Adjusting for Selection Bias in Evaluating Two-Dose Human Papillomavirus Vaccine Coverage Among Adolescents in the United States

Abstract

Recent changes to the recommended human papillomavirus (HPV) vaccine schedule include reduction from a 3- to 2-dose schedule (if initiated before age 15) and inclusion of a 9-valent vaccine (HPV9). To assess potential impact on vaccine coverage, we examined 2-dose coverage and timing among adolescents and explored the use of survival analysis methods to estimate coverage and identify factors associated with receipt of the second dose. Data from the 2015 National Immunization Survey-Teen were analyzed. Among 13-17 year-olds who initiated the series before age 15, the time to the second dose and factors associated with receipt of the second dose were estimated with and without adjusting for follow-up time. Among those who initiated before age 15, 88% initiated with the quadrivalent vaccine, 59% initiated at age 11-12 years, and 84% received a second dose. Among adolescents who received 2+ doses, median time between the first two doses was 3.0 months. Kaplan-Meier estimated median time between doses was 3.9 months. Logistic regression results indicated adolescents who initiated with HPV9 were less likely to receive a second dose (OR: 0.4, P < 0.0001); however, this was not observed in the Cox model (HR: 1.1, P = 0.5). Initiation at age 9-10 was associated with a greater likelihood of receiving a second dose (OR: 6.1, P < 0.0001). The effect was substantially smaller in the Cox model (HR: 1.2, P = 0.025). Standard methods for examining dose timing do not account for follow-up time, resulting in underestimation of dose timing and overestimation of the effect of vaccine type and age on receipt of a second dose. Substantial selection bias could affect individuals who initiated with HPV9 due to the limited duration of follow-up time. This research highlights the importance of using methods that account for variable follow-up time.