CHIS 2001: Assessing English/Spanish Data Quality

• Activity Number: 2
• Type: Invited
• Date/Time: Sunday, August 3, 2003 : 2:00 PM to 3:50 PM
• Sponsor: Social Statistics Section
• Author(s): Elaine Zahnd*+ and Gordon B. Willis and W. Sherman Edwards and Sue Holtby and David Grant and Charles A. DiSogra and Ninez Ponce
• Companies: Public Health Institute and National Cancer Institute and Westat and Public Health Institute and University of California, Los Angeles and University of California, Los Angeles and University of California, Los Angeles
• Keywords: behavior coding ; survey data ; cultural bias ; language bias ; data quality ; qualitative

Abstract

Cultural and language differences can introduce systemic bias in survey data, compromising data quality. To evaluate question performance in the context of these differences across respondent populations, a qualitative assessment of the California Health Interview Survey 2001(CHIS) was undertaken utilizing behavior coding techniques. CHIS is a RDD CATI survey administered to 55,248 adults, 5,801 adolescents, and 12,592 children in 2001. Verbal markers tracked "live" included: mental math problems, qualified responses, clarifications, rating scale problems, inadequate answers, and interviewer probes. A convenience sample totaling 226 interviews was monitored (123 English, 103 Spanish). Overall, survey items performed quite well. Items requiring recall combined with mental math or containing many complex response categories were problematic. Acculturation appeared to affect data quality among recent immigrants unfamiliar with the health care system. Some respondents struggled to fit their identities into the race, ethnicity and nationality categories. Some questions did not perform as well with respondents with lower economic/educational backgrounds.