PubMed1997
In the last decade violence has emerged as a public health issue, with concomitant interest in surveillance and prevention. This paper is an extension of earlier work seeking to expand the ecological analysis of violence across jurisdictional boundaries, sharpen the level of resolution of such analysis, and ultimately inform small area policy applications in terms of public health initiatives for violence reduction. The underlying model is drawn from stress theory and rests on a set of social indicators representing stress in the Baltimore area. In the earlier work, a set of 24 variables describing violence and socioeconomic conditions across some 1358 areas was factor analyzed and the resulting scores were mapped and interpreted. The present paper takes the analysis a step further in an attempt to identify groups of observations with common traits in order to assist public health professionals and other relevant decision-makers in the process of trauma surveillance, response, and prevention. Cluster analysis was used to combine most similar observations in terms of the three orthogonal factors, and the resulting cluster affiliations were mapped in geographic space. Although no spatial contiguity constraint was put on the clustering algorithm, many statistical clusters were also found to constitute geographic clusters. This implies that the process identified neighborhoods or parts of neighborhoods with shared traits in terms of the underlying set of stressors. Analysis of this type could be used by policy-makers to classify neighborhoods in terms of their needs for various services in addition to public health interventions, including policing, fire protection, building inspection, social work, and education.