print share

CHIS Data Quality

Assessing and Addressing Potential Noncoverage Bias

The rapid growth of cellular telephone use over the past decade may create coverage problems for CHIS and other telephone surveys. That's because CHIS has relied mainly on the widespread practice of selecting households through random digit dial (RDD) sampling, which historically has not included cell phone numbers. Cell phones, per se, are not problematic, but generate two issues that may lead to noncoverage bias in telephone surveys. First, an increasing proportion of cell phone users live in households that do not also have a landline phone. Second, traditional RDD sampling frames are limited to landline telephone numbers and thus exclude cell-phone-only households. To the extent that persons who live in cell-phone-only households differ with respect to health and health-related characteristics from persons who live in households with landline telephones, then noncoverage bias may result.

To assess noncoverage bias in CHIS 2007, the landline RDD sample was supplemented with a sample of adults living in households with only cell phones. For the cell-phone-only sample, a sample of telephone numbers designated for cellular use was drawn and screened; only cell phone users that did not have a landline telephone at home were eligible to complete the adult survey.

Noncoverage bias was assessed by comparing over 40 demographic and health-related characteristics among the landline RDD sample alone with the landline sample combined with the cell-phone only sample. Relative to the landline sample, the cell-phone only sample was younger, less likely to be married or own their home, and more likely to be male, employed, and living in a single-person household. The landline sample showed lower rates of risky health behaviors and better health care utilization and health insurance coverage. Once the estimates were properly weighted, noncoverage bias was substantially reduced. (The results of this study have been submitted to a peer-review journal and will be made available on this web site when the paper is published.)

CHIS 2007 data estimates are based on both the landline and the cell-phone-only samples. CHIS 2009 and subsequent CHIS surveys will include larger samples of cell phone households.

See further documentation on this