hispanic white women lived in neighborhoods with more amenities such as sidewalks, whereas non-Hispanic black women lived in neighborhoods characterized by more markers of incivilities. Based on theories of psychosocial etiology for adverse reproductive outcomes [42, 43], these very different environments may have important effects on racial disparities in preterm birth, a profound health disparity in the US, especially in the US south.
This particular neighborhood observation tool was chosen because the three theoretically informed constructs of physical incivilities, territoriality and social spaces are hypothesized to influence intermediate health outcomes during pregnancy such as stress level, diet, physical activity and weight status, as well as delivery outcomes of birthweight and preterm birth. Physical incivilities, characterized by poor housing, litter and abandoned houses, may directly and indirectly influence stress by increasing allostatic load or by influencing behaviors that help maintain low stress levels. Feelings of being unsafe might influence psychologically mediated pathways increasing stress and a physiological response to stress that over time increases a woman's allostatic load . This chronic stress condition has been presented as a weathering effect that over time influences poor health outcomes . Signs of physical incivilities that increased stress and decrease perceived safety may influence behavioral changes  such as the inability to exercise in one's own neighborhood  or increased gonorrhea rates . Conversely, territoriality is thought to communicate social control and have a protective affect on health, perhaps lowering allostatic load or increasing confidence to walk within one's neighborhood. The social spaces construct is hypothesized to promote personal interaction thereby increasing opportunities for social control and activity within one's neighborhood. To the extent that stress mediated pathways are involved in health outcomes, this neighborhood survey may be applicable for the study of other health outcomes such as weight status or chronic diseases.
Our research also sought to compare the attributes of a Raleigh, NC and its suburbs, a city of the new urban south, with those of Baltimore MD, a city with characteristics of the northern urban industrial center. Contrasting neighborhood attributes from various geographies is important because regardless of different developmental histories, similarities in neighborhood physical and observable manifestations that persist may help us understand how neighborhoods are important to health . Despite the scarcity of items representing incivilities in the Raleigh area, the physical incivilities scales had high internally reliability based on Cronbach's alpha scores, and territoriality and social spaces had moderate internal reliability. The low correlation estimates among the scales suggested that the scales captured distinct constructs and provided unique information about neighborhood attributes. We hypothesize that physical incivilities, territoriality and social spaces are importantly associated with reproductive health outcomes in Raleigh, NC and its surrounding suburbs, largely through psychosocially mediated pathways .
The fourth objective of this paper was to demonstrate that the unique neighborhood information obtained through direct observation is distinct from that of traditionally used census data. While the markers for incivilities, territoriality and social spaces may be used to estimate neighborhood deterioration, upkeep or resident investment, census variables can not replicate the information provided by these scales. Further, the theoretically informed scales suggest a mechanism regarding how neighborhoods can influence health outcomes. The inadequacy of using poverty as a surrogate for neighborhood dynamics is due to heterogeneity across low-income neighborhoods with regards to disadvantage, crime, and resources, as has been observed in previous studies . In a study of neighborhood effects on gonorrhea rates in New Orleans, LA, Cohen et al. found that a "Broken Windows" index – a directly observed measure combining housing condition, graffiti, accumulated garbage, abandoned vehicles and public high schools with problems – distinguished among low-income neighborhoods . Low-income, low broken windows indexed neighborhoods had significantly lower gonorrhea rates than low-income, high broken windows indexed neighborhoods. These illustrations show the importance of using directly observed data in combination with census or other administrative data; geo-referenced data such as parks, commerce, schools, zoning, alcohol outlets, and crime data [27, 47]; and perceived neighborhood environment data , to provide a rich picture of neighborhoods and their attributes, with minimal investment of time and expense, and to better understand mechanisms of neighborhood influences on health. In addition, increased accessibility to geocoded data has enabled more sophisticated modeling techniques and permit exposures to be characterized as simple counts or as rates for various units of geographic analysis [49, 50]. Geocoding allows one to observe the spatial distribution of an exposure over multiple geographies to identify hot spots, assess spatial autocorrelation, and allows the creation of accessibility measures and geo-simulation . The utility of different modeling techniques permits exploration of the most relevant exposure classification for health outcomes. In this way not only can the relationship of geography be better understood but the influence of changes in terrain on health can be assessed enabling researcher to explore causal mechanisms and move beyond simple associations.
Although newly developed southern US cities are notably less segregated than the industrial centers of the northeast , and patterns of poverty and neighborhood development are different because of the growth of these areas in an era since the demise of heavy industry as the basis for economic organization , the recent establishment of these communities may provide fewer social resources that could help to buffer effects of harmful environments. Furthermore, cities in which major growth has occurred since the automobile became ubiquitous are more geographically dispersed and may reduce easy access to facilities and amenities compared to cities with concentrated population centers and long-established urban transit systems. Reduced service concentration may be especially burdensome for poor individuals and families who may not own a car or have hours to devote to traveling between service facilities. Furthermore, recent growth in new south centers such as Raleigh, Charlotte and Atlanta has occurred since the era of suburban flight, meaning that center-city areas were never abandoned, since the center city never gained prominence in this later era. This implies a lower prevalence of the 'incivilities' that emerge when populations abandon decaying areas of the city for opportunities in newer suburbs. Yet, even with the lower prevalence of incivilities, their existence may influence health outcomes, and as population growth and development occurs, incivilities in poorly maintained neighborhoods may increase.
Future research is needed to corroborate data collection methods and findings. Directly observed neighborhood attributes can be combined with geographic information systems and resource inventories to validate findings, and can be augmented by these sources and census data to provide a detailed contextual database for the analysis of neighborhoods' influences on health outcomes. Longitudinal data collection and analysis of individuals and the neighborhoods in which they reside will be important as we move forward with this research. Analysis using the physical incivilities, territoriality and social spaces scales to predict health outcomes, particularly adverse birth outcomes is needed and forthcoming.