Bias in Telephone Surveys that Do Not Sample Cell Phones

An exploration of whether poststratification of data can reduce the bias inherent in failing to reach cell phone-only households (CPOH) in landline health surveys determined that poststratifying data is not always enough to remove bias.

Traditional random digit dialing can systematically exclude potential respondents who only use cell phones. The authors examined data from the 2008 National Health Interview Survey (NHIS) to determine whether people in CPOH differ from people not in CPOH in health care access and health behaviors, comparing NHIS estimates to reweighted estimates that excluded people in CPOH.

Removing CPOH from the NHIS data affects all health-related estimates. Poststratification reduced bias for the non-elderly, but does not reduce bias for Hispanics, Blacks or young adults. Bias reduction was greatest for estimates of uninsurance, and least for estimates of drinking, smoking and care that was delayed or neglected because of inability to pay. A lack of a usual source of care consistently had the largest bias.

Poststratifying data that excludes CPOH can reduce bias for estimates of health insurance among the total population, but does not effectively reduce bias for estimates of health behaviors and access to care. Nor does poststratification effectively reduce bias among subpopulations. Researchers must find ways to sample CPOH as cell phone use becomes more prevalent.