Measuring COVID-19 Vaccination Coverage to Support Healthcare Equity Decision-Making in Urban Areas


 Background: Limited studies have been conducted on access to COVID-19 vaccines and identifying the most appropriate health centres for performing vaccination in metropolitan areas. This study aimed to measure potential spatial access to COVID-19 vaccination centres in Mashhad, the second-most populous city in Iran.Methods: The age structure of the urban census tracts was integrated into the enhanced two-step floating catchment area model to improve accuracy. The model was developed based on three different scenarios: only public hospitals, only public healthcare centres, and the top 20% healthcare centres were employed as potential vaccination facilities. The weighted decision-matrix and analytic hierarchy process based on four criteria (i.e. service area, accessibility index, capacity of vaccination centres, and distance to main roads) were used to choose potential vaccination centres with the highest suitability for residents.Results: Our findings indicate that including the both public hospitals and public healthcare centres can provide high accessibility to vaccination in central parts of the urban areas. However, using only public healthcare centres for vaccination can provide higher accessibility to vaccination sites in the eastern and north-eastern parts of the study area. Therefore, a combination of public hospitals and public healthcare centres is recommended for efficient vaccination coverage.Conclusions: Measuring spatial access to COVID-19 vaccination centres can provide valuable insights for urban public health decision-makers. Our model, coupled with geographical information systems (GIS), provides more efficient vaccination coverage by identifying the most suitable healthcare centres, which is of special importance when only few centres are available.


58
The COVID-19 pandemic has posed substantial costs on individuals and societies, both 59 by direct impact on human physical and mental health, as well as indirectly through economic 60 and social restrictions. Non-pharmaceutical strategies such as social distancing, mask-wearing 61 and economic lockdown are effective control strategies for stop holding up transmission, but 62 notoriously difficult to fully reinforce (1,2). Amazingly, several effective vaccines could be 63 developed, produced and pass regulatory offices in different countries not much more than a year 64 after the first COVID-19 outbreak in Wuhan, China (3). Indeed, large-scale vaccination is 65 considered the best strategy to address this crisis (4) and so far 12 different vaccines have been 66 endorsed for full or restricted use by the World Health Organization (5), and many countries are 67 now striving to vaccinate their residents to reduce the risk. However, not only is vaccine 68 production lagging demand (5), but access to vaccination centres is a hurdle making vaccine 69 delivery a challenge; thus, careful planning is needed to ensure that everyone has appropriate 70 access to vaccination against this new virus. 71 Access to healthcare is a question of the degree of effort needed to reach required 72 medical services (6). It has five primary dimensions: availability, accessibility, accommodation, 73 affordability and acceptability (7,8). While availability refers to the number of available services 74 in the healthcare centres (9), and it is evident that each healthcare facility cannot provide all 75 different services that might be sought, we focussed on accessibility, i.e. the physical distance 76 between healthcare centres and those who might need its services. This is ordinarily given by the 77 length of, and how close it is to, the Euclidean distance (the straight line between 78 source and destination) and must be calculated considering all possible road 79 connections available (10, 11). For example, the drive time from individuals' homes to 80 healthcare centres has also been used in many studies to measure the accessibility dimension (12, 81 13). As accessibility is related to geographical factors it also labelled spatial access with 82 affordability, accommodation and acceptability considered non-spatial access dimensions, while 83 the availability dimension falls somewhere in-between (14, 15). Another classification 84 categorises access into revealed access and potential access, where the former refers to the actual 85 use of services, while the latter is a proxy of the ability of individuals to use these services (16). 86 In this study, potential spatial access (PSA) to health centres indicates the degree of geographical 87 access to them considering both the geographical distance and the capacity to identify areas 88 characterized by poor access to COVID-19 vaccination centres. 89 As shown by our research group previously, the two-step floating catchment area 90 (2SFCA) is a robust methodology to measure PSA to healthcare services (17). The method 91 consists of two major steps. First, it calculates the capacity-to-population ratio for each 92 healthcare location. Second, it sums the ratios for residential sites where healthcare locations 93 overlap. However, the 2SFCA approach has drawn sharp criticism for disregarding the 94 differences in accessibility within catchment areas assuming that all humans located within them 95 have equal accessibility (18). To address this issue, the enhanced 2SFCA (E2SFCA) was 96 developed by Lou and Qi (19) and has been further worked out by assigning geographical 97 weights to both steps of the calculation process, which differentiates the travel-time zones 98 through incorporating distance-decay (20).

99
Limited studies have examined the spatial accessibility of people to COVID-19 100 vaccination facilities. However, the accessibility for vaccination at a centre proposed as a pilot 101 COVID-19 vaccination programme in Hamilton, Ontario, Canada found that the selected sites 102 did not serve the rural and urban residents appropriately; moreover, the associated cost of travel 103 time was anticipated to be disproportionally borne by lower-income urban populations and rural 104 residents (21). Another study conducted in China compared four optimal vaccine distribution 105 scenarios, including random strategy, age strategy, space strategy as well as space and age 106 strategy finding that 30-40% vaccine coverage was needed to control the epidemic under the 107 space and age strategy, while 60-70% vaccine coverage was required for a random strategy (22).

108
A study conducted in the city of Warsaw, Poland, measured spatial access to COVID-19 109 vaccination sites using Thiessen polygons (also known as Voronoi polygons) (23). They 110 identified spatial inequalities and areas with poor access to vaccination sites and proposed 111 activating additional sites either located ad hoc or using mobile vaccination sites to achieve 112 uniform vaccination coverage. Importantly, the elderly population was found to be a significant 113 variable in their analysis (23). A study in Florida, USA, evaluated the spatial accessibility to 114 COVID-19 testing sites using the 2SFCA method by integrating both driving and walking 115 modes. Their results suggest that increased efforts are needed to improve accessibility to testing 116 sites among the elderly and those without private vehicles (24). Another Florida study assessed 117 the spatial accessibility of COVID-19 patients to intensive care unit (ICU) beds, using both the 118 2SFCA and the E2SFCA methods, developed an accessibility ratio difference index to evaluate 119 the difference between the models based on spatial access (25). They found that the 2SFCA 120 method overestimates the accessibility in areas with a lower number of ICU beds due to the 121 "equal access" assumption of the population within the catchment area (CA) (25).

122
A study in Brazil measured the geographic access to COVID-19 healthcare services using 123 a balanced float catchment area approach and identified substantial social and spatial inequalities 124 in access to health services during the pandemic (26). Their findings moreover indicated that 125 ICU equipment availability varied considerably between cities and was substantially lower European countries using a regional ratio of ICU beds to 100,000 population as the accessibility 128 index and the distance to the closest ICU and arrived at high indices in Germany, Estonia and 129 Austria, with the lowest in Sweden and Denmark. Importantly, the study identified a negative  With respect to COVID-19 vaccination, it is vital to prioritise the elderly population as it 141 has been shown that higher age increases the risk of mortality (31). In this study, we measured 142 the PSA to vaccination centres by developing a modified version of the E2SFCA model using a    To measure the PSA to vaccination centres, the age structure of each CT was integrated 173 into the E2SFCA method as an influential factor of COVID-19 mortality. We weighted each age where Pop x denotes population for age group x The CA is the basis of the E2SFCA method (19). According to previous studies (20, 34-  In step 1, the CAs were set at 1, 1.5, and 2-km distance to the j th healthcare location. We ratio R j within the CAs using Eq. 2 below following previous studies (11, 24, 41, 42).

193
Eq. 2 where P k is the population of the k th CT falling within the CA j (d kj ∈D r ); S j the vaccination Eq. 3 represents the accessibility vaccination centre for the population at location i; R j the 207 vaccination capacity-to-weighted population ratio at healthcare centre j that falls within the CA  The number of PHCs is almost 10 times higher than that of hospitals, and they are well-223 dispersed across the city. Therefore, the PHCs have a stronger potential when acting as 224 vaccination centres.  PVCs when applying buffer analysis in QGIS. The PVCs within the distance buffers were then 267 assigned weights, and the PVCs outside these buffer zones with no access to main roads 268 (especially for public transport users) were excluded from the analysis (Fig. 2-C). In this study, PHs and then PHCs were ranked according to their capacity, so that centres 271 with higher capacities were given a higher chance of selection. At the same time, the centres 272 without proper facilities and equipment (e.g., centres with two vaccinators or less) were not 273 given priority status (Fig. 2-D). After selecting the PVCs, the PSAs to these facilities were measured using the E2SFCA 291 methodology as described.

324
Preliminary results indicated that 864 CTs (66.4%) had low access to PHs (PSA 325 <0.000072) (Fig. 3-A). Also, those CTs with above-average access to PVCs are often located in     Table 1 shows the summary statistics for spatial autocorrelation for the three scenarios     There are several limitations in this study that are mainly associated with the data quality. poor areas of the city had the least access to PVCs. Therefore, due to the large size of the study 479 area and as it is common for people with lower socio-economic status to commute using public 480 transportations, it is suggested to provide vaccination services in neighbourhoods with better 481 access to public transportation.

482
The spatial accessibility models can measure the accessibility to potential vaccination 483 services so that all individuals would have adequate and equitable access to COVID-19 484 vaccination services. We found that using urban indicators in selecting the most appropriate 485 health facilities can help policymakers improve the accessibility to COVID-19 vaccination 486 services in a cost-effective and timely fashion. In addition, the proposed approach in this study 487 can be easily automated and broadly applied to various urban settings.