Measuring And Modeling The Social And Geographic Context Of Trauma.
Ichiro Kawachi, MD, PhD, Professor of Social Epidemiology, Harvard School of Public Health, Boston.

Multi-level analysis is a technique that is increasingly encountered in public health and epidemiological research. For example, researchers have used multi-level approaches to examine diverse problems, such as the link between neighborhood poverty and individual mortality hazard; or the association between institutional variables (e.g. nursing home staff ratios) and the use of sedative medication among elderly residents; or the influence of work-place stress in deleterious health behaviors (e.g. alcohol abuse). In each of these examples, the data structure consists of individuals nested within different social contexts. Multi-level analysis provides a method of simultaneously accounting for the clustering of data (including health outcomes), as well as explicitly modeling the contextual heterogeneity in the outcomes of interest.

The study of trauma often involves multi-level data structures - for example, earthquakes, floods, or other natural disasters that affect some communities more severely than others. Although multi-level analytical techniques have seldom been used to analyze the public health consequences of trauma, the nature of many traumatic events lends itself to this form of analysis, including examining the community variations in organized responses to the traumatic event. In this presentation, I will provide a non-technical introduction to the basic concepts and tools of multi-level analysis. The presentation will review the typology of contextual variables, as well as the different types of multi-level data structures, and discuss a framework for considering cross-level causal mechanisms and threats to causal inference (including the "ecologic fallacy").