Comparing dual phone type (landline and cell) respondents from cell sample versus landline sample in Oregon's 2011 BRFSS

Introduction: In 2009, CDC BRFSS phone survey methods included landline and cell samples. Cell sample respondents were eligible only if they had no landline. In 2011, Oregon adapted methods to include single and dual-phone-type respondents from both samples -- about 1,000 in the cell and 4,000 in the landline sample. Oregon's reasons for expanding respondent eligibility in the cell sample were to: a) increase sample efficiency and b) study potential differences between dual-phone-type respondents from the cell and landline samples. This presentation describes characteristics of Oregon's 2011 dual-phone-type respondents from cell and landline samples. Specifically, we were interested in determining potential for bias due to undercoverage of cell phone respondents who also have landlines in their household. Would they prove to be different from dual-phone-type respondents reached by landline for the interview? Methods: Using 2011 Oregon BRFSS data, we conducted descriptive and regression analyses to identify potential differences between dual-phone-type respondents from the two sample types. We modified BRFSS raked weighting procedures for phone type using state estimates from NCHS. We also examined methods for dealing with potential overlap of sample (between landline and cell samples). Results: We discuss analyses by sample type, including key weighting factors as well as key health indicators, to examine whether cell sample dual-phone-type respondents are different from their landline sample counterparts. Analyses include identification of cell-mostly, split or dual-use, and landline-mostly respondents. Discussion focus: Including dual-phone-type respondents from cell sample may be useful both in increasing sample efficiency and in reducing potential for bias.