Can Claims-Based Data Be Used to Recruit Black and Hispanic Subjects into Clinical Trials?

Since 1993, the National Institute of Health (NIH) has required that racial and ethnic minority groups be adequately represented in all NIH (2009) sponsored research. Increasing the participation of racial and ethnic minorities in clinical trials is recognized as a key strategy for the reduction of health disparities (Freedman et al. 1995; Institute of Medicine 1999; Corbie-Smith, Moody-Ayers, and Thrasher 2004). Yet the representation of minorities in clinical trials continues to be suboptimal in many areas of biomedical research (Swanson and Ward 1995; Hall 1999; Heiat, Gross, and Krumholz 2002; Sateren et al. 2002; Swanson and Bailar 2002; Sullivan et al. 2007; Ford et al. 2008; Williams et al. 2010). Reasons for underrepresentation of minorities in clinical trials include lower socioeconomic status, lack of insurance, non-English language, disease burden, and fear of harm (Corbie-Smith et al. 2003; Hussain-Gambles, Atkin, and Leese 2004; Ford et al. 2008). Investigators also commonly cite limited access to minority communities as an additional barrier (Williams and Corbie-Smith 2006). The use of administrative claims data may be one potential approach to rapidly and efficiently identify, and recruit minorities into clinical trials. A major barrier to using such an approach has been incomplete ascertainment of race and ethnicity in private and public datasets (Lauderdale and Goldberg 1996; Arday et al. 2000; National Research Council 2004). However, over the past decade investigators have been able to validate several probabilistic approaches to assign race and ethnicity to large cohorts of patients in such claims data (Fremont et al. 2005; Fiscella and Fremont 2006; Elliott et al. 2008). These existing methodologies include geocoding techniques and the use of Census-based surname lists. Further, the combination of both methodologies, particularly when using a Bayesian approach, has shown to improve the sensitivity and predictive value of either one alone at identifying minority groups (Fiscella and Fremont 2006; Wei et al. 2006; Elliott et al. 2008; Elliott 2009). To date, this approach has been primarily used in observational studies. We hypothesize this approach may also be a powerful tool for recruiting minorities into clinical trials. In this paper, we describe a pilot study that evaluates a three-step algorithm that incorporates these existing methodologies and the Medicare Race Code (MRC) to identify racial/ethnic minority subjects from a large health benefits carrier database. The study was performed in preparation for a NIH-funded randomized clinical trial consisting of a phone-based intervention for which we plan to recruit 250 black and 250 Hispanic subjects having recently undergone a percutaneous coronary intervention (PCIS).