This study highlights that two in five adult persons may be marginalized from existing COVID vaccination facilities, living more than 1 hour from the nearest facility. Only five of the 47 counties have more than 80% coverage, with four counties having less than 20% of their adult populations within an hour of the nearest vaccination facility. The country has made an ambitious plan of ensuring all adult populations are fully vaccinated against COVID by end of 2022 and using data driven decision making will be critical in ensuring universal access. In addition, as the country aims to use multi-pronged interventions such as community vaccination campaigns in populated areas like markets and spatial outputs presented in this study can be useful in designing which strategy works best.
This study highlights that 40% of the county’s population lives more than 1 hour from the nearest vaccination facility. To our knowledge, this is the first assessment of geographic access to COVID vaccination sites and outputs can be validated with per capita vaccination use and can be useful for monitoring distribution systems. For example, counties with high coverage, but poor uptake can be prioritised for heightened community engagement to increase uptake. Marginalization can also be extracted at geographic units smaller than counties for adequate intervention. Thus, we recommend the need to consistently monitor access and coverage of vaccine administration sites to ensure universal access. In this analysis, we used freely available geospatial datasets that would allow scaling up such a similar analysis to other LMICs.
The importance of measuring geographic accessibility is in highlighting where interventions are needed. These include where to locate new facilities or improving road networks. Here, we use geospatial tools to propose potential facilities whose cold chain and human resources can be improved to increase vaccination coverage. Using 80% as a near-term goal, our modelling approach identified 560 additional facilities based on population density, that can be optimised to improve access. Our results also highlight that primary care facilities will be critical in improving coverage. However, given that improving immunization to all facilities in the country would still leave 1 in 10 adult persons marginalized, then other methods of improving coverage such as community campaigns and mobile clinics can be emphasized in areas of low population density to improve coverage. Furthermore, the level of selection can be used to identify which level of facility can be selected to ensure maximum benefit. For example, the one large facility can be used to select a hospital (level 4 to 6) in its closest proximity, while small facilities selected can be used to select dispensaries in their close proximities.
This study had several limitations. First is that geographic access may not necessarily translate to uptake, and as such increased coverage must be accompanied with adequate community communication strategies on the importance of getting vaccinated. Secondly, travel time may be affected by other transport factors such as weather, waiting times and different travel speeds. Thirdly, some of the datasets used such as population distribution may have errors which can be propagated to the results of accessibility. Capacity can also be more fluid than expressed here. In addition, cost implications of transport should be considered. We were also unable to geolocate all the 7,877 health facilities intended to be used to improve coverage and this would have affected the final access metrics.
Nonetheless, our aim was to provide a subnational comparison and such outputs are still important in identifying areas of poor coverage. Although we hard bound for selecting access, our aim is proof of concept, and methodology can be adapted at national and subnational levels at different stages of expanding vaccine coverage. We therefore recommend the need for adapting local parameters for defining geographic accessibility, measures of capacity, and validating these validated with partners working in communities. The process also used AccessMod version 5.0, a WHO tool for measuring geographic access with free to use data meaning it can easily be scaled up and adapted to different contexts and scenarios to aid in decision making.