Insights from the COVID-19 pandemic on building resilience capabilities in vaccine supply chains supporting mass vaccination Open Access

Mass vaccination is a commonly used strategy to quell the uncontrollable spread of a pandemic disease. The rationale is that if vaccines can be rapidly and ubiquitously delivered to potential recipients, then susceptible populations can be immunized quickly, and disease transmission can be stemmed, reducing potential negative consequences. During normal times, pharmaceutical supply chains make tradeoffs between speed, cost and service (Cafone, 2022). However, the success of a mass vaccination endeavor during a pandemic is premised almost entirely on speed (Cafone, 2022; Budish, 2021). Any delay in vaccine development, production, delivery or inoculation could prolong the toll on human life and escalate social and economic impacts. Consequently, the benefits of mass vaccination are highly dependent on a resilient vaccine supply chain (VSC) that is capable of withstanding and rapidly surmounting unanticipated disruptions that emerge while orchestrating an urgent campaign.

One of the deadliest pandemics in recent times, COVID-19, resulted in over 767 million cases and 6.9 million deaths globally during the period from March 11, 2020, to May 31, 2023 [World Health Organization (WHO), 2023]. Because of the threats posed by the pandemic, mass vaccination was used as a critical weapon against the disease. Yet, despite the unprecedented speed in developing a viable vaccine, loss of life and economic costs continued to escalate following disruptions in the VSC (Bown, 2022, p. 10; GlobalData Healthcare, 2022; Trump et al., 2022).

Furthermore, several studies conducted at the onset of the COVID-19 pandemic point to the inability of supply chains to handle large-scale disruptions of the magnitude of a global pandemic (El Korchi, 2022). In the case of pharmaceutical supply chains, and specifically VSCs, decades of efficiency strategies (e.g. implementation of lean principles), coupled with numerous interdependencies, have increased the fragility of the supply chain and reduced the ability to handle unexpected shocks. To minimize disruptions, resilience must, therefore, be intentionally designed in the supply chain (Christopher and Peck, 2004; Trump et al., 2022), and VSCs are no exception.

Despite the importance of VSC resilience, however, past research has focused narrowly on VSC efficiency to the exclusion of resilience (Sheffi, 2020; Simchi-Levi and Simchi-Levi, 2020; Golan et al., 2021a, 2021b; Trump et al., 2022). A few researchers have looked at some challenges in the VSC (Hartner et al., 2024; Ferranna, 2024; Feddema et al., 2023; Trump et al., 2022; Alam et al., 2021; Finkenstadt and Handfield, 2021) and in the pharmaceutical supply chain in general (Ganguly and Farr, 2024; Ferdous et al., 2023; Bastani et al., 2021; Bastani, 2021), but most studies have focused on optimizing or improving only a portion of the VSC. As a result, the literature lacks thematic content on building end-to-end resilience in VSCs during a pandemic.

Research on supply chain resilience before COVID-19 has also focused primarily on localized or contained risks, for example, natural disasters, manufacturing plant disruptions and supplier shortages (Golan et al., 2020; El Korchi, 2022). These risks were studied as isolated events; however, COVID-19 revealed that an unexpected catastrophic event could trigger further events that result in multiple simultaneous risks. Recognizing the limitation of supply chains to handle an exacerbated situation of compound risk events that result in a serious disruptive state, researchers have called for increased research to generate new insights into resilience and recovery (Chowdury et al., 2021; Sarkis, 2020). As global pandemics are infrequent events, the COVID-19 experience offers a valuable opportunity to investigate the capabilities essential to VSC resilience in pandemic situations. Consequently, this research aims to answer the following questions based on insights from the COVID-19 pandemic:

To address these questions, this study undertakes an exploratory analysis using a diversity of data sources, such as white papers, reports, academic papers and audiovisual presentations of over 100 subject-matter experts in all areas of the COVID-19 VSC. A study of this scope, which reviews the collective opinions, insights and experiences of a large group of professionals globally, has not been previously undertaken. The results make several contributions that should be useful to researchers and practitioners interested in VSC resilience during pandemic situations. From an academic perspective, it advances the literature on resilience using the case of a massive pandemic that touched the entire world. This study also contributes to the growing body of literature on the resilience of complex supply chain systems and, in particular, mass vaccination supply chains. As an empirical study that used data from direct high-level discussions to reduce and prevent disturbances in the COVID-19 VSC, the findings may help advance a theory on VSC resilience. Researchers should also find it useful as an additional source of data in performing systematic reviews and triangulation of data pertaining to the COVID-19 VSC. From a practitioner perspective, the results serve as a comprehensive resource on the capabilities required to rapidly design, plan and manage resilience in all or part of a global mass vaccination supply chain during the urgency of a pandemic. As the case of a rare disaster, it also demonstrates how resilience can be built to augment traditional risk management practices.

The remainder of this paper is structured as follows: Section 2 provides some background on supply chain resilience. Section 3 reviews the literature on VSC resilience, and Section 4 details the methodology used for the study. Section 5 presents the results, which are further discussed in Section 6. Section 7 provides some concluding remarks on the study and opportunities for future research.

Interest in supply chain resilience has soared in the past decade (Tukahumabwa and Stevenson, 2015; Behzadi et al., 2020) particularly since the COVID-19 pandemic (Castillo, 2022). According to Datta (2017), the globalization of supply chains, implementation of lean inventory philosophies and lack of supply chain redundancies have increased susceptibility to disruption and triggered negative tailspin effects in production and delivery. Pharmaceutical supply chains, being no exception, have come under scrutiny for their increasing fragility and vulnerability to disturbances because of the implementation of these efficiencies. VSCs deployed during a pandemic are large complex ecosystems (Gunderson et al., 2010; Swann et al., 2020), with all the characteristics of a system of systems as shown in Table 1. Because of this complexity and the compound effect of their fragility, size, number of interactions and dependencies and exposure to unpredictable external events (Thacker et al., 2017; De Weck et al., 2011; de Bruijne and van Eeten, 2007), disturbances can quickly lead to a breakdown that delays achieving herd immunity goals.

Supply chains support multiple flows from one node to another across diverse links (Christopher and Peck, 2004). Both nodes and links are susceptible to disruption by events that may or may not be anticipated, are unknown or only partially understood. Unless capabilities to resist and overcome potential disruptions are built into the supply chains, disturbances can lead to a serious decline in performance.

As a complex system, the VSC has a defined boundary within which the component entities and systems interact. During a pandemic, these include supply chain participants (Sauser et al., 2008); the government; the public health sector; regulatory agencies; the private sector; multilateral organizations; community organizations; and other critical entities. Each subsystem seeks to achieve its mission, while also acting jointly with other systems to accomplish the shared mission of the system (Sauser et al., 2008).

The concept of resilience was first studied in ecological systems and defined as an intrinsic property that allows a system “to absorb a change in state variables, drive variables, and parameters, and continue to persist” (Holling, 1973). Since then, resilience has been explored in several academic disciplines, including psychology, sociology, disaster and emergency management, engineering and supply chain management (Ponomarov and Holcomb, 2009). However, because of the diversity of these fields and the vastly differing applications of the resilience concept, a comprehensive and universal definition is still lacking in academic literature (Ponomarov and Holcomb, 2009; Mensah and Merkuryev, 2014; Hosseini et al., 2016). In general, resilience is understood to be an intrinsic property achieved through the development of diverse capabilities that allow a system to absorb the impact of an unexpected shock, adapt to a disturbed state and restore performance to a desirable state.

Several scholars have since identified different phases of system resilience, from pre-disturbance to recovery, and have developed models and formulas to depict and quantify resilience. Bruneau’s (2003) resilience triangle is one of the classic models that graphically depicts the behavior of a system to a shock – an earth tremor. Upon the onset of the shock, the system immediately falls to its lowest performance level and then immediately commences a gradual recovery back to full performance (Figure 1).

Bruneau et al. (2003) developed a deterministic equation, as shown in equation (1), to quantify the resilience loss over time:

Where:

t₀ = time that the disruption takes place;

t₁ = time that total recovery to pre-disruption performance levels is restored; and

Q = amount of restored performance as a function of time.

While Bruneau’s model applies to several systems, its assumptions may be limited in the case of supply chains where the buffering role of inventory prevents a precipitous and instantaneous drop in performance following a shock. An alternative model proposed by Baroud et al. (2014) may better explain a supply chain’s behavioral states following a disturbance (Figure 2).

Baroud’s (2014) model identifies four states, beginning with a state of reliability (initial performance), followed by states of vulnerability, survivability and recoverability. System performance at any time t is denoted as Ф(t), with higher values indicating a preferred state [equation (2)]. Overall system resilience (ᴙ) at time (t) is observed if there is a disruptive event quantified on a scale from 0 to 1 as the ratio of recovery time to disruption time. A value of 1 denotes a perfectly resilient network.

Performance of a fully resilient system at time t_r, given a disruptive event e^j: i:

Where:

ᴙ Ф (t_r|e^j) = system resilience from initial performance to recovery, given a disruptive event, e^j;

Ф (t_r|e^j) = performance of a fully resilient system at time t_r given a disruptive event, e^j;

Ф (t_d|e^j) = performance of the system at time t_d given a disruptive event, e^j;

Ф (t₀) = performance of the system at time t₀;

t₀ = time at which the system is operating at its original state;

t_e = time at which disruption occurs;

t_d = time at which system reaches maximum disrupted state;

t_s = time at which recovery begins;

t_f = time at which full recovery is reached;

Ф(t₀) = performance of the system at time t₀;

Ф(t_d) = performance of the system at time t_d; and

Ф(t_f) = performance of the system at time t_f.

In the supply chain domain, resilience is still a relatively new concept (Christopher and Peck, 2004; Craighead et al., 2007; Falasca et al., 2008) under the general purview of, but distinct from, traditional risk management (Ponomarov and Holcomb, 2009; Pettit et al., 2010). Although several authors have attempted to describe and define the term, Castillo (2022) reports that an overarching theory on supply chain resilience remains to be developed.

Christopher and Peck (2004) characterized supply chain resilience as a system’s ability to shift back to an initial performance level or more desirable state following a disturbance. Sheffi and Rice (2005) proposed a similar definition but focused on the ability to return to normal performance following a disruption. Ponomarov and Holcomb (2009) concluded that supply chain resilience is the ability to adapt to unexpected events, react when disrupted and recover from disturbances by maintaining the required level of integration and control to continue operations. Datta (2017) described resilience as preventing and rapidly responding to disruptions to regain stability and ensure competitiveness. Falasca et al. (2008) viewed resilience as the ability to reduce disruption risk by decreasing the likelihood, impact and recovery time associated with disruptions. Kochan and Nowicki (2018) characterized supply chain resilience as the ability to shift performance toward a desirable state to avert supply chain failures. Implicit in all these definitions is maintaining the desired performance following an unexpected shock.

Researchers have also attempted to collate disparate views of supply chain resilience. Weiland and Durach (2021), for example, identified two divergent views that build on Holling’s (1996) work. One takes an engineering perspective, perceiving the supply chain as a configurable closed system detached from the external environment and capable of being optimized to improve efficiency. This view emphasizes the importance of maintaining an equilibrium state or bouncing back after a disruption. It is supported by Simchi-Levi et al. (2014), who suggest quantifying supply chain resilience in terms of time to recovery.

The second view takes an ecological perspective, perceiving the supply chain as an open or complex adaptive system interacting with its environment – a perspective supported by Choi et al. (2001). Gunderson and Holling (2001) also argue that viewing resilience as a single optimum state only works in systems with low uncertainty and not in dynamic, constantly evolving environments subject to high uncertainty. They emphasize that it is more important to maintain existence rather than efficiency. The ecological view underscores the difficulty of total control and optimization. Instead, it proposes building capabilities that allow appropriate responses to unexpected changes so that a system can cope with disruption. Wied et al. (2019) define two possible outcomes following a recovery that support the view of a different performance level:

1. bounce-back resilience, in which the disturbed system returns to its pre-disturbance performance levels; and

2. bounce-forward resilience, in which the system’s performance after a shock is either worse (fragile system) or better (anti-fragile system) than pre-disturbance levels.

Table 2 summarizes the various definitions of resilience.

The general idea behind resilience theory is to design systems that can handle unexpected events by minimizing the potential negative consequences (Wied et al., 2019). The authors (Wied et al., 2019) posit that a system’s performance following an unexpected shock is driven by the impact of the shock and the extent to which the system’s attributes can buffer the disturbances. This relationship is expressed as a function of two variables:

1. a set of unexpected events or disruptors; and

2. a set of resilience attributes that minimize the impact of the shock [equation (3)].

As shown in Figure 3, VSC performance is directly related to the resilience properties that counter a disruptive event: (Wied et al., 2019)

Where:

p = set of variables that define system performance;

C = set of uncertain events or shocks;

R = set of variables that mediate the shock and determine the ultimate; and

f = relationship between P, C and R.

Considerable variation exists across disciplinary areas on how resilience is achieved. The three-pronged model emphasizing absorptive, adaptive and restorative properties (El Korchi, 2022; Biringer et al., 2010) offers a structural approach for aligning resilience capabilities to post-disruption phases. In the supply chain literature, frameworks incorporating resilience capabilities have been proposed – the most popular being that developed by Christopher and Peck (2004) which contains four vital capabilities:

1. supply chain (re)engineering;

2. collaborative risk identification and management;

3. agile response to risk events; and

4. development of a risk management culture.

Each is further supported by several essential capabilities. Other researchers have emphasized attributes such as knowledge of internal and external supply chain risks, sustainability, redundancy, flexibility, visibility, information technology (IT) capability, information sharing, robustness, awareness, security, velocity, adaptability, market position, risk control, revenue sharing, public-private partnerships and supply chain network design (MIT Technology Review, 2020; Singh et al., 2019; Christopher, 2011; Sheffi and Rice, 2005). Zsidisin and Wagner (2010) emphasize building flexibility and redundancy to offset supply disruptions. A useful structural approach for identifying resilience capabilities and potentially linking to the three-pronged model is offered by Wied et al. (2019) who suggest asking three questions: “Resilience of what?”, “Resilience to what?” and “How should resilience be designed?”

Given this general conceptualization of resilience, VSC resilience can be described as the ability to buffer a disruption following an unexpected disturbance of the VSC flows by cultivating a set of capabilities that facilitate recovery to a desired or planned performance level. During a mass vaccination campaign, the VSC goes through several phases. Initially, production is ramped up rapidly, after which a relatively steady state of alignment between supply and demand occurs, followed by a decline as the campaign comes to an end. While disturbances can occur at any point, the initial phase is typically the most vulnerable.

Research on mass vaccination supply chain resilience capabilities during a pandemic is sparse (Qrunfleh et al., 2023; Joshi and Sharma, 2022; Golan et al., 2021b) and generally reflective of early investigations into a new area. Interest in supply chain resilience, and more specifically VSC resilience, only sparked during the COVID-19 pandemic following disruptions in multiple global supply chains.

To evaluate the work already undertaken on VSC resilience capabilities and establish a gap in the literature, SCOPUS and PubMed databases were searched using the keywords “COVID-19” AND “vaccine* AND “supply chain” AND “resilien*.” The search retrieved 70 documents from SCOPUS and 22 from PubMed. In addition, 14 articles were selectively retrieved from Google Scholar. From a total of 106 articles, duplicate papers, books and publications not relevant to the topic were removed. The remaining 31 articles that were reviewed fell broadly into two groups: qualitative analysis and the development of mathematical models.

In the area of qualitative research, various authors have conducted exploratory studies to identify the challenges in different parts of the COVID-19 VSC and recommend solutions to cope with individual challenges using a risk management approach. Hartner et al. (2024) conducted a rapid review of factors impacting immunization in low- and middle-income countries. The analysis revealed that health-care staffing shortages and vaccine stockouts were key disruptive factors. To combat the challenges, the authors recommended strengthening community outreach and health-care provider support but did not describe how this should be done. Ferranna (2024) conducted a comprehensive review that identified vaccine nationalism, intellectual property rights, manufacturing capacity limitations, poor health-care systems and vaccine hesitancy as disruptors to equitable vaccine allocation in low-income countries. The authors recommended increasing local manufacturing capacity; investing in health-care systems, cold chain equipment and infrastructure; training health-care workers; collecting data; engaging in outreach activities; and negotiating international agreements to resist the disruptions. Maramba et al. (2023) investigated the performance of supply chains and developed a case study for the acquisition and distribution of COVID-19 resources in South Africa. The authors reported that material shortages, broken distribution systems, insufficient vaccines and poor allocation of vaccine administration staff were major issues. To solve these problems, they suggested bolstering transport and logistics capacity. Mekonen et al. (2024) conducted a systematic review of the difficulties, impacts and opportunities of the global health commodities supply chain, including the COVID-19 VSC. The authors reported that insufficient manufacturing capacity, poor transport and cold chain systems, difficulty in monitoring vaccine uptake, poorly organized local entities, lack of vaccine monitoring bodies, monetary constraints, forecast inaccuracy, vaccine hesitancy and a lack of volunteer interest were responsible for disruptions. They recommended that stakeholders define, map and prioritize the challenges associated with vaccine development, manufacturing, delivery and administration for developed and low- and middle-income countries. However, specific recommendations on resilience capabilities were not offered. Golan et al. (2021a) reviewed the resilience literature pertaining to nanotechnology-based product supply chains, including VSCs. They recommended that standardization, network science, stakeholder accountability and scalability principles be applied in designing resilience. Bown (2023) reported on the market failures that impacted the supply of COVID-19 vaccines and justified domestic subsidies to strengthen supply chain performance. Pennisi et al. (2024) identified cold chain logistics, infrastructure, partnerships, cultural competency, community engagement, social factors and health-care infrastructure as essential to implementing VSC resilience. Gress et al. (2021) undertook a systematic review of newspaper comments on the Mexican Government’s handling of COVID-19 vaccine distribution and based on their findings, proposed a new design for the humanitarian supply chain. Marceau and Parwani (2021) examined the role of the World Trade Organization in facilitating cross-border vaccine flows and equitable distribution, while Fahrni et al. (2022) identified opportunities to strengthen distribution equity, improve logistics and enhance vaccine stability.

Among the researchers focusing on the COVID-19 VSC, Feddema et al. (2023) and Forman et al. (2021) stand out for using a systematic approach to identify potential disruptive forces and propose corresponding risk mitigation strategies. Feddema et al. (2023) conducted a scoping review in addition to undertaking 80 interviews and roundtable discussions. Based on the results, the authors proposed increased focus on production, allocation, technology transfer and regulatory alignment. However, this study was limited to disruptions in manufacturing operations. On the other hand, Forman et al. (2021) addressed disruptions in the end-to-end value chain, highlighting disturbances at the research and development, production and allocation stages. Disruptors identified included shortages of materials, low production capacity, poor cold chain infrastructure, criminal activity, pricing differentials, vaccine nationalism, vaccine hoarding and vaccine hesitancy. Suggestions offered to improve performance emphasized inventory buffers, joint patent purchases, compulsory licensing, lifting of export bans, security and transparency, flexible appointment scheduling, inclusive vaccine campaigns, rapid response to side effects, data interoperability and legal risk mitigation. Despite having an extensive list, there are gaps in the recommendations.

Some authors have conducted bibliometric analyses to summarize the researchers, countries and topics contributing to the VSC disruption and resilience literature. Moosavi et al. (2022) performed a literature search, bibliometric analysis and citation analysis and summarized resilience strategies as production, supply chain redesign and government intervention. Qrunfleh et al. (2023) used Bibliometrix software to assess the importance of VSC resilience and mitigation strategies; and Joshi and Sharma (2022) used Bibliometrix and Biblioshiny packages to review articles from 2000 to 2020. However, because of the period selected, the analysis pertained mainly to the 2009 H1N1 virus and the polio epidemic of 2014, both of which were of a significantly smaller scale than COVID-19.

Other qualitative research studies fall more broadly under the pharmaceutical supply chain umbrella. For example, Ganguly and Farr (2024) applied Interpretive Structural Modelling to determine and rank the hurdles to resilience in pharmaceutical supply chains in India. Bastani et al. (2021a) applied a Grounded Theory Methodology to elucidate information on supply chain resilience models in the pharmaceutical industry during disasters. This study, however, was limited to the Iranian context without specific emphasis on VSCs or pandemic situations. In a closely related study, Bastani (2021) conducted a content analysis to identify approaches to improve pharmaceutical supply chain resilience under politico-economic sanctions. Although the study identified several important factors, it also focused on the Iranian sector and was not specific to VSCs.

The second strand of research on VSCs focuses on developing mathematical models to optimize, improve or enhance VSC performance. Bamakan et al. (2021) identified 18 factors that caused the bullwhip phenomenon in the COVID-19 VSC and had 13 of these factors ranked by 72 pharmaceutical industry experts using Analytical Hierarchy Process. A fuzzy cognitive map of the influential factors was developed from this analysis. Golan et al. (2021b) conducted a systematic review of modeling techniques applied to improve supply chain resilience. The authors recommended the use of emerging technologies to better quantify the tradeoffs between efficiency and resilience. They also suggested that modeling of VSCs should incorporate the impact of one vaccine network on another because of common nodes inserted in the supply chains (e.g. raw material suppliers, cold chain infrastructure and logistics providers). Other mathematical models that may be of interest to the reader include the works of Jahed et al. (2024), Mahmud et al. (2023) Rinaldi et al. (2023), Apoorva et al. (2022), Kazancoglu et al (2022), Shahparvari et al. (2022), Sinha et al. (2022), Verma et al. (2022) and Miguel et al. (2021).

After reviewing the current literature, one obvious limitation of the studies is that resilience is equated to traditional risk management, where each disruptor is viewed as being contained within a part of the VSC and matched to a corresponding solution using a linear cause-and-effect approach. The VSC, however, is a complex system (Trump et al., 2022) consisting of multiple interacting sub-systems. Because of these interconnections, a disruption in one part could quickly spiral out of control if the system lacks the capabilities to surmount the disturbance on its own and prevent performance loss. Christopher (2011) recognizes the vital capability of being able to shift back to a prior performance level and identifies key resilience characteristics to resist such disturbances. From a more systemic perspective, Trump et al. (2022) propose implementing VSC resilience capabilities through intentional design (e.g. qualification of multiple suppliers and use of interchangeable parts) and intentional intervention (e.g. government investment in multiple manufacturers) (Trump et al., 2022).

Given this analysis, there remain a few gaps in the literature to be filled. The core resilient capabilities of VSCs first need to be identified and then mapped to specific operational and policy solutions for each of the seven critical flows in the VSC as identified by Lawrence (2024). Another opportunity exists to break down the top-level capabilities into detailed supporting capabilities. Thirdly, among the studies reviewed, only one researcher (Joshi and Sharma, 2022) applied cluster analysis to bibliometric analysis. Because of the systemic nature of the VSC, an opportunity also exists to perform a cluster analysis to identify the associations and relationships between the resilience capabilities.

To extract data on the insights, opinions, experiences and recommendations from experts, content analysis was used as an exploratory technique. Content analysis is well regarded as a systematic, structured and replicable approach for analyzing written, oral and visual data (e.g. transcripts, focus group interviews, survey responses and books) and for describing, analyzing and synthesizing the data into meaningful categories to understand a phenomenon or event (Elo and Kyngäs, 2007; Cho and Lee, 2014). Content analysis is also used to extract and identify patterns and contextual meaning to better understand a situation for which little information is available or for which actions have been taken during a past event (Van Note Chism et al., 2008) While other qualitative methods develop theory through an open-ended coding process, content analysis is useful for extracting data to answer specific research questions that typically ask “what?”, “why?” and “how?” (Heikkilä and Ekman, 2003, p. 138 as cited in Cho and Lee, 2014).

To perform a content analysis, the researcher collects the data by studying the output of participants in a naturalistic setting where the problem is addressed or experienced (Creswell and Creswell, 2018). From the data gathered, only relevant data is coded and categorized. Content analysis is flexible in allowing inductive, deductive or combined inductive/deductive coding approaches (Creswell and Creswell, 2018; Cho and Lee, 2014). Inductive analysis derives themes from the bottom up by organizing data into increasingly broad and abstract categories through an intensive iterative process that analyzes, compares and aggregates the codes, while deductive analysis establishes code categories at the outset of the study or from theory or past research (Elo and Kyngäs, 2007). In either case, categorization is flexible and allows the researcher to code both explicit and implied meanings in the data (Hsieh and Shannon, 2005). Given these characteristics, we found content analysis to be the most suitable method to answer the research questions for the following reasons:

• The research questions were specific rather than open-ended.

• The approach could be applied to diverse data types, for example, audiovisual presentations and documents.

• The method allowed for an appropriate sample size, given time constraints and the impracticability of convening high-level experts during the COVID-19 crisis.

• Large volumes of data on VSC challenges and solutions to support mass vaccination campaigns were disseminated in near real-time in the public domain, providing a rich stream of information that would have otherwise been difficult to replicate at the end of the pandemic.

• A significant amount of data in the public domain was produced in verbal form. Requiring experts to provide this information in written form via surveys or questionnaires would likely have resulted in data being filtered, condensed, diluted or omitted.

• Obtaining the viewpoints and experiences of high-level subject matter experts globally would have been difficult without the intermediary role of organizations that hosted symposia, webinars and other meetings.

• The intervening role of national governments and international organizations in mass vaccination endeavors required private sector companies, for example, pharmaceutical companies, to cooperate in providing information that might otherwise have been withheld.

• Content analysis diversifies the research methodologies commonly used in supply chain management research (Williams et al.., 2017; Kauffman and Denk, 2011).

To achieve the goals of this study, diverse data sources were used. In all, 36 data sources comprising audiovisual presentations of symposia, workshops and webinars and documents consisting of reports, white papers, Web pages and academic publications were selected (Table 3).

To extract data on COVID-19 VSC resilience capabilities, YouTube was selected as the primary source for capturing data from the audio-visual presentations of senior professionals and subject matter experts. YouTube is frequently used as a data source for academic research in the medical, sociological and biological domains (Allgaier, 2020) and particularly for studies on emergency disasters (Rout et al., 2024; Jamil Uddin et al., 2023; Slick, 2019; Owens et al., 2013; Hussin et al., 2011; Ravitz et al., 2010).

Furthermore, YouTube is the most used streaming video platform globally (Restream, 2022) and the most popular in the USA after Google (Statista, 2022). Daily mobile views surpass 1 billion with an average viewing time exceeding 45 min (GMI, 2020). In a study conducted during the COVID-19 pandemic, over 81% of respondents claimed to have used YouTube (Suciu, 2021), demonstrating its worth as an important information-sharing platform. YouTube was used by many highly regarded organizations to rapidly share emerging information discussed at meetings to address the challenges and solutions of the COVID-19 VSC. The proceedings of high-level panel discussions organized by public and private entities, including intergovernmental organizations, academic groups and consulting companies studying, designing, leading, and managing the VSC, were posted on YouTube. Given the time constraints for this study, the resources available and the difficulty in reaching high-level professionals during a global emergency, YouTube offered reachability and accessibility to a critical source of data that might otherwise have been ignored.

For this study, a sample size of 20–30 was targeted per the recommendations for Grounded Theory (Marshall et al., 2013: Sandelowski, 1995), which can also be applied to content analysis. The search term “COVID vaccine supply chain” was used to select appropriate data sources on the COVID-19 VSC during the two years from March 2020, when the World Health Organization (WHO) declared the COVID-19 pandemic, to April 2022, when advanced countries began removing pandemic restrictions. The search term was intentionally kept broad to obtain all relevant YouTube presentations. The top 201 audiovisual recordings retrieved plus one additional recommended presentation were selected for further review and analysis.

The recordings were initially sorted based on title; duplicates were removed; and the remaining presentations were examined for compliance with the inclusion criteria. A salient requirement was the credibility of the organizations providing the data. International organizations, government agencies, universities, consulting companies, professional associations and trade associations deemed credible were considered acceptable. The content also needed to be delivered by subject-matter experts in an area pertinent to the COVID-19 VSC. Examples of acceptable individuals included senior officers from pharmaceutical, distribution, logistics, transportation and freight forwarding companies; international organizations; regulatory bodies; trade, customs and border protection; health-care administrators and practitioners; government employees appointed to lead and manage the VSC; and university professors in engineering, supply chain, logistics, management, international trade, economics and health care. Presentations by politicians, such as Congressional Hearings, were excluded. Other requirements were a minimum duration of 30 min for each event, which eliminated approximately 90% of YouTube presentations on short news stories, updates and advertisements. The final criterion was the delivery of the presentations in the English language.

From an initial number of 202 audiovisual presentations, 21 events consisting of 108 participants and representing 30 h of recorded time were selected. Transcripts accounting for 18.5 h were downloaded and reviewed for accuracy, spelling and punctuation against the audio recordings to ensure the correct capture of the transcribed data. This was a necessary, though time-consuming, step as the transcripts often lacked proper punctuation and the voice recognition software sometimes failed to transcribe correctly because of speaking style, pronunciation, enunciation, etc. The remaining presentations (10.5 h) were transcribed manually to text using MS Word over several weeks, and the transcribed data was cross-checked twice. The first check was concurrent with manual transcription and involved replaying the audio presentation multiple times to ensure correct transcription. The second review was a verification of each transcript against the original audiovisual presentation. The complete process of accessing, correcting, checking and rechecking the transcripts was performed over approximately four months. Table 3 and  Appendices 1 and  2 provide summary and detailed information on the titles and affiliated organizations of the 108 contributors/panelists in the study.

Other important data sources were the documents authored by academics, companies and individuals and interviews published on websites. In all, 15 documents consisting of white papers, reports and academic papers were selected from targeted authors and reputable organizations. The documents varied in scope and impartiality, and following a review for relevance and appropriateness in the context of COVID-19 VSC resilience, 14 papers were selected for data extraction. Additionally, data was extracted from short interviews published on the website of an international organization. Table 4 summarizes the number and type of all data sources used in the study.

The transcripts and other documents were imported into NVivo 12 software, and each document was read meticulously, line by line. Immersion was used to gain a full understanding of the contents to guide the development of categorical codes. A single coder was used throughout to ensure coding consistency. An inductive coding approach was initially applied to the data using an iterative process that required repeated review, query and reflection (Moretti et al., 2011). To reduce the number of nodes, the codes were condensed into increasingly broader categories, with the highest-level categories determined using a mixed inductive/deductive approach based on the researcher’s extensive supply chain management knowledge, experience and previous exploratory work (Lawrence, 2024). Deductive codes were developed to align with mass VSC flows prone to disruption during a pandemic. For this study, data collection, cleansing and coding were conducted over approximately nine months (Figure 4).

To further identify similarities in the data, cluster analysis was used to analyze the sources, primary factors and sub-factor in a comparison matrix. The sources of data were clustered based on word similarity and evaluated using Pearson’s correlation coefficient (r), while the coded data was clustered based on coding frequency and evaluated using Jaccard’s coefficient. Pearson’s correlation coefficient (r) measures the extent of linear association by calculating the covariance between two variables of interest. As in quantitative analysis, values range from −1.0 to 1.0, with 1.0 indicating a perfect positive correlation between the two variables and −1.0 indicating perfect negative correlation. An r value of 0.7 or greater indicates a very strong relationship between two variables [equation (4)]:

Where:

• r = Pearson correlation coefficient;

• x_i = value of x-variable in sample;

• $x¯=mean value of x variables$⁠;

• y_i = value of y-variable in sample; and

• ȳ = mean value of y variables.

Jaccard’s similarity coefficient is used to measure the similarity between two finite sets of data by comparing the size of the intersection to that of the union of the data sets. Values range from 0 to 1, with 0 denoting an intersection of 0 and 1 an intersection equal to the union of the two sets [equation (5)]:

Where:

• A = Data set 1 (Source A); and

• B = Data set 2 (Source B).

Criteria to evaluate a content analysis for validity and reliability is lacking in the literature. However, the recommendation is to use the same criteria for general qualitative research (Elo and Kyngäs, 2007; Cho and Lee, 2014). In qualitative research, validity refers to the accuracy of inferences drawn from a study from the researchers’, participants’ or reviewers’ perspectives (Creswell and Miller, 2000). According to Creswell and Miller (2000), the validity method used should depend on the lens researchers select to validate their studies and the researchers’ paradigm assumptions.

For this study, the methods proposed by Creswell and Miller (2000), Cho and Lee (2014) and Creswell and Creswell (2018) were used to establish validity from the researcher’s perspective and from constructivist and critical perspective paradigms. Regarding the first, validity was established through the length of the study, the use of appropriate sample size and composition, the achievement of an acceptable level of saturation and an iterative process for reflecting on and revising the assigned codes to ensure accurate interpretation and categorization of participants’ contributions. An iterative process was used to narrow the categories to relevant themes and eliminate overlapping themes.

Validity was also established using constructivist and critical perspectives paradigms based on trustworthiness (also described as credibility, transferability, dependability and confirmability), authenticity (e.g. fairness) and reflection on potential researcher bias. Credibility was established using data triangulation, whereby the same approach was applied to analyze data from different sources to verify or falsify trends perceived in the data. Transferability (applicability) was established by setting the context of the study as the COVID-19 pandemic. According to Creswell and Creswell (2018), a detailed description that allows readers to determine whether the data can be transferred to other scenarios can be used as a validation method. Dependability (consistency) was established by ensuring all coding was meticulously recorded. The role of the researcher’s supply chain background was also examined in deriving the codes. While the codes were primarily determined using an inductive approach, the researcher’s understanding and knowledge of supply chain management helped to develop some code descriptions deductively.

While there is a lack of consensus on what constitutes reliability in qualitative research, it is generally agreed that consistency in coding, rather than exact replication of methods and results (Carcary, 2009), is appropriate to establish reliability. Mason (2002) notes that from an interpretivist’s view, reliability is about providing evidence that the researcher has not invented or misrepresented data or been negligent in data recording and analysis. Grossoehme (2014) recognizes that the researcher is an integral part of the subject of the study and will influence it; however, this does not diminish the reliability of the study but, rather, improves it. Creswell and Creswell (2018) suggest documenting the procedures employed in detail so that another researcher can follow the methodology and produce similar results. Silverman (2009) proposes constant comparison and the use of comprehensive data, including data that deviate from expected results, while Gibbs (2007) recommends that transcripts are checked for accuracy; a single coder is used to ensure consistency and a codebook is established to document the results. The data derived from sources must be continually checked for accuracy and context by a single coder or multiple coders.

For this study, reliability was established in three ways. First, through an extensive iterative process, the codes were reflected on and refined by an experienced supply chain specialist to categorize chunks of data, while also comparing similarly coded data from different sources to ensure consistency. Secondly, multiple sources were used to increase the reliability of the findings. Thirdly, as coding was conducted by a single person, reliability was established through intra-rater repeatability by coded documents at least six months apart. In all, 50 samples were pulled from the data sources using a simple random number generator. For each document, excerpts were randomly selected, coded and compared to the previously coded document and assessed using a binary yes/no approach. The results indicated over 90% agreement at the sub-factor Level 1 and sub-factor Level 2 echelons.

In this study, 36 sources of data were imported into NVivo 12, read, analyzed and coded. Through the iterative process of code refining, two of the data sources, which were both academic papers (Sources 22 and 29) ( Appendix 1), were found to provide little data in support of VSC resilience capabilities, essentially reducing the number of sources to 34.

What primary capabilities are essential to building resilience in VSCs supporting a mass vaccination operation during a pandemic?

Through an iterative coding process, 2,246 coded references were categorized into seven primary factors essential to resilience in VSCs during a pandemic. Agility accounted for over 51% of the codes, stressing the vital importance of rapid assembly and deployment of the VSC to meet critical volume requirements. Risk Mapping, Modeling, Planning and Control (Risk MMPC) followed at 15.2% of the codes. Inter-organizational, Cross-sector and Cross-Border Collaboration (ICC Collaboration) amassed 12.9% of the codes, and Network Planning and Design (Network P & D) received 9.9%. The final three categories accounted for 10.8% of the codes: Supply and Demand Planning and Coordination received 4.3%, Complex system Management Skills 4.2% and Pre-pandemic Preparedness 2.3% of the codes (Figure 5).

What sub-factor resilience capabilities underpin the primary resilience capabilities and how should these be categorized and prioritized?

An analysis of the sub-factors that underpin the resilience properties of each of the seven primary factors provides deeper insights into where and how resilience needs to be built in VSCs. The data were sorted into 44 Level 1 and 145 Level 2 sub-factor categories, with Level 1 sub-factors disaggregated to align to each of the seven critical VSC flows as identified by Lawrence (2024) – that is, product flow, financial flow, information flow, vaccine recipient flow, knowledge flow, skilled worker flow and waste flow ( Appendix 2). Level 2 sub-factors detail specific ways in which resilience can be operationalized in each VSC flow. Figure 6 demonstrates the coding hierarchy.

Based on the flows identified in the VSC during a pandemic (Lawrence, 2024), seven Level 1 categories of sub-factor capabilities were identified to support the primary factor Agility. Data analysis revealed that Agility capabilities of Product Flows were the most frequently coded category, followed by Agility capabilities of Information, Financial and Knowledge Flows in that order. Significantly less emphasis was placed on the Agility capabilities of Skilled Worker, Vaccine Recipient and Waste Flows. The radial graph in Figure 7 depicts the relative importance assigned to each of the respective Level 1 factors.

Further data analysis indicated that 52 Level 2 sub-factors are required to support the Agility Level 1 sub-factors. Of these, 14 fell in the top 80% based on coding frequency (Table 5). A complete list of Level 2 sub-factors supporting Agility Level 1 sub-factors is provided in  Appendix 3.

Mounting a global mass vaccination campaign involves an exorbitant amount of risk, from the selection of vaccine technology to determining the quantity and timing of vaccine doses to meet herd immunity goals. Because of the unpredictable nature of a pandemic, risks are generally not known or understood in advance. Capabilities must, therefore, be developed to rapidly map, model, plan and control the risks as they occur during a pandemic.

Based on the flows identified in the VSC during a pandemic (Lawrence, 2024), seven Level 1 categories of sub-factor capabilities were identified to support the primary factor Risk MMPC. Analysis of the data revealed that Risk MMPC capabilities of Product, Financial and Waste Flows were coded most frequently. Risk MMPC of Vaccine Recipient Flows was addressed moderately, while minimal emphasis was placed on Risk MMPC of Knowledge, Information and Skilled Worker Flows (see the radial graph in Figure 8).

Further data analysis identified 27 Level 2 resilience sub-factors, supporting Level 1 Risk MMPC sub-factors. Of these, six fell in the top 80% of Risk MMPC Level two factors based on coding frequency (Table 6). A complete list of Level 2 sub-factors supporting Risk MMPC Level 1 sub-factors is provided in  Appendix 4.

Mass vaccination is an endeavor that requires global collaboration, as well as local organizational and cross-sectoral collaboration. Based on the COVID-19 pandemic, the most frequently discussed type of collaboration entailed partnerships and alliances to enable rapid product flow.

Based on the identified flows in the VSC (Lawrence, 2024), seven Level 1 categories of sub-factor capabilities were identified to support the primary factor ICC Collaboration. Of these, ICC Collaboration to build resilient Product Flows was cited most frequently. ICC Collaboration to facilitate Financial Flow resilience was also highly discussed. ICC Collaboration pertaining to Information, Knowledge and Vaccine Recipient Flows was discussed moderately to minimally. ICC Collaboration to achieve Waste Flow resilience received minimal attention and ICC Collaboration for Skilled Worker Flows received no attention (see the radial graph in Figure 9).

From the data, 17 Level 2 resilience sub-factors supporting ICC Collaboration Level 1 sub-factors emerged. Of these, five fell in the top 80% of ICC Collaboration Level two factors based on coding frequency (Table 7). A complete list of Level 2 sub-factors supporting ICC Collaboration Level 1 sub-factors is provided in  Appendix 5.

VSC networks supporting routine immunization during normal times quickly collapse during a pandemic because of speed, volume, scalability and reach requirements. To maintain high stakeholder satisfaction levels, a well-designed resilient end-to-end supply chain network is essential.

Based on the flows identified in the VSC, seven Level 1 categories of sub-factor capabilities were identified to support the primary factor, Network Planning and Design. Of these, Network Planning and Design to facilitate Product Flow resilience overwhelmingly received priority. All other factors were minimally addressed (see the radial graph in Figure 10).

From the data, 14 Level 2 resilience sub-factors supporting Network Planning and Design Level 1 sub-factors were identified. Of these, three fell in the top 80% of Network Planning and Design Level 2 factors based on coding frequency (Table 8). A complete list of Level 2 sub-factors supporting Network Planning and Design Level 1 sub-factors is provided in  Appendix 6.

During normal times, planning and coordination of inventory flows require tremendous skill to avoid stockouts and potential penalties. During a pandemic, the processes are even more critical, as the needs of all potential recipients must be met expediently and equitably.

Based on the flows identified in the VSC during a pandemic (Lawrence, 2024), seven Level 1 categories of sub-factor capabilities were identified to support the primary factor Supply and Demand Planning and Coordination (see the radial graph in Figure 11). Of these, Supply and Demand Planning and Coordination supporting the resilience of Product Flows was the most extensively discussed.

From the data, ten Level 2 resilience sub-factors supporting Supply and Demand Planning and Collaboration Level 1 sub-factors were identified. Of these, three fell in the top 80% based on coding frequency (Table 9). A complete list of Level 2 sub-factors supporting Network Planning and Design Level 1 sub-factors is provided in  Appendix 7.

Although a pandemic is a rare catastrophic event, lessons from the COVID-19 pandemic suggest that some actions can be taken in advance to support the resilience of mass VSCs.

Based on the identified flows in the VSC (Lawrence, 2024), seven Level 1 sub-factor capabilities are required to support the primary factor Pre-Pandemic Preparedness (see the radial graph in Figure 12). Of these, Pre-pandemic Preparation to support the resilience of Product Flows was the most heavily discussed. Pre-pandemic Preparation of Financial, Knowledge and Vaccine Recipient Flows was moderately discussed, while Pre-pandemic Preparation of Information, Skilled Worker and Waste Flows was not addressed.

From the data, 13 Level 2 resilience sub-factors supporting Pre-Pandemic Preparedness Level 1 sub-factors were identified. Of these, three fell in the top 80% based on coding frequency (Table 10). A complete list of Level 2 sub-factors supporting Pre-pandemic Preparedness 1 sub-factors is provided in  Appendix 8.

The VSC deployed during a pandemic is a complex system of systems that requires a set of unique management skills to achieve the overall goal of herd immunity. As for other complex systems, achieving this outcome requires significant interactions between various entities and sub-systems. From the data, two broad categories of Complex System Management Skills emerged – namely, system management skills and globally responsible leadership. The overwhelming number of references fell under system management skills (Figure 13).

Complex System Management Skills comprised 12 of the Level 2 sub-factors. Of these, seven skills fell in the top 80% (Table 11). However, an important complex system management skill – the ability to work with ambiguity, complexity and uncertainty was not considered to any great extent, falling below the top 80% range. A complete list of Level 2 sub-factors supporting Complex System Management Skills Level 2 sub-factors is provided in  Appendix 9.

What similarities and associations are evident between the primary resilience capabilities and the sub-factor resilience capabilities?

To determine the similarities and connections within the data, a cluster analysis was further performed on the sources of data, primary factors and Level 1 sub-factors. Data sources were clustered based on word similarity and evaluated using Pearson’s coefficient. Primary and Level 1 sub-factors were clustered based on coding similarity and evaluated using Jaccard’s coefficient. In each case, the coefficients were calculated for pairs of items in a matrix. The resulting similarities were depicted in hierarchical dendrograms and circle graphs. Hierarchical dendrograms indicate the degree of similarity through nested relationships, with the innermost nests having the strongest similarities. Lines connecting different factors in circle graphs indicate a relationship or similarity, with thicker lines indicating stronger similarities.

A cluster analysis of the data sources based on word similarity and assessed by Pearson’s correlation coefficient (r) indicated two large clusters roughly divided between the audiovisual presentations and the written sources, regardless of sector/industry. This is evident in the horizontal dendrogram in Figure 14. The Pearson coefficient (r) ranged from 0.1275 to 0.8738.

The results depicted as a circle diagram indicated very strong similarities (r = 0.7–1.0) between 15 of the 21 audiovisual presentation sources when Pearson’s correlation coefficient was used to measure the associations. Moderate correlation (r = 0.5–0.7) was evident among all the sources (Figure 14).

The primary resilience capabilities were analyzed based on coding frequency, with clusters indicating similarities or associations arising from coding of a reference to more than one node. Several primary factor pairs exhibited strong similarities when Jaccard’s coefficient was calculated. The strongest relationships were denoted by the thickest lines, with very strong associations observed between Agility and Risk Mapping, Modeling, Planning and Control (J = 0.9412) and between Agility and Network Planning and Design (J = 0.8824). Table 12 summarizes the relationships with a Jaccard coefficient of 0.7–1.0. In the moderate range of 0.3–0.5, several similarities were also identified between the factors. These similarities may help to explain actual relationships (See Figure 15 for a dendrogram of these relationships; See  Appendix 10 for circle graph relationships; and Table 12 for connected pairs of primary factors in the matrix).

A cluster analysis of Level 1 sub-factors based on coding similarity revealed that the strongest associations (Jaccard coefficients > 0.80) were observed between Supply and Demand Planning and Coordination and Network Planning and Design supporting financial flows and skilled worker flows; Inter-organizational, cross-sector and cross border collaboration and Risk mapping, modelling, planning and control supporting financial flows; and between Agility capabilities supporting product and information flows. These associations are represented in the horizontal dendrogram in Figure 16 and circle diagram in  Appendix 11. Table 13 indicates significant similarities (J > 0.7) for 13 pairs of Level 1 sub-causal factors.

This study aimed to identify the critical resilience capabilities required to handle unexpected disruptions in VSCs deployed to support a mass vaccination campaign during a global pandemic. With the help of NVivo 12 software, audiovisual presentations of over 100 global experts, along with white papers, reports and other documents on the COVID-19 pandemic, were reviewed and analyzed using a rigorous iterative coding process over several months. More than two thousand raw coded references were derived from the source data and gradually consolidated into increasingly higher categories in accordance with the practice of content analysis. The results yielded a set of primary factors and supporting sub-factors essential to achieving VSC resilience in seven critical flows of mass VSCs.

The results of the study indicate that based on the COVID-19 context, VSC resilience during a pandemic can be achieved by developing resilience capabilities along seven primary lines:

1. agility;

2. risk mapping, modeling, planning and control;

3. network planning and design;

4. inter-organization, cross-sector and cross-border collaboration;

5. supply and demand planning and coordination;

6. pre-pandemic preparedness; and

7. complex system management skills.

Given the unpredictable nature of a pandemic and its global spread, Agility capabilities emerged as the most critical primary factor. This is not surprising as pandemics occur rarely and without warning and the scale needed to arrest the disease may not justify the building of infrastructure that is unused during normal times. Instead, the ability to rapidly reconfigure supply chains to support all critical flows becomes imperative.

Regarding specific VSC flows, the data largely indicate that there is a strong awareness of the need to build resilience in product and financial flows, and to some extent, information flows. However, the limited attention afforded to building resilience in Knowledge, Vaccine Recipient, Skilled Worker and Waste Flows suggests that these areas may be assumed to be adequate to meet the needs of VSCs deployed in response to a pandemic. The infrastructural, policy and organizational decisions to support a global emergency effort of the scale of COVID-19 require greater attention.

To further strengthen resilience properties, the appropriate complex management system skills are integral to achieving the overarching goal of herd immunity. One of the most critical complex system management skills – the ability to work with ambiguity, complexity and uncertainty – was not emphasized in the discussions on the VSC. This capability fell in the bottom 20% of complex system management skills discussed in the data sources, suggesting that the systemic nature of mass vaccination supply chains may be unrecognized or poorly understood. Future considerations to develop a pipeline of complex system management talent will be essential to strengthen the deployment of resilient VSCs during a future pandemic. The implications are also relevant for team building and overall leadership and management of a complex VSC system.

While Pre-Pandemic Preparation was least emphasized in the discussions, it should not be construed that this dimension is irrelevant. Sub-factors supporting pre-pandemic preparedness reduce VSC risks that could lead to disruption and lay the groundwork for building agile response capabilities.

Several sub-causal factor capabilities to support resilience in the critical VSC flows emerged from the data. Of these, the major factors included the capability to: expedite cross-border flows through national and international policies; ramp up capacity across the end-to-end supply chain in manufacturing, cold chain infrastructure, air cargo, dry ice production and last mile delivery; implement redundancy strategies in vaccine research and development, manufacturing and logistics; implement flexibility, substitutability and interchangeability across raw materials and component parts; map end-to-end supply chain bottlenecks; develop strategies to minimize stockouts; collaborate at national and international government levels in subsidizing capacity, research and development and regulatory approval; develop information sharing platforms both physically and electronically; invest at risk in all stages of the VSC, particularly research and development and manufacturing; forecast, manage, synchronize and allocate materials and products in accordance with supply chain techniques; develop readiness strategies in vaccine research, production, regulation, option contract negotiation and investment planning; develop diverse strategies to promote vaccine safety and efficacy and avert vaccine hesitancy; and develop strategies to minimize waste through waste handling and sustainability policies and strategies.

From the Jaccard coefficient values, sub-causal factor associations were also identified, with the strongest associations between financial, skilled worker, product, information and vaccine recipient flows supporting the primary factors, Supply and Demand Planning and Coordination, Network Planning and Design, Inter-Organizational, Cross-Sector and Cross-Border Collaboration, Agility and Risk Mapping, Modeling, Planning and Control. What this suggests is that there are many dependencies, interconnections and cause-and-effect relationships that, when combined concurrently, reinforce resilience in mass vaccination supply chains.

Cluster analysis of the data provided some interesting insights into both the primary and Level 1 sub-causal factors. Cluster analysis performed on a qualitative data set provides exploratory rather than causality insights. However, the close association observed between several pairs of factors, for example, Agility and Risk MMPC; Agility and Network Planning and Design; and Agility and Inter-organizational, Cross-sector and Cross-border Collaboration, points to a complex VSC network with dependencies that exceed linear relationships and may need to be built and executed simultaneously to reinforce or achieve VSC resilience.

A pandemic of the scale of COVID-19 requires urgent containment to avoid disastrous consequences to human life, society and the economy. Presently, mass vaccination is considered to be the most effective way to immunize a population against a pandemic disease. However, because mass vaccination supply chains are complex systems, disruptions in one part can produce a domino effect of breakdowns that delay the attainment of herd immunity goals. To prevent this outcome, resilience capabilities must be built throughout the VSC to allow it to quickly adapt and recover from disruptions itself.

This study provides deep insights into the factors essential to VSC resilience during a pandemic. As an academic study, the results serve to complement and validate the information provided by practitioners and others on the COVID-19 pandemic. Using data from a wide spectrum of published and audiovisual sources, this study reports seven categories of resilience capabilities to reduce disturbances in mass vaccination supply chains. Each category of resilience capabilities is further aligned to the seven critical flows in the VSC (Level 1 sub-factors) and broken down into specific activities (Level 2 sub-factors) to achieve resilient flows. Through cluster analysis, associations between the resilience capabilities are identified, indicating where several strategies may be required simultaneously to reinforce VSC resilience.

While conducted as a comprehensive investigation of discussions at the highest levels globally, the research can be improved in several ways. First, the study used passive data extraction primarily from YouTube presentations, white papers, reports and academic publications, to reach a large number of high-level individuals involved with the VSC and the mass vaccination campaign. However, further clarification of data could be obtained through the use of focus groups, semi-structured interviews or surveys using a select pool of participants.

Opportunities also exist to verify and validate the priorities or weightings of the primary dimensions and underlying sub-factors. For example, weightings could be compared to the code frequency in this study to validate the findings. Data could also be limited to a particular country or region to focus on the specific issues manifested during the COVID-19 pandemic.

Other opportunities include using the results of this study as a starting point for conducting quantitative research related to hypothesis testing, structural equation modeling and the development of quantitative models to better explain resilience in VSCs. While data was extracted during the peak of the pandemic, retrospective reflection at the end of the pandemic could possibly elicit new details.

Through cluster analysis, this study provided insights into resilience capabilities that may be dependent or complementary. However, cluster analysis is an exploratory technique and further analytical work would be necessary to derive causality relationships.

Table A1 

Table A2 

Table A3