We constructed datasets for vaccination policy and doses administered as of 8 February 2022 that contain country-specific metrics. We used these metrics to develop three coverage indicators and estimated demand for vaccine doses. Detailed descriptions of data collection methods and data sources are provided in Additional file 1.
Metrics
Policy metrics include authorization status of COVID-19 vaccines (approved, conditionally approved, or authorized for emergency use), vaccination schedules, indicated age groups, contraindications, whether vaccines are sold/donated by or received by the country, and whether local residents need to pay for vaccine (Additional file 1: Table S1-S4, Fig. S1). Vaccination schedules mainly include dose intervals according to regulatory authorities, and special doses, such as additional doses for immunocompromised individuals and booster doses. Contraindications are medical conditions for which individuals should not be vaccinated, either temporarily for conditions that resolve or permanently.
The dose-administered dataset included four time-varying, age-specific, platform-specific metrics: number of vaccine doses administered, number of people receiving at least one dose, number of people fully vaccinated according to country schedule, and number of people receiving additional/booster doses. People receiving at least one dose are those who have received the first dose of a multiple-dose vaccine or the dose of a one-dose vaccine. The number of people fully vaccinated indicates those who have completed their primary immunization series according to their country’s vaccination schedule. Additional doses are doses for people with medical conditions requiring doses beyond the primary series to achieve immunity (e.g., moderately to severely immunocompromised individuals); booster doses are doses given after the primary series to counter waning immunity or decreasing protection [16].
Target populations
Target populations are eligible individuals who are within approved age groups and without contraindications according to the country’s immunization policy. When country-specific information was unavailable, we used the World Health Organization (WHO)-recommended ages (as of 8 February 2022), in which people 5 years and older are recommended to receive BNT162b2 vaccine and people 18 years and above are recommended to receive any COVID-19 vaccine [17]. Contraindications vary by country and may include pregnant women, people with certain underlying conditions (i.e., bleeding disorders and immune suppression), and/or those previously infected with SARS-CoV-2 (Additional file 1: Table S2, Table S5) [1, 18,19,20,21]. We estimated sizes of target populations for primary, additional, and booster doses according to country-specific policy (Additional file 1: Table S6). To estimate the number of individuals with bleeding disorders or immune suppression, we adapted Clark’s method to adjust for effects of clustering and multimorbidity using data from the Global Burden of Diseases [21] and two multi-morbidity studies [22, 23], considering that one person can simultaneously suffer more than one sub-category of such a disease [20]. Methods for estimating target population sizes are detailed in Additional file 1.
Vaccine coverage
We constructed vaccine coverage indicators for both total populations and target populations. The indicators are full vaccination coverage, proportion receiving at least one dose, and doses administered per 100 people. Because not all countries report vaccine doses administered on a daily basis, we used linear extrapolation between reported data points for missing internal data points. For example, if vaccination volumes on March 1 and March 3 were reported but no volume was reported for March 2, we assumed that the vaccination volume on March 2 was midway between the March 1 and March 3 values. We estimated vaccine coverage through 31 January 2022. For countries that did not report data to 31 January 2022, we used latest data reported on or after 17 January 2022 as the 31 January value, but if no data were reported on/after 17 January, we indicate missing values for 31 January.
Exploring factors associated with coverage
We identified a list of economic, social, health spending, health resources-related factors that are known or believed to be associated with vaccine coverage (Additional file 1: Table S7) [19, 24,25,26,27,28], including socio-demographic index (SDI), healthcare access and quality (HAQ) Index, GDP per capita adjusted for purchasing power parity (PPP), physician density, as well as PPP-adjusted government health spending per capita. SDI is a composite indicator developed by the Institute for Health Metrics and Evaluation, which reflects a country’s socio-demographic level and was proven to correlate highly with health outcome variables [29]. Through calculating geometric means of lag-distributed income per capita, average education level, and fertility rate under 25 years, values range from 0 to 1 and can be divided into five categories: high, high-middle, middle, low-middle, and low SDI [30]. HAQ is a country-specific index that quantifies the accessibility and quality of personal healthcare, ranging from 0 to 100 [25]; physician density mirrors the capacity of healthcare services [19, 27]. PPP-adjusted GDP per capita [26] and government health spending per capita [28] reflect a country’s overall economic and health-related economic level, respectively.
Demand
We determined demand for vaccine as the number of doses needed to complete vaccination of countries’ target populations according to national immunization program policy. We estimated demand for primary immunization and additional/booster doses separately. We used a simplifying assumption that all COVID-19 vaccines primary series require two doses, as one-dose (Ad26.COV2.S, Ad5-nCoV) and three-dose (ZF2001, CIGB-66) vaccines are thus far a small (though unknown) proportion of total doses administered. We estimated additional and booster doses needed based on sizes of certain populations allowed by regulatory agencies; we assumed one additional/booster dose per person recommended or indicated. We calculated total current demand as the sum of doses required minus doses administered by 31 January 2022, stratified by primary and additional/booster dose demand.
Data sources
In priority order, we obtained data from government websites, health department websites, official media, vaccine manufacturers’ websites, authoritative media, and local media. We also used data from public databases that systematically collect and cross-check such information, such as Gavi-COVAX [31], COVID-19 Vaccine Market Dashboard [32], and Our World in Data [33]. Whether countries sell/donate or receive vaccine was obtained from COVAX data [31].
We collected raw data through a combination of manual and automated means. For countries that released data on a regular basis through official sources [34, 35], we collected information manually on a weekly basis; for countries that released official data through a public online dataset (e.g., GitHub repository) [36], we accessed data with a compiled R Script. Countries were included in our doses administered datasets if they reported at least two data points. We verified data transcription accuracy with double-entry.
We derived 2020 population estimates from United Nations World Population Prospects [37]. Among 194 WHO Member States, age-specific United Nations population proportions were available for 183 countries. Age-specific population estimates for the 11 remaining countries (Andorra, Cook Islands, Dominica, Saint Kitts and Nevis, Monaco, Marshall Islands, Niue, Nauru, Palau, San Marino, and Tuvalu) were compiled from WorldPop datasets [38].
Statistical analysis
We explored univariate associations of coverage with SDI, HAQ Index, PPP-adjusted GDP per capita, physician density, and PPP-adjusted government health spending per capita. Multicollinearity of those covariates was accessed by using the variance inflation factor (VIF) in linear model; we did not perform multivariable modeling considering that most variables were collinear (VIF > 5) (Additional file 1: Table S8). Instead, we individually investigated univariate associations of each selected variable with vaccine coverage. We used a linear or non-linear regression model to explore the relationship between vaccine coverage and a series of variables. We chose a linear regression model if a clear linear relationship was observed in scatter plots and the standard residuals were uniformly distributed around zero. Otherwise, we used non-linear regression to fit. For regressing vaccine coverage with PPP-adjusted GDP per capita and government health spending per capita, since there were clear non-linear, logarithmic relationships observed through scatter plots, we used a nonlinear self-starting regression model (logistic growth model) to determine adjusted relations. Model selection was based on scatter plots, regression coefficients, and residual plots. Statistical analyses and visualizations were done using R (version 4.0.2).