Back to Annual Meeting

Back to Annual Meeting

APHA Scientific Session and Event Listing

Recorded presentation

Donna H. Odierna, DrPH, School of Public Health, University of California, Berkeley, and Alcohol Research Group, PO Box 19064, Oakland, CA 94619, 510 531-5303, dodierna@berkeley.edu

Background: The people and groups who are most vulnerable to social inequalities and poor health may also be more difficult to recruit for population health research and to retain in longitudinal studies. When the instruments we use to look at the relationships between social inequality and health are themselves affected by social inequality, the risk of systematic research bias is increased.

Methods: Welfare/TANF and General Assistance recipients (n=688) in a Northern California county were surveyed in 2001 and then again one year later (RR=86%). Extensive tracking procedures were used to locate respondents. Baseline data were used to construct a nine-item index of marginalization. Relationships among marginalization, health, and attrition were examined.

Results: Nineteen percent were defined as severely marginalized (4+indicators). In bivariate analysis, severe marginalization significantly predicted attrition (p<.05), while individual indicators of marginalization did not (two indicators achieved borderline significance). Multivariate analysis showed that severe marginalization (OR=2.57), co-occurring substance dependence/psychological distress (OR=3.83), residential mobility (OR=2.10), age 24-35yrs. (OR=2.68), and being on General Assistance (OR=3.72) were significantly associated with attrition. Poor/fair health status, heavy substance use, and several other variables were associated with severe marginalization but not with attrition. There was a significant positive relationship between problem drinking and attrition in bivariate, but not in multivariate, analysis.

Discussion: Even after extensive tracking of the sample, and adjusting for other significant variables in analysis, the additive measure of severe marginalization predicted attrition while the component indicators did not. This suggests that there may be a cumulative effect, and merits further investigation.

Learning Objectives: