Estimation of Small Area Child Immunisation Coverage in Rural Nigeria Using Fay-Herriot Model

Direct survey estimates of child immunisation coverage for small geographic areas such as rural local government areas in Nigeria have large sampling errors that limit their utility for planning, necessitating model-based small area estimation methods that borrow strength from auxiliary data. This study applies the Fay-Herriot small area estimation model to estimate LGA-level child immunisation coverage in rural Nigeria using the 2021 MICS6 survey data as the direct survey input and administrative health facility and vaccination coverage data as auxiliary information. Fay-Herriot empirical best linear unbiased predictors were computed for 180 rural LGAs in 10 states across the North West, North East, and North Central zones. Variance smoothing using the generalised additive model addressed zero-direct-estimate LGAs. Mean squared error of small area estimates was estimated by parametric bootstrap. Direct survey estimates had coefficients of variation exceeding 30 percent in 67 percent of LGAs, confirming their unreliability for planning. Fay-Herriot model-based estimates reduced mean coefficient of variation from 38 to 14 percent. Estimated LGA immunisation coverage ranged from 11 to 78 percent across the 180 LGAs. Seventeen LGAs showed less than 20 percent estimated coverage, requiring urgent immunisation programme attention. Fay-Herriot estimates were significantly more stable than direct estimates in repeated resampling validation. The study provides UNICEF Nigeria and the NPHCDA with a statistically validated small area estimation framework for immunisation monitoring at LGA level. Keywords: small area estimation, Fay-Herriot model, immunisation coverage, rural Nigeria, MICS survey