Assessing non-coverage bias among children 19-35 months and adolescents 13-17 years in the National Immunization Survey
Christina Gail Dorell, Centers for Disease Control and Prevention
Background Children 19-35 months and adolescents 13-17 years are targeted by the random-digit-dialing National Immunization Survey (NIS); in 2009, age-eligibility rates were 2.8% and 8.8%, respectively. The proportion of children living in cell-phone only households is 29% and may contribute to non-coverage bias in random-digit-dialing surveys. Methods Nationally-representative samples of telephone numbers in the United States were selected from cell-phone and landline sampling frames from March-July 2009 in the NIS. The cell-phone sample included households with a working cell-phone number used in a cell-phone only/mainly household. Households were screened for children and adolescents. Weighted vaccination rates for the landline sample and a dual-frame landline and cell-phone sample were compared to assess potential non-coverage bias associated with excluding cell-phone only/mainly households. Results The proportion of cell-phone only/mainly households among the cell-phone sample was higher among children (89.4%) than among adolescents (69.7%). The differences between landline only and dual-frame vaccination rates varied, ranging from a modest -1.25- 0.82 percentage points for childhood-administered vaccines among 19-35 month olds; few statistically differences were identified. Among 13-17 year olds, differences between landline and dual-frame vaccination rates were larger, ranging from -14.7- 6.4 percentage points for childhood- and adolescent-administered vaccines. Conclusion Although cell-phone only/mainly households were more common among children than adolescents, excluding cell-phone only/mainly households from the NIS had a larger impact on adolescent vaccination coverage estimates. Future analyses will examine trends in differences between landline and dual-frame vaccination rates among adolescents from 2009-2011. Weighting alternatives will be pursued to address non-coverage bias among adolescent vaccination estimates.