A comparative analysis of supply chain factors impacting COVID-19 vaccine security in high-income countries (HICs) and low-income and middle-income countries (LMICs)

Study countries

Using the four-stage vaccine supply chain framework proposed by Duijzer et al,5 we sought to perform a comparative analysis of supply chain factors in a purposive sample of five HICs, namely USA, Israel, Canada, UK, Japan and five LMICs, namely Brazil, Peru, Vietnam, India and Guinea. Our sample was chosen after a knowledge synthesis to account for (1) heterogeneity in predictor supply chain variable capabilities, (2) geographical variety and (3) adequacy of publicly available critical data elements. Where needed, the LMICs were further subcategorised into upper middle-income countries (UMICs), that is, Brazil and Peru, lower middle-income countries, that is, Vietnam and India, and low-income countries, that is, Guinea. We used 2021 World Bank country classifications by income level that are based on gross national income (GNI) per capita in current US dollars (USD).10

Data collection

All data for this study were obtained from the following major publicly available databases: Our World In Data,11 COVID-19 Vaccine Tracker,12 COVID-19 Market Dashboard,13 World Bank’s World Integrated Trade Solution (WITS) software,14 Launch and Scale Speedometer,15 Duke Global Health Innovation Center16 and World Bank Total Population Statistics.17 These databases had comprehensive data at multiple timepoints for all the measures described below except COVID-19-specific vaccine production capacity. Raw data was downloaded from these databases by one author (MP) and corroborated by the remaining authors. Specific search terms used for each data extraction and data analysis methods are described in detail in the sections below. As there were no publicly available databases with comprehensive data on COVID-19 vaccine production capacity at multiple timepoints, we obtained data from Statista18 and Global Commission for Post-Pandemic Policy19 reports.

Measures

Data for national vaccination rates was from Our World In Data using their ‘Data explorer’ with the metric ‘People vaccinated’ and the interval ‘Cumulative’.11 For this study, the percentage of a country’s population that received one dose of any COVID-19 vaccine at or before two timepoints, 31 March 2021 and 30 April 2021, was taken as a proxy for vaccine security. These timepoints were chosen after an analysis of national vaccination rates in the 6 months post emergency use authorisation (December 2020 to May 2021) revealed wide variation between the countries in our sample.11 12 We focused on earlier timepoints as vaccine availability was more likely to be the main health systems bottleneck to vaccine security, prior to downstream Tanahashi factors, such as accessibility and acceptability becoming barriers to vaccine security. As our study focused on the ‘high demand and low supply’ early phase of the pandemic described by Bongers et al,20 our observations are most applicable to ‘initial vaccine security’, and may not be extrapolated to the later ‘low demand, high supply’ phase of the pandemic when factors such as hesitancy significantly impacted vaccine security.

To determine the best supply chain predictor variables, we studied two factors at each of the four stages of the vaccine supply chain. To characterise the product stage, that is, the ability of countries to develop a novel COVID-19 vaccine, we first retrospectively examined the number of COVID-19 vaccine makers in each country as of June 2023. Data was derived from the COVID-19 Market Dashboard using the tabs ‘Vaccines’, ‘Products’ and ‘By developer country’.13 Both vaccine products ‘approved for use’ and at ‘all development stages’ were compared. Given heterogeneity in vaccine platforms and the emergence of new vaccine development partnerships during the pandemic, we postulated that countries with a greater number of vaccine makers in June 2023 would have had a more robust domestic vaccine development programme at the outset of the pandemic. Next, we examined global vaccine trade data from the World Bank’s WITS software prospectively to determine if countries were net importers or exporters of vaccines in 2019, just prior to the pandemic.14 We theorised that net vaccine exporters would possess greater ‘know-how’ prior to the pandemic and have increased vaccine development capacity.

Given limited public data on vaccine production capacity, we assessed COVID-19 vaccine specific manufacturing capacity by comparing the number of COVID-19 vaccine doses produced in each country as of 3 March 2021. Data was from Statista18 and the Global Commission for Post-Pandemic Policy.19 To our knowledge, there are no public sources with this data as of 31 March and 30 April 2021. In addition, we compared the number of unique COVID-19 vaccine manufacturing facilities (any stage of production) in each country as of 30 June 2022 from the Launch and Scale Speedometer using the tab ‘All vaccines’.15 This was corroborated with data from 30 June 2023 from the COVID-19 Market Dashboard using the tabs ‘Vaccines’, ‘Capacity’ and filtering by ‘Production locations’.13 Once again, we are unaware of any public data sources as of 31 March and 30 April 2021.

To understand how manufacturers chose to allocate their limited initial supply of COVID-19 vaccines, we examined the number of vaccine purchases by each nation via bilateral agreements and multilateral partnerships as of 31 March 2021. As dose regimens differed by vaccine platform, we divided the total number of doses purchased by the total population of each country to establish AMCs as a percentage of population (AMC %). Data was derived from the Duke Global Health Innovation Center using the selector ‘March 2021’16 and from World Bank Total Population Statistics.17 In addition, we examined COVID-19 vaccine price data in USD to understand if paying more led to greater vaccine security. Mean price data as of 30 June 2023 was from the COVID-19 Market Dashboard using the tabs ‘Vaccines’, ‘Price’ and filtering by ‘Country/Group’.13 Given attempts by both manufacturers and countries to closely guard prices, this source only contained data for 7 of the 10 countries in our sample. We corroborated this data with mean prices paid by HICs and LMICs as of 31 March 2022 from WHO.21

To quantify external vaccine distribution from manufacturer to purchaser, we divided the total vaccines delivered to each country as of 31 March 2021 by the total population of each country (Delivery %). Data was from the COVID-19 Market Dashboard using the tabs ‘Vaccines’, ‘Delivery’ and filtering by ‘Month’ for March 2021, and then again by ‘Country’.13 As a proxy for internal distribution, we studied the impact of the domestic public health response, as reflected by a Stringency Index as of 31 March 2021, on vaccine security. We used the Stringency Index from Our World In Data via their ‘Data explorer’ with the metric ‘Stringency Index’, the filter ‘Country’ and the interval ‘31 March 2021’.11

Data analysis

We used basic descriptive statistics to characterise the data obtained for each variable in our sample. To compare data between HICs and LMICs, we used basic statistics (two-sided t-tests). To explore the relationship between each supply chain predictor variable and our outcome variable of vaccine security, we used basic scatter plots. We compared at least two variables for each stage of the vaccine supply chain and used lines of best fit to visualise any relationship with vaccine security. All data analysis was conducted using Minitab V.21.4.1.

Patient and public involvement

Our study was conducted by analysing publicly available aggregate data. As such, patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this research.

Vaccine security

Our analysis of vaccine security reveals wide variation between countries, with mean vaccine security at 16.3% (95% CI 1.4% to 31.2%) at the end of March and 21.7% (95% CI 5.5% to 38.0%) at the end of April. Mean vaccine security at the end of March among HICs was 29.7%, while among LMICs, it was 2.9%, resulting in a difference of 26.8% (95% CI 1.8% to 55.3%, p=0.06). Mean vaccine security at the end of April among HICs was 37.9%, while among LMICs, it was 5.6%, resulting in a difference of 32.2% (95% CI 4.2% to 60.3%, p=0.03). While the difference between HICs and LMICs improved proportionately at the end of April, the gap remained wide.

Figure 1A shows vaccine security against GNI per capita. In general, it appears that countries with greater GNI per capita had stronger vaccine security. However, we report wide variation in vaccine security among HICs, with Israel and the UK overperforming, and Canada and Japan underperforming. Interestingly, as shown in figure 1B, there appears to be an even stronger association between vaccine security and GNI per capita when data for all nations are compared categorically by income groupings.

Request permission

Product

The mean number of vaccine products at all stages of development among HICs was 44.0 and among LMICs was 11.8, resulting in a difference of 32.2 (95% CI −113.2 to 48.8, p=0.33). Figure 2A shows vaccine security against number of vaccine products at all stages of development. There appears to be no correlation between vaccine security and vaccine products (all stages). This lack of relationship is even more apparent when the USA is excluded from analysis. Once again, Israel and the UK overperform, and Japan underperforms relative to its HIC peers.

Request permission

The mean number of vaccine products developed by HICs that received regulatory approval was 2.6, and by LMICs was 1.4, resulting in a difference of 1.2 (95% CI −7.45 to 5.05, p=0.66). Figure 2B shows vaccine security against number of vaccine products that received approval. Again, there appears to be no clear relationship between vaccine security and number of novel approved domestic vaccine products. While the USA had more vaccine products approved than any other country, Israel was able to achieve a high level of vaccine security with no domestic products that received approval.

Figure 2C shows the relationship between vaccine security and net vaccine exporter status in 2019, prior to the onset of the pandemic. Our analysis does not show a significant difference in net vaccine exporter status between HICs and LMICs (95% CI −0.95 to 0.55, p=0.55). Again, there appears to be no clear relationship between vaccine security and net vaccine exporter status. While India’s overperformance on vaccine security relative to other LMICs as a net exporter stands out, we note that Israel was a net importer that still achieved a high degree of vaccine security.

Production

Figure 3A shows the relationship between vaccine security and number of COVID-19 doses produced domestically in early March 2021. Interestingly, there does not appear to be any clear correlation between vaccine security and domestic vaccine production. This lack of association is even more obvious when the USA is excluded from analysis. Furthermore, there is no significant difference in domestic vaccine production between HICs and LMICs (difference of 14 522 000 doses, with 95% CI −70 657 888 to 41 613 888, p=0.54).

Request permission

In figure 3B, we examine the relationship between vaccine security and number of COVID-19 vaccine manufacturing facilities in each country in June 2022. Again, there appears to be no correlation between these variables. HICs had a mean of 9.4 COVID-19 manufacturing facilities and LMICs had a mean of 4.8, with no significant difference (3.6, 95% CI −20.8 to 13.5, p=0.63). This lack of association between vaccine security and the number of COVID-19 vaccine manufacturing facilities in each country was also evident in June 2023 data obtained from a different public source.13

Allocation

On average, the countries we studied secured AMCs to cover 265% of their population with one dose of the COVID-19 vaccine by the end of March 2021 (95% CI 89.8% to 440.2%). HICs secured significantly more doses than LMICs (difference of 336%, with 95% CI 45% to 628%, p=0.03). As shown in figure 4A, there appears to be a modest correlation between vaccine security and AMCs secured. Among HICs, Canada and Israel were significant outliers, with Canada’s vaccine security lower than expected given its high AMC (%), and Israel having high vaccine security despite a modest AMC (%). In fact, if Canada were excluded from our analysis, the relationship between vaccine security and AMC (%) becomes much stronger.

Request permission

Despite having access to incomplete price data, we report a very strong relationship between vaccine security and the mean price per dose paid by countries (figure 4B). As shown in figure 4C, this relationship remains strong even when data is analysed by region from a different public source.21 HICs paid significantly more per vaccine dose than LMICs, with a mean difference of US$23.17 (95% CI, US$19.12 to US$27.22, p<0.001).

Distribution

In figure 5A, we report a strong relationship between vaccine security and vaccines delivered. On average, HICs and LMICs had received deliveries to vaccinate 55.8% and 7% of their populations, respectively, by the end of March 2021, resulting in a difference of 48.8% (95% CI −118.7% to 21.1%, p=0.13).

Request permission

Interestingly, as seen in figure 5B, vaccine security does not appear related to the Stringency Index at either timepoint. HICs and LMICs had a mean Stringency Index of 59.2 and 64.0, respectively, resulting in a difference of 4.75 (95% CI −10.31 to 19.81, p=0.47). There was wide intragroup heterogeneity in the public health responses within HICs and LMICs, with Canada having the highest Stringency Index and Japan having the lowest.

Vaccine supply chain resilience refers to the ability of a vaccine supply chain to anticipate, respond to and recover from disruptions while maintaining its critical functions and ensuring high vaccine availability.22 For the purposes of our study, high vaccine security (ie, vaccination rates) may be taken as a proxy for a highly resilient vaccine supply chain as our study focuses on the early phase of the pandemic when vaccine availability was the main Tanahashi bottleneck. Overall, our data show higher vaccine security in HICs and lower vaccine security in LMICs, with a significant difference both between and within these groupings. Our study of supply chain variables between these groups reveals a significant difference only in the allocation phase predictor variables of AMC (%) and mean price per dose. It also appears that there was large variation in supply chain resilience within HICs, with Canada, Israel and Japan being outliers on various occasions, and within LMICs, with India standing out. Our data also show a stronger relationship between vaccine security and supposedly downstream supply chain variables (vaccine purchases, price per dose and deliveries) over upstream variables (vaccine development capacity, production and manufacturing facilities). However, the presence of notable outliers in each situation suggests that one or two supply chain variables alone may be insufficient to explain all the variation in vaccine security.

Lesson 1: Paying more helps, but financial muscle alone cannot ensure vaccine security

Among the nations in our study, Peru paid the lowest mean price per dose. As shown in figure 1A, Peru and Brazil have similar GNI per capita. However, Brazil paid more per dose than Peru and had greater vaccine security (figure 4B). Peru also had a much smaller delivery (%) to AMC (%) ratio than Brazil (1.9% in Peru vs 12.9% in Brazil). Vaccine security in Peru was further hampered by distribution challenges, with the ‘vaccine gate’ scandal of February 2021 revealing that experimental COVID-19 vaccine doses were administered to influential government officials not part of the trial.24 When Peru is compared with Brazil, it appears that a willingness to pay more per vaccine may help.

Despite having some product development capacity and securing a far higher AMC (%) than other HICs, Canada had a far lower vaccine delivery (%) compared with the US, the UK and Israel. Canada’s ratio of delivery (%) to AMC (%) was only 2%, far fewer than the UK (13%), the USA (16%) and Israel (48%). Several factors may have contributed to this situation. Canada placed significant bets on vaccines that happened to enter the market late. When it eventually placed orders for mRNA vaccines, manufacturers prioritised other nations due to funding agreements and limited Canada’s access to larger orders.25 Furthermore, Canada’s purchase contracts did not include advantageous delivery schedules. Last, a lack of domestic vaccine manufacturing capacity made Canada reliant on vulnerable global supply chains. Unlike other HICs that invested in infrastructure expansion, Canada’s historical underfunding of vaccine manufacturing hindered its ability to produce vaccines on a large scale. This lack of capacity prevented Canada from using manufacturing contracts to control supply or accelerate deliveries.26 Canada’s experience appears to suggest that vaccine supply chains are non-linear, and a narrow focus on apparent downstream supply chain variables such as vaccine purchases, while neglecting upstream elements, may not necessarily increase vaccine security.

It has been widely reported in the literature that excess procurement by HICs can drive inequity and that any perceived benefits to HICs may be limited and short lived.27–29 Furthermore, vaccine allocation proportional to wealth may be detrimental to all, and sharper disparities between HICs and LMICs have been shown to lead to earlier and larger outbreaks of new waves.27

Even after securing a far higher AMC (%) than other HICs, Canada was one of the few HICs to request doses from COVAX that were intended to build global equity.25 Despite the above, Canada had lower vaccine security than Israel, the UK and the USA. While securing more AMCs overall appears to increase vaccine security (figure 4A), the interconnectedness of vaccine supply chain components means that financial muscle alone cannot ensure vaccine security or make up for other deficiencies.

Lesson 2: Good governance is critical to vaccine security

Our data from Israel, Japan, Brazil, Vietnam and Guinea all suggest that good governance is vital to vaccine security and may even compensate for shortcomings in certain elements of the vaccine supply chain.

Israel far outperformed its HIC peers on vaccine security, despite ranking last on product and production phase variables, and second last on AMC (%) within the HICs we studied. Nevertheless, Israel was able to secure the highest number on delivery (%). Rosen et al outline reasons for Israel’s exceptional performance.30 Broadly, they can be categorised into: (1) good governance (a well-established framework for swiftly addressing national emergencies), (2) strong healthcare system (a technologically coordinated system that extends to community-based providers) and (3) novel factors specific to the COVID-19 vaccine (including innovative technical solutions to address Pfizer-BioNTech vaccine’s cold storage requirements and early participation in clinical trials). The Israeli experience illustrates that while preparation prior to a pandemic helps, in a time of crisis, a swift national response built on well governed public institutions and a nimble private sector may outweigh overpreparation and even make up for relative weaknesses in other elements of the vaccine supply chain.

Japan ranks 8 among the 10 countries for vaccine security, despite having novel vaccine discovery and manufacturing capacity. Japan’s ratio of delivery (%) to AMC (%) was 0%, the lowest in our study. The anomalous performance of Japan is the result of a unique governance challenge—a deliberately cautious approval process by regulators due to a high level of vaccine hesitancy among the Japanese people stemming from historical safety scandals.31 This highlights the importance of proactively working with public and private partners across all phases of the vaccine development cycle, from discovery to postregulatory approval, to avoid unnecessary roadblocks for otherwise promising vaccine candidates.

Despite having some manufacturing capacity and having the highest vaccine security among LMICs, Brazil still lagged well behind the USA, the UK, Canada and Israel. Our AMC (%) data corroborates reports that UMICs, such as Brazil, may have exhibited greater caution in securing AMCs compared with HICs, preferring to negotiate with manufacturers only after some efficacy signals were present.32 Furthermore, Brazil’s approach to production and procurement, built on its history of domestic manufacturing and technology transfers, was slowed by regulatory hurdles and conflicts between national and state governments.33 The case of Brazil again illustrates the importance of pandemic preparedness cooperation among public and private partners to expedite system-wide response times.

Vietnam had the lowest vaccine security in our sample, despite having a higher GNI per capita than Guinea and India. Interestingly, in 2020, with few COVID-19 cases and deaths, Vietnam was lauded for its efforts to control the spread of the virus.34 This over-reliance on a containment approach by Vietnamese health authorities may have hampered efforts to internally distribute even delivered doses of the vaccine, as Vietnam had a vaccination rate to vaccines delivered ratio of just 5% in March 2021, the lowest among LMICs.

With no product or production stage advantages and over-reliance on the African Union’s (AU) pooled purchasing power and donations, Guinea had the second lowest vaccine security in our sample. This was further complicated by political instability, with the AU suspending Guinea’s membership in September 2021 after a military junta overthrew the government.35

Lesson 3: The presence of some domestic manufacturing capacity may provide leverage against vaccine insecurity

While domestic manufacturing capacity may not guarantee vaccine security, it appears to guard against vaccine insecurity. Despite having the second lowest GNI per capita in our study and securing the lowest AMC (%), India was second only to Brazil among LMICs when it came to vaccine deliveries and vaccine security. Possessing manufacturing capacity may have given India flexibility. Once India faced a severe second wave in April and May 2021, it enacted export restrictions to prioritise its domestic population, highlighting the importance of investing in regional manufacturing capacity in LMICs that may accelerate response times and provide a buffer against an insular global supply chain.36

Our data suggests that the USA excelled in the product and production phases. Despite a willingness to pay high prices, the USA did not secure as many AMCs as Canada or the UK by the end of March 2021. The American strategy appears to have focused on consolidating its upstream supply chain advantages through Operation Warp Speed to achieve large-scale vaccine manufacturing. This resulted in the production of hundreds of millions of doses by early 2021.37 The American experience suggests that countries and transnational organisations with the means must continue to invest in and scale critical enabling competencies to further accelerate vaccine development. Despite possessing by far the greatest domestic manufacturing capacity (figure 3A), the USA did not have the highest vaccine security. In contrast to Canada, which relied too heavily on downstream elements of the supply chain, the USA appears to have counted too much on upstream variables. Despite some shortcomings, we note that the three nations with the greatest domestic manufacturing capacity (USA, India and UK) were all relatively vaccine secure compared with their peers, suggesting that having some domestic manufacturing capacity still helps. However, relying on only one or a few aspects of preparedness will not produce success.

Lesson 4: A resilient vaccine supply chain requires modest strength in multiple domains

The UK stands out in our study for its high level of vaccine security based on broad strength in all four phases of the vaccine supply chain, despite not achieving the highest scores in any one element. Baraniuk chronicles the reasons for British success, which include a focus on early novel vaccine development by Oxford University scientists that created a snowball effect, advantaging manufacturing, purchasing and delivery schedules.38 Strengthening multiple elements of the vaccine chain may have a synergistic effect, and the redundancy created by this approach may shield against unanticipated threats posed by a novel pathogen.