Assessment of Potential Bias in the National Immunization Survey (NIS) Estimates using Data on Telephone Status from the 2007 NHIS

The Quality and validity of survey estimates depend on coverage of the target population and the magnitude of bias in estimates. According to estimates from the National Health Interview Survey (NHIS), wireless substitution is rapidly increasing among adults. The prevalence of wireless-only households with children less than 18 years of age has increased from 3.7% in 2003 (January-June) to 19% in 2008 (July-December). Households with wireless-only telephone service accounted for more than 83% of households without landline telephone service. In addition, households with mixed telephone service where members primarily use wireless telephones for most calls (wireless-mostly) may not respond to a survey if called on their landline telephone. In 2007, wireless-mostly and wireless-only households accounted for 13.5% and 20.1% of children aged 1-4 years, and 16.2% and 8.7% of teens aged 13-17 years, respectively. The increasing prevalence of wireless-only and wireless-mostly households may affect the validity of estimates from telephone surveys such as the National Immunization Survey (NIS). The NIS is a large telephone survey of households with children aged 1935 months followed by a mail survey of providers to obtain vaccination records; the NISTeen uses the same methodology for adolescents aged 13-17 years. Data from households with children aged 1-4 years from the 2007 NHIS are used to evaluate potential noncoverage bias and methods to adjust sampling weights for noncoverage using information from wireless-mostly and wireless-only households. Noncoverage bias is evaluated by comparing characteristics of household members and estimates of selected outcome measures associated with vaccination coverage by type of household telephone service. The purpose of this paper is to continue the evaluation of potential noncoverage bias in the NIS survey estimates due to noncoverage of wireless-only and wireless-mostly households, with emphasis on evaluating weighting methods to reduce non-coverage bias.