Assessment of Potential Bias in Telephone Survey Estimates due to Noncontact and Noncoverage where Respondents Primarily use Wireless Telephones or do not have Landline Telephones

The quality of survey estimates depends on coverage of the target population, response rates, and measurement errors. Landline telephone surveys exclude households that do not have landline telephones including wireless-only households. Among children under 18 years of age, the prevalence of living in wireless-only households increased from 3.7% in 2003 (January-June) to 19% in 2008 (July-December) according to the National Health Interview Survey (NHIS). In addition, households with mixed telephone service where members primarily use wireless telephones for most calls (wireless-mostly) may not answer their landline telephones which results in survey noncontact. In 2007, wireless-mostly and wireless-only households accounted for 15% and 13% of children under 18 years of age, respectively. In the U.S., the prevalence of the wireless-only and wireless-mostly population has been consistently increasing and may affect the validity of estimates from telephone surveys such as the National Immunization Survey (NIS). Telephone status and interview data for children under 18 years of age from the NHIS-Sample-Child file are analyzed to compare alternative methods to adjust sampling weights and evaluate potential bias in weighted estimates. We used interview data to compare characteristics of children and estimates of selected outcome measures that are associated with health conditions and vaccinations. The purpose of this paper is to evaluate potential bias in RDD survey estimates after adjusting for noncoverage of households that do not have access to landline telephone.