Assessing bias in administrative database studies of RotaTeq vaccine completion due to exclusion of subjects with incomplete follow-up.

Abstract

BackgroundRotaTeq® pentavalent human rotavirus vaccine (RV5) is effective against rotavirus illness and rotavirus-related hospitalizations and death. Effectiveness depends on adherence to the dosing schedule, which includes 3 doses at ages 2, 4 and 6 months. Two studies have used automated claims databases to estimate the proportion of vaccinated infants who complete the dosing schedule, but excluded from analysis vaccinated infants who were not enrolled in the database for a sufficient period to observe all 3 doses. Restricting study populations based on duration of follow-up can introduce bias if a large number of subjects are excluded due to insufficient follow-up, and if their outcomes differ from subjects who are included. To address the possibility that exclusions may have been extensive and led to biased estimates of completion rates, we conducted a claims database analysis in the HealthCore Integrated Research DatabaseSM to evaluate the proportion of rotavirus vaccinated infants who completed the 3 dose series of RV5. We evaluated potential error introduced by restricting analyses to infants with complete follow-up by estimating completion rates among infants with complete follow-up, and using Kaplan-Meier analyses to estimate completion rates including infants with incomplete follow-up.ResultsThe inclusion criterion requiring continuous enrollment for the first year of life resulted in only 108,533 (40%) of 233,143 vaccinated infants from 2006–2012 being included in the analysis. After relaxing inclusion criteria, we were able to include 86% of vaccinated infants. The estimated completion rate among infants with continuous enrollment from birth through the first year of life was 78.1% (95% confidence limits [CLs] 77.8%, 78.3%), and among the expanded population the estimated completion rate was 77.4% (95% CLs 77.2%, 77.6%).ConclusionsThese results indicate that most infants were not followed in the database through the first year of life, but the impact of excluding infants with incomplete follow-up was negligible when assessing RV5 completion rates for this commercially insured population. Nonetheless, to increase the size of study populations and reduce the potential for bias, it is preferable to include subjects with incomplete follow-up in automated database analyses, and adopt more robust approaches to defining and analyzing study populations that account for missing data.