Infectious Disease
Bias in rotavirus vaccine effectiveness estimates due to missing vaccination cards in a Uganda case-control evaluation Jordan Cates* Jordan Cates Hellen Aanyu-Tukamuhebwa Julia Baker Cissy Nalunkuma Eleanor Burnett Kaudha Elizabeth Annet Kisakye Goitom Gebremedhin Weldegebriel Jacqueline Tate Jason M. Mwenda Paul Gastañaduy
Background: Card-confirmed vaccination status is essential to estimate vaccine effectiveness (VE) accurately, yet card ascertainment is often incomplete, especially in low-resource settings. We aimed to quantify the potential bias in rotavirus VE estimates due to exclusion of children without card-confirmed vaccination status.
Methods: Children aged 4-11 months were enrolled in a hospital-based test-negative case-control evaluation of rotavirus VE in Uganda from July 2018- December 2022. Logistic regression was used to estimate the odds ratio (OR) of at least one dose of rotavirus vaccine between cases (children with laboratory-confirmed rotavirus diarrhea) and non-rotavirus “test-negative” diarrhea controls, adjusted for age, season, and site. VE was calculated as 100*(1-adjusted OR). VE restricted to children with card-confirmed vaccination status was compared to VE when imputing the vaccination status for children without a vaccine card. Multiple imputation was performed assuming a multivariate normal distribution with 40 imputations, including variables for rotavirus test result, receipt of rotavirus vaccine, and demographics associated with missing vaccine cards.
Results: Among 186 cases and 632 controls, 48% and 44% were missing vaccine cards, respectively. Among 96 cases and 352 controls with card-ascertained vaccination status, 92% and 94% received at least one dose of rotavirus vaccine, respectively. Among 90 cases and 280 controls with imputed vaccination status, an estimated 88% and 93% were vaccinated. Among children with card-confirmed vaccination, the adjusted VE was 56% (95% confidence interval [CI]: -10%, 82%). In the multiple imputation analysis, the adjusted VE was 50% (95% CI: -25%, 80%).
Conclusions: Rotavirus VE restricted to card-confirmed children was slightly higher compared to multiple imputation. This illustrates the use of multiple imputation to assess the magnitude and direction of potential selection bias due to low card ascertainment.