American Journal of Epidemiology Advance Access originally published online on May 17, 2006
American Journal of Epidemiology 2006 163(12):1083; doi:10.1093/aje/kwj178
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Response to Invited Commentary |
Wight et al. Respond to "Considerations about Multilevel Thinking"
1 Department of Community Health Sciences, School of Public Health, University of California, Los Angeles, CA
2 Division of Geriatrics, David Geffen School of Medicine, University of California, Los Angeles, CA
Correspondence to Dr. Richard G. Wight, UCLA School of Public Health, 650 Charles E. Young Drive South, Box 951772, Los Angeles, CA 90095 (e-mail: rwight{at}ucla.edu).
Received for publication March 20, 2006. Accepted for publication March 23, 2006.
Galea and Ahern (1
) have provided an insightful commentary on our article (2), raising several noteworthy observations pertaining to the current state of and future directions for multilevel epidemiologic research. We concur that theory-driven multilevel studies are needed, along with explication of the mechanisms that link levels of influence on health, particularly since these mechanisms may be more readily amenable to intervention than structural determinants of health. For example, in our analyses of the association between neighborhood education and late-life cognition (2), we recommend promoting educational attainment in the general population in disadvantaged neighborhoods because it may prove cognitively beneficial to aging residents of these neighborhoods. An argument could be made instead for promoting alteration to the hypothesized mechanisms by which we believe educational attainment would benefit aging residents, such as engaging in physical and social activities or increasing access to cognitively stimulating material resources, because these mechanisms are less wieldy from a policy standpoint and are more proximal to the target population. However, without longitudinal data to test the impact of these mechanisms, this possibility is purely speculative. By specifying causal mechanisms, theory-based investigations are well poised to distinguish mediators from confounders or analytic noise.
As Galea and Ahern (1) note, a drawback to implementing theory-based multilevel investigations concerns the current lack of appropriate data. With growing evidence for apparent area-level effects, there will hopefully be greater impetus to develop data sets that specify the potential mechanisms linking levels of influence and that account for the inherent nesting of individuals within contexts. To date, many multilevel studies have been undertaken with data not originally designed to assess multiple levels of causality (as was the case with data from the Study of Assets and Health Dynamics Among the Oldest Old (AHEAD) analyzed in our study (2)), thereby hindering examinations of heterogeneous associations or limiting the set of variables available for inclusion in multivariate models. For example, the power to detect cross-level interactions is limited when there are too few cases per macro-level unit to provide sufficient variability in variables of interest. Although there are approaches for solving the problem of "data sparseness" (e.g., cluster analysis, sample stratification), the problem would best be addressed by designing projects, a priori, to contain ample data points per contextual unit. Similarly, testing causal mechanisms between levels can occur only if these pathways are operationalized as variables. Analysis of data sets specifically developed to test such pathways would represent a sizable improvement over secondary analysis of data sets developed for other purposes, which leave the investigator with the option of making use of only those constructs that happen to be available.
By deriving theory-driven research models and applying them to appropriate data sets, future multilevel epidemiologic studies stand to make substantial contributions to our understanding of not only why but how contexts matter to health.
 |
ACKNOWLEDGMENTS |
|---|
Conflict of interest: none declared.
|
References |
|---|
 TOP References |
|---|
- Galea S, Ahern J. Invited commentary: considerations about specificity of associations, causal pathways, and heterogeneity in multilevel thinking. Am J Epidemiol 2006;163:1079–82.
[Free Full Text] - Wight RG, Aneshensel CS, Miller-Martinez D, et al. Urban neighborhood context, educational attainment, and cognitive function among older adults. Am J Epidemiol 2006;163:1071–8.
[Abstract/Free Full Text]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||