Unpacking proxy variables: Cultural factors in adaptation to type II diabetes

Social scientists routinely employ sociodemographic variables such as race, ethnicity, and sex as independent variables in their research. These "social address" variables typically stand in, either explicitly or implicitly, for the more explanatory variables believed to underlie them. For instance, race and ethnicity often serve merely as proxies for the values, beliefs, and behaviors (i.e. culture) that are assumed to correlate with them.;"Unpacking" proxy variables---directly measuring the variables believed to underlie them---can provide a more reliable and more interpretable way of looking at group differences in patterns of illness, service use, and outcomes. The present study examines the factors hypothesized to underlie ethnicity as it relates to adaptation to, and outcomes of, managing type II diabetes in a veteran population.;Two instruments were developed to measure seven domains believed to correlate with ethnicity: economic marginality, domestic and family workload, domestic help, family relations, saliency of religion, proactive response to illness, and negative impact of illness. It was hypothesized that these domains would have an impact on the relationship with health care provider, severity of illness, utilization of urgent health care services, and quality of life.;Twelve Anglo veterans and 16 Hispanic veterans with type II diabetes were interviewed using the semi-structured Ecocultural Veteran Interview (EVI). These veterans, and an additional 17 Anglo veterans and 10 Hispanic veterans, also received a self-report instrument modeled after the EVI, the Ecocultural Veteran Self-Report (EVSR). Multitrait-Multimethod analyses were used to compare the reliability and validity of the two instruments. Sequential hierarchical general linear models were used to assess the utility of the measures in accounting for variance in the outcome measures.;Results indicate that the EVSR taps into the same domains as the more resource-intensive EVI. Furthermore, the domains are correlated with self-reported ethnic identification. These domains directly predict the relationship with provider, utilization of urgent health care services and quality of life. In addition, the domains interact with patient characteristics to predict severity of illness. The evidence from this study suggests that research focusing on improving the measurement of ecocultural variables in health services research is likely to be fruitful.