2006 HSR&D National Meeting Abstract
2007 — Self-Report Measures in Disparities Research: Assessing Measurement Bias in Diverse Populations
Kallen MA (METRIC)
Kuykendall DH (METRIC)
Among the more important measurement issues in health disparities research is that of measurement bias: Measurements are biased when scores on the trait of interest (e.g., health status, patient satisfaction) are influenced by factors other than that trait, such as culture and literacy. Methods for identifying measurement bias include analyses for assessing differential item and test functioning (DIF/DTF). DIF is the tendency of an instrument item to exhibit bias across different subpopulations. Although a single biased item-level measurement does not necessarily make the overall scale’s measurements biased, bias accumulates. DTF analyses assess whether enough item-level bias exists to bias scale measurement. Thus, DIF and DTF analyses can be used to assess the cultural equivalence of measures, and hence, whether items and scales used to measure self-reported health status and other constructs of interest in disparities research are culturally fair. Investigation of DIF and DFT should be considered in studies using self-report measures to assess differences among diverse populations (e.g., populations differing by race, age, gender, or veteran status). After attending the workshop, participants: 1) should understand what DIF/DTF analyses are, 2) be able to describe how health disparities research benefits from DIF/DTF assessment, and 3) appreciate that DIF/DTF analysis is underutilized in health disparities research.
The concepts of DIF and DTF will be reviewed, with an emphasis on the relevance for health services research and on disparities research in particular. Alternative methods for assessing DIF/DTF will be presented and demonstrated. Practical suggestions for investigators will be offered, and discussion will focus on how DIF/DTF assessment can benefit health disparities research. A sample data set and analysis programs will be made available on the METRIC website for researchers wishing to gain experience with DIF/DTF techniques.
This workshop is for health services researchers who are interested in obtaining a basic and conceptual (not technical) understanding of how to evaluate whether individual items and scales exhibit bias across different subpopulations, and hence, whether items and/or scales might be inappropriate for use in disparities research.