Vaccination Trends in Bangladesh and Mozambique: Small area estimation of routine childhood vaccinations 1991-2013

Abstract

<bold>Background </bold>In many countries, vaccine-preventable diseases remain major causes of child mortality. Childhood vaccination is one of the most cost-effective and efficient ways of protecting the health of children across the globe. As countries strive towards universal vaccination coverage, the value of timely, robust and high-quality estimates of vaccination coverage at local-levels becomes increasingly valuable. We produced sub-national estimates of vaccination coverage for BCG, measles and the third dose of oral poliomyelitis (OPV3), and diphtheria-tetanus-pertussis (DTP3) in Bangladesh and Mozambique. <bold>Methods</bold>We estimated childhood vaccination coverage by analyzing unit record data from household surveys. Data was mapped to the lowest administrative unit possible either from GPS coordinates or a survey variable. We used a small area geospatial model with intrinsic conditional autoregressive (ICAR) random effects to model spatial variation in districts and provinces in Mozambique and districts and upazilas in Bangladesh. A spline function was used to model temporal trends, and spatial-temporal interaction terms were specified to capture the variation in local-level patterns of vaccination coverage. Time series were generated for the years 1991-2013 for the 64 districts in Bangladesh and for the period 1997-2008 for 426 upazilas in Bangladesh. Estimates for Mozambique's 11 provinces were produced for the period 1994-2013 and for 148 districts in 2011. <bold>Findings </bold>Vaccination coverage in Bangladesh and Mozambique increased across antigens in the last twenty five years. Nationally, coverage of DTP3 rose from 67.5% (44.3-81.4) in Bangladesh in 1995 to 98.9% (86.1-99.9) in 2013. Mozambique achieved smaller relative growth and at lower absolute levels: DTP3 coverage was 61% (58-64.6) in 1994, increasing to 76.9% (73-80.6) in 2013. The lower administrative level estimates highlight intra-district and intra-province variability that suggest vaccination coverage does not follow administrative boundaries. The disparities between highest and lowest performing upazilas in Bangladesh decreased over time - the mean standard deviation for DTP3 coverage was 15.5% (14.2-17.1) in 1997, shrinking to 6.8% (6.2-7.7) in 2008 as more upazilas achieved higher levels of vaccination coverage. Variation among provinces in Mozambique also decreased - the standard deviation for measles vaccination between provinces was 17.5% (14.8-19.2) in 1995 and remained at 11.6% (9.1-13.7) in 2013. Among districts in Mozambique, full vaccination had the largest standard deviation at 16.0% (14.1-19.5%) in 2009. BCG vaccination coverage was the most equitable with a standard deviation of 10.5% (9.9-11.1). <bold>Interpretation </bold>The sub-national analysis in both countries illustrates a more nuanced story of success and missed opportunities, as well as distinctive geographic patterns to vaccination which are masked at higher administrative levels of analysis. These results should be taken as a call to action for more local-level data collection and data analysis efforts, and should bring attention to sub-national efforts to improve service delivery and access to childhood vaccinations in Bangladesh and Mozambique.