Background: The current study considers potential nonresponse bias and data quality issues in the Multiethnic Cohort, a prospective study of lifestyle-cancer associations among adults ages 45 to 75 years from five ethnic groups in Hawaii and California.
Methods: We examined determinants of early versus later response to the baseline questionnaire using logistic regression with response wave regressed on measures of demographics, history of illness, health behaviors, medication, and supplement use.
Results: Participants who were more health conscious tended to respond earlier. Female sex, more education, personal experience with cancer, more physical activity, and regular use of aspirin were associated with early response. Race other than white and current smoking were associated with later response. Of note, African-Americans and Latinos, with lower response rates, and Japanese, with the highest response rate overall, were more likely to respond to a later mailing. Results were generally consistent across sex, age groups (under 65, 65+), and ethnic groups and over time. Although repeated mailings increased the proportion responding and the diversity of participants, the percent of missing item responses increased somewhat with response wave.
Conclusion: Multiple mailings are recommended for recruiting ethnic minority groups but the tradeoff may be more missing data among later respondents.