Using Administrative Data to Identify Spanish-Preferring Seniors
Spanish-preferring seniors with limited English proficiency (SPLEP) are a particularly vulnerable group, receiving poorer care than others seniors, including immunization for influenza and pneumonia at half the rates of English-speaking non-Hispanic Whites. In addition to the linguistic and health literacy barriers they often face in receiving care, they are underrepresented in surveys of patient experience, making it difficult to assess their experiences and needs. CMS Medicare records contain information regarding beneficiary surname, address, insurance coverage type, and limited demographic information, including a race/ethnicity variable know to have less than 50% sensitivity for Hispanics. Residential address was geocoded and surnames were matched against a recently constructed surname list (Word 2007); these sources of information were combined to produce Bayesian posterior probabilities of race/ethnicity (Elliott et al. 2009). The 2000 Census proportion of SPLEP in the geocoded residential block was derived. Logistic regression predicted having responded in Spanish (by request) among 257,518 respondents to a 2008 national survey of Medicare beneficiaries. This model predicted a preference for Spanish response with a concordance (area under the curve) of 97%, using SPLEP%, Bayesian predicted probability Hispanic, income<150% FPL, having managed care coverage, and evidence against being Black or Asian being important predictors. This model can be used to estimate disparities for the Spanish-preferring population in the absence of self-reported information, to target interventions, and to direct appropriate survey resources.