How Sensitive Are Multilevel Regression Findings to Defined Area of Context?

Abstract

The authors develop a hybrid model of health care use that blends features of the traditional AdayAndersen behavioral model with the socioecological modeling perspective. They use the model to conceptualize the various levels of influence expected from socioecological variables in individuals' mammography use decisions, build contextual variables from fine-grained data into four different types of geographic areas, and then use two-and three-level modeling of personal and area-level contextual factors to explain observed behavior. The central focus is on whether differentiating the conceptualized levels of influence seems to materially affect regression findings. The test could conceivably be confounded by the modifiable areal unit problem, but little evidence for this is found. Findings for California women suggest that distinctions do matter in how the levels of influence are defined for local neighborhood contextual factors. Studies using only county-level contextual factors will miss some meaningful associations related to interpersonal/proximate-level factors. Keywordsaccess to care; mammography use; modifiable areal unit problem; multilevel model; socioecological model; breast cancer screening There are many different ways to combine microdata into variables describing small areas, and how this is done may affect statistical properties of the variables (Ali et al., 2005;Tatalovich, Wilson, Milam, Jerrett, & McConnell, 2006). Greenland (2002) and Pickett and Pearl (2000) discussed what this might mean for multilevel analyses. Empirical small area analyses (with areal-level aggregates for both the dependent and the explanatory variables) have found inconsistencies across models using different areal aggregations, which has been named the modifiable areal unit problem, or MAUP (Cockings & Martin, 2005;Fotheringham & Wong, 1991). To our knowledge, there have been no published studies with empirical assessments of MAUP in multilevel regression models, which is a major contribution of this article.To test for MAUP in multilevel models requires careful conceptualization of the relevant zones of influence for contextual variables (i.e., the appropriate level of aggregation for use in the analysis). A strong conceptual model is needed to guide the assessment because coefficient estimates can change for other reasons besides MAUP. These include differences in the relevance of the zones of influence defined (Pickett & Pearl, 2000), differences in statistical properties of the constructed variables because of MAUP (Greenland, 2002), or omission of important variables (Dietz, 2002 To identify all important variables and their relevant zones of influence, we developed a hybrid conceptual model that blends the best features of traditional behavioral models of access and use with socioecological models of health outcomes. This model situates the individual decision maker, distinguished by enabling-predisposing need characteristics, into an ecological context that has different zones of influence (fundamental-dista...

Multilevel built environment features and individual odds of overweight and obesity in Utah

“…It is widely acknowledged that effective interventions on health behaviors and outcomes occur on multiple levels (Nader, Bradley, Houts, McRitchie, & O'Brien, 2008). Mobley, Kuo, and Andrews (2008) examined how contextual variables in four types of geographic areas (post code areas, primary care service areas, medical service study areas, and county) affected the use of mammography service, and found inconsistent results across the four levels. Another study offered some insights speculating that small local areas might reflect social support while a large area unit might reflect geo-political units and minorities' political influence (Kuo, Mobley, & Anselin, 2011).…”

Spatial Sufficiency of 5% Medicare Standard Analytic Files

“…We use the socio-ecological model from previous research to specify a comprehensive set of variables to include in the multilevel model (Mobley et al, 2008a; 2008b; 2010). These include person-level, local area level, and county level variables.…”

“…The socio-demographic factors are defined for smaller areas known as primary care service areas (PCSA), which were formed from the aggregation of ZIP code tabulation areas to reflect Medicare patient travel to primary care providers (Goodman et al, 2003). PCSAs are smaller than counties, and are thought to better represent local neighborhood conditions (Mobley et al, 2008a). …”

No man is an island: disentangling multilevel effects in health services research

“…This complication highlights how it is often difficult to derive meaning from the area being measured—whether this represents a community, a healthcare service area or a policy planning area. Others have grappled with this issue and found that the findings vary by the choice of area-level unit 5. The lack of a common, reliable and valid method for measuring the concept of a community undermines our capacity to determine its contribution to healthcare outcomes.…”