Mitigating personal protective equipment (PPE) supply chain disruptions in pandemics – a system dynamics approach

3.1 The interview study

The main reason for using an interview study was to capture not just the phenomenon of disruptions and mitigation strategies, but especially their underlying causes and interrelations as they occurred in a pandemic context with global effect. Since the phenomenon and the literature on supply chains in the pandemic context is limited, we took an exploratory approach to investigate how organisations responded to supply chain disruptions caused by the COVID-19 pandemic and how they enhanced their ability to avoid upcoming disruptions. Interview respondents were identified through a combination of snowball and criterion sampling, complemented by convenience sampling. First interviews were with project end users. These interviewees were also used to identify other relevant organisations and experts. Further experts were identified through the networks of project members. In addition, medical NGOs also contacted us directly to express their interest to contribute to the study.

In total, 80 organisations were contacted for this study, resulting in 38 interviews with experts during May–September 2020. Respondents came from different geographical areas (Europe, China, US and Canada) and included experts from non-governmental organisations (NGOs), health care providers, health professionals (medical doctors and nurses from hospitals), pharmaceutical professionals, decision-makers of ministries of health, medical item and equipment suppliers, and PPE manufacturers (see Table A1). To better understand the logistics challenges and the mitigation actions, we targeted roles related to the supply chain, procurement, and logistics within these organisations. We talked to logistics directors, procurement and supply chain managers, sourcing directors and emergency response managers. The mix of the different roles that we interviewed helped us to understand the supply chain disruptions that occurred along with the corresponding mitigation actions that different types of organisations used.

Prior to the interviews, a semi-structured interview guide was developed. Semi-structured interviews allow to capture emerging, new insights, as respondents can express their experiences and add insights that were not anticipated (Fawcett et al., 2008). The interview questions are guided by the conceptual framework of supply chain disruptions (see Figure 1).

Due to travel restrictions and social distancing rules, all interviews were carried out online via MS Teams or Zoom. Interviewees were contacted via email to set up the calls and consented actively to their participation. Most of the interviews were held in English, except for two that were conducted in, and later translated from, Cantonese. Interviews lasted between 45–60 min each. They were recorded, and then transcribed using the automated transcription function of NVivo, a qualitative data software program that has also the ability to organise and code data (Dean and Sharp, 2006). Transcripts were manually cross-checked by the research team as well as the respondents.

Using NVivo, the data was coded with respect to the causes of PPE supply chain disruptions which are identified from the literature and mitigation strategies to overcome them. The coding paradigm of Corbin and Strauss (2015) was followed, which consists of open, axial, and selective coding, as it provides a thorough and structured approach for examining the phenomenon of interest while leaving room for upcoming new categories.

First, open coding was introduced based on the four types of disruptions identified by the literature. In NVivo, open codes were introduced as free nodes (direct effect of COVID-19, policy induced disruption, strategic induced and pandemic induced disruptions) and measurements taken or suggested to overcome those disruptions. Then, axial coding was used to identify themes and subcategories. During this process, we identified the effect of the disruptions on different areas. We continued with the selective coding as the final stage in the data analysis, resulting in selected core categories and sub-categories (capacity, regulations, strategic dimensions as well as on the behaviours). Table A2 shows the resultant nodes and their subcategories, what causes them (fourth type of disruptions identified from the conceptual framework and validated by our data), include representational quotes from the data and the mitigation strategies and the representative quotes.

The quality of the empirical study can be assessed by the dimensions of pre-understanding, credibility, transferability, dependability, conformability, integrity, and utilisation, as recommended by supply chain researchers for qualitative studies (Flint et al., 2002; Halldórsson and Aastrup, 2003; Kaufmann and Denk, 2011). The researchers in the study have worked with medical supply chains before, shaping their pre-understanding. Regular discussions among project team members and with end users further contributed to the deeper understanding of medical supply chains in pandemics, the credibility of the analysis, especially with regards to emerging themes and their categorisation, and the integrity of interpretations. Integrity was further attested by the respondents who cross-checked their transcripts. Transferability was addressed by using respondents from various geographical areas and expert roles.

During interviews, respondents reflected on their expertise and previous experience. Knowledge was constructed together in the research team, increasing the conformability of the research. The utilisation of results was increased by the cross-checking of transcripts by respondents, but also by the organisation of a stakeholder workshop in October 2020 to which end users, respondents, and other stakeholders were invited, including a wider audience of experts in the same and similar roles of expertise. To ensure triangulation, we compared responses across respondents and with the existing literature and we had access to the reports of the organisations that are involved in the study related to their response to COVID-19.

3.2 System dynamics modeling

The results of the interviews data analysis informed the SD model. At its core, SD models complex dynamical systems and helps to understand them by building a causal theory of the interaction of their numerous parts represented in a causal loop diagram (Pruyt, 2013). It also allows taking into account non-linear relations between system elements, feedback loops and delays, which are critical for supply chain planning. While insights from the SD methodology and related concepts can generate managerial insights (Turner et al., 2018), an alternative is to develop a conceptual model, which later can inform a simulation model (Zhao et al., 2019). In this study we are focused on the second. Importantly, SD is not the only methodology to model complex systems. Other examples are complexity science (CS), with its core concept of a complex adaptive system (CAS) and discrete-event simulation (DEVS). CAS has a set of novel and potentially useful for supply chain concepts such as resilience, self-organisation, etc (Van Dam et al., 2012). CAS is typically either a network or an agent-based model (Mittal and Risco-Martín, 2017). The individual elements of these models “grow” the overall behaviour of a system through a set of interactions at micro-level (Epstein, 1999).

The second methodology - DEVS, is more conventional (Mansharamani, 1997). DEVS models a system through interaction between entities (e.g. trucks) in a series of events (e.g. load-unload) where they change their state (e.g. empty-full). One helpful way of distinguishing these three modelling methodologies is how they look at the system of interest. SD assumes that a system can be modelled “top-down” - a modeller has enough information to capture its mechanisms from the “bird’s eye view”. Such an approach is useful for models on a large scale: national economy, population growth and so on. On the contrary, CAS operates “bottom-up” - a modeller does not try to model a particular system’s behaviour. Instead, they look for an emergent phenomenon generated by those interactions by defining a simple set of interactions. CAS is often used for small and mid-scale models: virus spread, segregation in a city (Tisue and Wilensky, 2004). DEVS takes a mid-standpoint instead. A modeller has to have a bit of “top” and “bottom” knowledge about the system of interest.

All three modelling methodologies can potentially bring insight into a complex system. Our rationale to choose SD is threefold: (a) the scale of a system - a global supply chain, the information derived from the interviewees, (b) delays, focusing on key actors such as suppliers, customs, etc., and finally, (c) the availability of data - at the time when the study was conducted there was a lack of reliable data about the impact of the pandemic on the key elements of the global supply chain.

In supply chain management, SD has been used for modelling supply chain uncertainty under different conditions including aspects of interest to our study, such as capacity planning (Cheng et al., 2008), behavioural patterns and the related bullwhip effect (Langroodi and Amiri, 2016; Özbayrak et al., 2007), but also humanitarian supply chains (Besiou et al., 2011), medical supply chains (Attridge and Preker, 2000; Mirchandani, 2020), and even to demonstrate the repercussions of a workforce being sick in an epidemic and its impact downstream the supply chain (Kumar and Chandra, 2010).

In this article, we used the results from the qualitative study to construct a conceptual SD model in the form of a causal loop diagram (CLD). The purpose of this model is threefold: (1) connect the main elements of the PPE supply chain into a system, (2) explain the potential causes of disruptions with feedback loops and delays, and (3) propose a set of interventions with mitigation strategies to understand systemic supply chain risk. Our approach for developing the model is as follows. We use existing models on supply chain disruptions as a base, and interview data to guide the further work. Based on the empirical data from the COVID-19 pandemic, we then adjust and adapt the model to this case. To do so, we follow the stepwise approach proposed by Pruyt (2013), where we start by identifying the main problems formulated by the interviewees and end by formulating a dynamic hypothesis explaining how disruptions are generated by the model structure, and how the identified strategies can contribute to mitigate these disruptions. More specifically, we started by mapping the “elements” of the system - the global supply chain of PPE, from the interviewees’ perspective: stock of PPE items, PPE items in production.

The SD methodology allows us to incorporate not only “physical” elements of the system (e.g. stock of PPE items) mentioned by the interviewees but also other, more “conceptual” (e.g. panic buying). Next, we connect those elements with casual relations - how does one element affect another and is this link positive or negative? For example, how panic buying affects the number of items in production. The effect of one element on another can be nonlinear and with a delay: an increase in customs clearance time may result in fewer items in stock. If an interviewee indicated a delay, we added it to the causal loop diagram. The identified elements and casual relations create feedback loops: several connected elements looped on themselves. In case of a conflict of opinion, one interviewee says that the causal link is negative while the other thinks it is positive; we contacted both of them to resolve it. Overall, we intend to create a more comprehensive description of the system. Finally, we search for potential ways to manage the system’s behaviour through a set of mitigation strategies proposed by each interviewee. We specifically focus on the mitigation strategies that are special to the pandemic to estimate their potential effects.

4.1 Empirical results

Our empirical results showed that all four types of disruptions identified from the literature had an impact on the different aspects of the PPE supply chain (see the coding scheme and representative quotes from the interviews in Table A2).

The pandemic itself had a direct effect on the organisations involved in the PPE supply chain through sudden surges (and later fluctuations) in demand, at the same time as PPE producing companies experienced capacity constraints in their production due to their own workforce being sick, or due to lockdowns.

Policy-induced disruptions such as travel bans and export bans had strong negative effects on the transportation, import and export of PPE. The dominant supply chain strategies of single sourcing and lean health care led to further disruptions. Most PPE producers came from China, and when these companies experienced the direct effects of the pandemic (especially sick workforce), the lack of alternative suppliers, and alternative regions whence to procure PPE resulted in a lack of PPE altogether; not being able to meet the global surge in demand. SCRM literature often notes the dangers of single sourcing; what is novel here is the accentuation of not just a single supplier failing, but sourcing from one single region.

Pandemic-induced behaviours further exacerbated disruptions in the PPE supply chain. Surges in demand not only led to surges in prices but also to gaming with those through withholding supplies from the market until prices would rise. At the same time, governments outbid one another for the same PPE, and their panic buying resulted in a lose-lose situation for all. Sadly, this situation gave rise to fraudulent behaviour as well, with fake companies, counterfeit products, and even cargo theft.

Apart from unearthing all these disruptions, our interviews also discovered which mitigation strategies were used to counter them. In the past, typical mitigation strategies to counter supply chain disruptions in epidemics had included pre-positioning (Altay et al., 2009; Comes et al., 2020; Kovács and Spens, 2009; Tang, 2006; Toyasaki et al., 2017), kitting (Jahre and Fabbe-Costes, 2015; Vaillancourt, 2016), multiple sourcing (Yang et al., 2018; Berger and Zeng, 2006) and pre-qualifying (alternative) suppliers (Kumar and Chandra, 2010; Tang, 2006).

We found evidence of all of these mitigation strategies (see Table A3), but also some new nuances to them, and some further ones that were used in the pandemic. In the area of pre-positioning, health care systems that had some PPE inventory fared better in the first wave of the COVID-19 pandemic and bought themselves the necessary breathing time to search for alternative suppliers. Pre-positioning is a common mitigation strategy employed by humanitarian organisations globally, as well as disaster management organisations in their respective countries and is proved to work effectively (Toyasaki et al., 2017).

Kitting proved very useful for emergency health care. While typically health care systems work with pull supply chains and stock items individually, in an emergency, it is important to have all items at hand that are needed in a specific operation. However, PPE kits for different roles differ, and this differentiation was highlighted in the interviews. Kitting itself is a method of standardising what is in a package. Standardisation was even more highlighted due to a current lack of global PPE standards and technical specifications. Sizes, quality requirements, and other technical specifications differ for PPE in e.g. South-East Asia vs. North America vs. Europe. Interestingly, it is the same PPE producers that produce PPE for all these regions in separate lots, later exporting these lots to different environments. A lack of global PPE standards reduced the possibility for using PPE from one geographical region in another, and created further quality control (and compliance) issues around the world if the PPE was delivered did not match the national requirements. Our results indicate that the standardisation of PPE technical specifications would facilitate the response to COVID-19 from production to transpiration to quality compliance.

In the case of face masks, for example, the standardisation of technical specifications does not extend to performance standards like filtration levels alone, but also to labelling, certifications, and quality management systems in manufacturing. Importantly, we refer to the quality requirements on PPE products, e.g. in terms of their required filtering efficiency. There are different ways to employ such a mitigation strategy: either by agreeing on one product technical specifications standard that is globally acknowledged, or by harmonising existing standards and recognising corresponding ones as equals across different countries.

Multiple sourcing and pre-qualifying alternative suppliers are also recommended as mitigation strategies to the disruptions by our data. Single sourcing can be seen as the root cause for many supply side disruptions also in the COVID-19 pandemic. Many organisations were left without medical supplies, because they relied on just one supplier and had to find last minute alternatives. Multiple sourcing with flexible supply bases enables companies to shift production among suppliers promptly (Yang et al., 2018). As our data shows, organisations in the PPE supply chain are now re-considering their sourcing strategies. This is not only to find alternative suppliers, but also specifically for geographical diversification of the supply chain, to ensure that PPE is produced also in different countries (i.e. not only in China), and even on different continents.

An additional, procurement-related mitigation strategy is that of joint procurement. While little of this was evident in the PPE supply chain prior to the COVID-19 pandemic, organisations did eventually come together either nationally, or across similar interests. We consider joint procurement special to the pandemic since it motivated countries like the EU member states as well as organisations like the International Federation of the Red Cross and Red Crescent Societies (IFRC) to engage in joint procurement of essential products to fight the pandemics. IFRC in Geneva, according to one of our interviewees, consolidated orders of PPEs for the whole membership to make sure that the products will arrive at the needed quality and quantity). Later, joint procurement played a bigger role in for COVID-19 vaccines. Overall, joint procurement increases the buyer’s negotiation power, especially under uncertainty (Xianglinga and Ping, 2018). Joint procurement, as well as joint pre-positioning, even as virtual stock, increases the flexibility of the use of items across organisations and regions.

Furthermore, production changeover was used across many countries and companies to support pandemic response. As global PPE supply did not meet global demand during the first wave of the COVID-19 pandemic, increasing global supply became the prime issue. Thereby started a race to set up extra PPE production facilities. As highlighted by our interviewees, there is a need to be able to produce a percentage of critical items locally to ensure that they are available in time of emergency. Reshoring and domestic production became a political mantra in many countries. Thus, governments supported domestic production not only through direct financial support and risk sharing of such investments, but also by fast-tracking regulations and certification processes (Sodhi and Tang, 2021).

Production changeover was used across several industries, e.g. the car industry switching to the production of respiratory ventilators, distilleries to manufacture sanitizers, and within PPE, the fashion, garment and textile industries switching to face masks. For example, in the United States of America, in March 2020, the government invoked the Defense Production Act (DPA) which gives the president the authority to compel the private sector to work with the government to provide essential material goods needed for national defence. Different companies made contracts to produce respiratory ventilators. But since there is no mechanism to drive domestic demand towards those producers once cheaper global options become available, there is the risk of losing resources of this investment.

Air transportation was itself disproportionately hit by the pandemic, especially due to policy-induced disruptions (travel bans). However, many medical items including PPE are frequently transported as belly cargo of passenger planes. Applying production changeover principles, airlines converted passenger planes to cargo planes to be able to support pandemic response. The Project Airbridge in the US is an example of successful production changeover that funded flights to shorten the amount of time it takes for US medical supply distributors to bring PPE and other critical medical supplies into the U.S. during the COVID-19 pandemic response.

In summary, our results highlight both different types of supply chain disruptions and also mitigation strategies for the PPE supply chain. Many of these mitigation strategies can be applied to other types of disruptions, and disasters, as well. However, some of them are either emergency health care-specific (such as kitting), and others are specific to a pandemic situation. Such is the accentuation of a need to standardise or at least harmonise PPE across the globe, and to apply production changeover also to transportation.

Most importantly, however, none of these mitigation strategies can be seen in isolation but they are highly interlinked with one another. Disruptions create incident evolutions and cascades, and mitigation strategies may either stop, or change those. Our findings clearly show that no single mitigation measure will be sufficient to address the different types of supply chain challenges encountered. Rather, we argue that a coordinated intervention across the different options is needed. In the following section, we will analyse the interdependencies of the different types of SC disruptions to identify feedback loops and identify the impact of the mitigation measures in the dynamic SC system.

4.2 A system dynamics model of PPE supply chain disruptions

We further analyse the data and the literature to understand the links between the various causes of PPE supply chain disruptions and propose a SD model that captures these causal mechanisms during the first wave of the COVID-19 pandemic. As PPE is primarily produced in China and then exported to the rest of the world (ROW), the modelling is done from the perspective of a Chinese-ROW PPE supply chain. In essence, the model focuses on the supply chain from final manufacturing and assembly (here of medical respiratory face masks in China) to use in health care centres in the rest of the world, with air transportation as common for PPE. Figure 3 depicts the key elements of the PPE supply chain. While a supply chain may have more elements and is more complex, we have focused on those highlighted by our respondents: from the PPE producer to PPE delivery to their users - hospitals. To be more comprehensive, we also highlight two other nodes critical to the global supply chain of PPE but left out of the scope of the study: raw materials and disposal.

The first model presents the baseline performance of the global supply chain system during the first wave of the COVID-19 pandemic in the rest of the world (with data from Apr–Sep 2020) is in Figure 4. It consists of 19 elements: system’s variables or stocks (plain text), connected into a system with causal links, four main feedback loops (italic in bold): COVID-19 outbreak, Lack of items, Lack of transportation capacity and Global supply chain disruption, and four policies imposed by different actors (text in a rectangle). Altogether, the model’s structure generated the undesirable behaviour that led to a lack of PPE in the country of the outbreak aka “customer country” (ROW).

The outbreak itself, i.e. the COVID-19 outbreak feedback loop, is at the core of the model (see Figure 4). It is a negative (balancing) explaining the rise in the PPE demand given the rise in infections. Given a low initial number of PPE in stock and without a policy to tackle the virus spread (e.g. a lockdown), this feedback loop will seek to maximise the country’s infections until it reaches a goal equal to a certain percentage of the population. With an increase in Items in use, the Stock of items decreases, and it initiates the order placement represented with (Quality) orders stock. On the other hand, Stock of items is influenced by a key policy that was employed by various countries in the COVID-19 pandemic: an export ban. Such policies were implemented to counteract the quick depletion of the extant stock and provided a temporary effect.

Pandemic-induced behaviours (panic buying) significantly impacted how orders were placed. They lead to a perceived rush in procuring the surge capacity needed in the outbreak area. Another affected critical aspect is order’s “quality”. A quality order in the model represents the one that fits the needs of the healthcare centre, takes all necessary steps of quality compliance, and complies with standards and regulations. Interviewees highlighted that they had spent less time on the Pre-order quality check, which often led to a decrease in order quality. In the case of face masks alone, that would extend to any batch delivered to, e.g. EU countries complying with FFP standards and their required levels for use in health care (with FFP3 for intensive care units, surgical masks in health care otherwise). Other PPE included gloves, eye protection, hazmat suits and powered air-purifying respirators. Notably, the requirements of what was supposed to be used were changed several times during the first wave of the COVID-19 pandemic.

The supply side was affected by the pandemic due to both the direct effects of the outbreak, and due to policy-related disruptions. Sick or infected Factory personnel could not participate in the manufacturing process anymore. In addition, workers could not get back to factories after the Chinese New Year due to Local travel restrictions (lockdowns) imposed by the government. Since factories were often operating at maximum capacity due to the number of orders received, further orders were adhered to with a delay. Such delays propagated throughout the supply chain, with customers receiving orders later than anticipated.

Further, we highlight the supply chain strategy-related disruptions. Especially the number of signed Contracts with PPE manufacturers is a matter of single vs multiple sourcing. They impact both on the quantities produced as well as the potential to add surge capacity elsewhere, if personnel at the single source supplier is sick, for example.

Produced items still need to be delivered to their destination, however. A surge in deliveries leads to an increase in transportation orders. The bottleneck here is modelled in the lack of transportation capacity feedback loop, with remaining transportation capacity decreasing with further orders, but also being affected by international travel bans and their related decrease in the number of passenger flights. The latter is significant since medical items including PPE are often transported in those.

Further steps remain before the PPE is delivered. Here, customs clearance time increases with the surge in deliveries, negatively impacting on overall delivery times. Any bullwhip effect, and any mistakes induced by panic buying will play out as critical delays and potentially mismatched items in the overall global supply chain disruption feedback loop.

Overall, the model represents a system’s perspective at representing global PPE supply chain disruptions. First, the direct impact of the pandemic limits production capacity. Workforce that has fallen sick cannot participate in the production process anymore and forces the country to impose the lockdown, which in turn will make it impossible for other workers to get back to factories. The single-sourcing strategy further increases dependence on specific suppliers. The thereby reduced capacity with limited potential to rapidly extend production meets an increased demand as countries and people prepare themselves for the pandemic. In combination with pandemic-induced behaviour (e.g, panic buying), this creates a sustained demand shock with limited capacity. The increased pressure on customs combined with low-quality orders further delays delivery of critical medical goods. This leads us to Proposition 1:

Further, we observe that transport is affected by both: the pandemic and subsequent policies to respond to it. Here, we find that the policies that are designed to protect people or countries on the short-term from the virus create bottlenecks on the longer-term further down in the supply chain, for instance via the combination of travel and export bans. The final complication is caused by the mismatch between ordered and delivered items, driven by the occurring delays, rushed orders and the lack of common global PPE standards.

4.3 Analysing the effects of mitigation strategies for PPE supply chain disruptions

In this section, we present the mitigation strategies that were highlighted as specific to the pandemic and estimate their effects on PPE supply chain disruptions: introducing global standards for PPE, joint PPE procurement, and using production changeover to convert passenger planes to cargo flights. In addition, we added multiple sourcing. These mitigation strategies are noted in blue in Figure 5. Below, we present the causal mechanisms between the selected measures and how they reduce disruptions, before developing propositions on mitigation strategies.