Modeling performance indicators of resilient pharmaceutical supply chain Open Access

Over the past few decades, corporate world has become more turbulent and vulnerable in nature due to rapid globalization, varied customers' interest and demand, technological changes, unexpected pandemic outbreaks etc. Firms are facing enormous pressure to manage and control their supply chains effectively. To secure the market position, business organizations are searching for proper strategies that may enhance their strengths in some key areas of their supply chains and operations. Establishing resilient supply chain is one of the fruitful strategies to overcome the adverse situation tactfully as well as to improve the overall business performance (Rajesh, 2018). Over the past few years, a lot of academicians and industry practitioners have shown much interest for the establishment of resilient supply chain (Chowdhury et al., 2019). Moreover, recently, a novel and wide-ranging disruptive pandemic called COVID-19 outbreak has put resiliency into the spotlight. This pandemic has hit the global supply chains adversely. Amid such turbulence, this outbreak has provided organizations a golden opportunity to revisit the strategy and accelerate their capabilities for the adaption of long-term resiliency in their network for managing future challenges (Ivanov, 2020).

A resilient supply chain can sidestep avoidable risks and has the potentiality to bounce back fast from uncertain occurrences. It is to be noted that more than 80% of global firms are now interested for the adoption of resilient supply chain (Bhatia et al., 2013). However, after establishing resilient supply chain, it is required to monitor and measure the performance of supply chain resilience continuously. It is essential for the business organizations to establish relevant and significant performance indicators (PIs) to measure the resilience of the supply chain (Behzadi et al., 2020). Researchers and industry practitioners have performed many studies to identify the most pertinent PIs for measuring supply chain resilience (Hohenstein et al., 2015; Bai and Satir, 2020). Ali et al. (2017) identified some non-financial key performance indicators (KPIs) for measuring supply chain resilience. Stone and Rahimifard (2018) reviewed 137 papers on supply chain from various renowned databases. Their findings also showed that most of the previous literatures were focused only on identification of the PIs of resilient supply chain. However, the companies and supply chain managers require not only the identification but also the understanding of relationships among these PIs to establish better performance evaluation strategy for resilient supply chain. To the best of our knowledge, the resilient supply chain PIs have not been modeled yet to better understand the relationships among them.

As an emerging sector of Bangladesh, pharmaceutical industry has also focused for the proper adoption of resilient supply chain concept. Daily commodity products like rice, oil, electricity etc. have almost simple ad static supply chains. On the other hand, having complex and dynamic structure, PSC face various challenges to establish resilience in the chain. In the context of Bangladesh, this sector plays a vital role providing lifesaving medicines. However, it becomes very difficult to maintain the smooth flow of medicines because huge population of Bangladesh causes the constant fluctuations in customer demand. So, in order to maintain the smooth flow of medicines from producers to customers amid such difficulties, a resilient supply chain is inevitable. There are some significant PIs that, if improved upon, will help the decision makers in making a supply chain more resilient. So, motivated by the current situation, this study proposed a structured framework for modeling the PIs of resilient PSC. However, after reviewing the previous literature and analyzing the research gap, this study raises the following research questions (RQs):

Based on the above research questions, this study has formulated the following specific research objectives (ROs):

To achieve the above objectives, this study has deployed the Delphi based fuzzy DEMATEL approach to identify and analyze the most significant PIs of PSC and developed a SD model for predicting the resilience of the supply chain. The Delphi based fuzzy DEMATEL approach has the advantages of extracting the research data from the experts based on their prior experiences and developing cause and effect diagram along with prominence rank under fuzzy environment (Abdullah and Zulkifli, 2019). After examining the cause and effect relationships among the indicators, SD model is developed. Therefore, integrating fuzzy DEMATEL with SD helps to construct more robust models.

This section presents the review of previous literature for building a theoretical foundation of this study.

Over the last few decades, the researchers and supply chain practitioners are working out to establish resilient supply chain to deal with uncertainty and disruptions (Pettit et al., 2019). This concept of supply chain resilience has gained attention because it deals with designing the supply chain to enhance the capability of retuning to its original state from unexpected events or disruptions (Christopher and Peck, 2004; Ralston and Blackhurst, 2020). According to Fiksel et al. (2015), supply chain resilience is an essential proactive approach which helps the companies to reduce traditional risks. Resilient supply chain has the ability to identify the risks, ameliorate the impact and come back rapidly from man-made or natural occurrences (Singh et al., 2019a, b; Lee and Rha, 2016). In today's dynamic market, establishing resilience in the supply chain has become one of the utmost challenges for supply chain specialists and managers.

PSC is considered to be more complex than the other industries (Zahiri et al., 2017). Nowadays, supply chain resilience has become a widely used area of interest among researchers and supply chain practitioners of pharmaceutical companies. In the supply chain resilience literature, various quantitative approaches have been previously applied by researchers for resilience assessment (Hossain et al., 2019; Yaroson et al., 2019). Pettit et al. (2013) proposed a tool for supply chain resilience assessment and management which was called the SCRAM tool. Another assessment in the literature is the fundamentals of the grey systems theory to evaluate and analyze the supply chain resilience (Rajesh, 2020).

In the recent studies, Rajesh (2018) used VIKOR method for assessing supply chain resilience. Aigbogun et al. (2018) proposed another quantitative cross sectional design for resilience assessment. Another study was conducted with the integrated method of ANP and DEMATEL by Leksono (2019). Ahmed et al. (2020) developed a framework to model the barriers of closed loop supply chain. A logical framework for assessing supply chain resilience based on PIs was provided by Song et al. (2019). The proposed framework provided an insight about resilient supply chain by representing linkages between some indicators and a set of performance measures. Banihashemi et al. (2019) explored the relationship between reverse logistics and sustainability performance.

Many researchers have already identified a number of PIs for resilient supply chain. To find out the PIs in the context of resilient PSC, relevant articles were reviewed from various databases (Google scholar, Scopus etc.)using key words such as “supply chain resilience”, “supply chain resilience performance indicators”, “pharmaceutical supply chain performance indicators” etc. After reviewing the literature, we found 19 most relevant supply chain resilience PIs for further analysis. Table 1 represents these 19 PIs with definition. Many researchers emphasized on supply chain collaboration to be an important PI for supply chain resilience although it is very difficult to implement (Christopher and Peck, 2004; Papadopoulos et al., 2017). Some researchers have found that supply chain visibility plays a significant role in establishing resilient supply chain (Swift et al., 2019). Hafezalkotob and Zaman (2019) identified sustainability as an important PI for resilient supply chain. Production capacity also determines the performance of a resilient supply chain (Chowdhury and Quaddus, 2016; Jain et al., 2017; Kaur et al., 2020).

For the past few decades, researchers have contributed a lot in the field of resilient supply chain performance measurement. All the research studies provide resilience supply chain PIs for overall organization. However, identification of proper PIs and modeling those indicators to assess supply chain resilience for pharmaceutical industries are still required further investigations. Therefore, the goals of this study are to identify and analyze the appropriate PIs and to develop a model for predicting the resiliency level in the context of Bangladeshi pharmaceutical company. To address these research gaps, the contributions of this study and problem definition are stated as follows:

1. Finding the major PIs in the field of pharmaceutical industry supply chains.

2. Evaluating the ranking of these indicators according to their influence and identifying the cause and effect relationship.

3. Building a SD model for predicting the resiliency level.

In this section, the details of the proposed methodology for analyzing the PIs and assessing the resilience of PSC have been explained. Figure 1 shows the flow diagram of the proposed research framework.

In this study, the Delphi method has been applied to identify the significant PIs of resilient PSC in two phases after reviewing the previous literature. It is an iterative process of collecting and assessing data from a group of experts through a series of structured questionnaires (Jason and Glenwick, 2016; Lee and Seo, 2016). This tool is extensively used in supply chain, demand forecasting and complex decision making. (Belton et al., 2019; Moktadir et al., 2019).

In phase 1, most relevant indicators of resilient PSC were selected by the industrial experts using primary questionnaire ( Appendix 1) and in phase 2, evaluation of comprehensive relationships among these indicators was done by evaluators using the secondary questionnaire ( Appendix 2). Using the judgmental sampling technique, 10 pharmaceutical companies of Bangladesh were selected that were trying to identify and evaluate the PIs for improving their performance level. Due to confidentiality, symbolic names were used instead of the real names of these companies. For phase 1, 20 respondents were selected using purposive sampling method (Guarte and Barrios, 2006). The profile of the case pharmaceutical companies and the respondents are represented in Table 2.

Initially, a total of 19 PIs of resilient PSC were identified after reviewing previous literature resources (Table 1). The primary questionnaire ( Appendix 1) was sent to the respondents through email to select the most relevant PIs from these 19 PIs. The purpose of the study was communicated with them and they were requested to provide their feedback to check the validity of those PIs for PSC. They were given the chance to add or remove any indicator of resilient PSC. Based on the experts' opinions, a total of 12 PIs were sorted out. Table 3 presents the identified PIs through Delphi method.

The fuzzy theory was proposed and introduced by Zadeh (1965). This tool mainly deals with the vagueness problems of linguistic variables in reality. It has the capability of assessing the subjective and imprecise judgments provided by individuals or expert panels. In this study, a widely used and well- established triangular fuzzy number (TFN) has been considered. Before using fuzzy based tools, fuzzy numbers are required to convert into crisp values (defuzzification).Among various methods, Converting Fuzzy data into the Crisp Scores (CFCS) is extensive in terms of providing better crisp values than other similar methods (Wu and Lee, 2007). According to membership functions, the total score can be found as a weighted average. Let, $Aij=(uijn,vijn,wijn)$⁠, mean the degree of criterion $i$ that affects criterion $j$ and fuzzy questionnaires $n$ (⁠$n=1,2,3,…….., h)$⁠. The crisp value of ith criteria can be obtained by four steps as follows (Opricovic and Tzeng, 2003):

• Step 1. Perform normalization

where $Δminmax=maxwijn−minuijn$⁠;

• Step 2: Calculate the right (⁠$w$s) and left (⁠$u$s) normalized values

• Step 3: Estimate total normalized crisp values

• Step 4: Compute crisp values

The Decision Making Trial and Evaluation Laboratory (DEMATEL) approach is a simple method for evaluating the relationship among various alternatives based on the pair-wise comparison. This method was first proposed by Fontela and Gabus (1974). It was widely used in analyzing complex structural models of causality between complex factors. Integrating fuzzy systems with DEMATEL enables the decision-makers to evaluate the complex factors easily (which are difficult to quantify and usually expressed with linguistic variables). Table 4 shows the triangular fuzzy linguistic scale for direct relation criteria. The steps of fuzzy-DEMATEL are presented as follows (Kazancoglu et al., 2018):

• Step 1: Development of the fuzzy initial direct relation matrix $A$ by integrating the inputs from experts.

• Step 2: Development of crisp direct relation matrix $Z$ using Eqns (1)–(7).

• Step 3: Construction of standardized matrix $X$ using the following equation

where $S=max{max∑j=1nZij,max∑i=1nZij}$

• Step 4: Total relation matrix $T$ construction using the following equation

where $I$ indicates the identity matrix.

• Step 5: Then the degree of influential impact $(D)$ and the degree of influenced impact $(J)$ are calculated as follows:

• Step 6: In this step, $(D+J)$ and $(D−J)$ are calculated. $(D+J)$ indicates the prominence vector and $(D−J)$ indicates the relation vector.

• Step 7: Finally, the causal diagram is developed using the prominence vector and relation vector.

For verifying the proposed methodology, a pharmaceutical company located in Tangail, Bangladesh was selected to analyze the PIs and predict the resilience level of the supply chain. Due to confidentiality, the name of the case company has been kept anonymous. For the assessment of the relationships among the performance indicators using fuzzy DEMATEL for the case company, 5 evaluators were selected. In fuzzy DEMATEL, 4–12 evaluators are suitable for data collection. In this study, five evaluators were selected to keep the calculation simple. They were selected based on their working experience and positions in the relevant department. The profile of these five evaluators has been shown in Table 5. The questionnaire (Table A4) was sent through email after explaining to them the objectives of the study.

At this stage of the study, performance indicators of resilient PSC of the selected pharmaceutical company were assessed using the fuzzy DEMATEL method. The resilience performance indicators represented in Table 3 were selected for the case company. Then, through the secondary questionnaire ( Appendix 2), the data were obtained from the five professional evaluators for fuzzy DEMATEL. The linguistic scale shown in Table 4 was used through the one-to-one interview for data collection. The fuzzy DEMATEL method was applied through the following steps.

At first, the pair wise comparisons among the resilience PSC indicators were constructed by using the fuzzy linguistic variables. The linguistic assessment by the five evaluators is shown in Table A5 in  Appendix 3. These linguistic variables were then converted into triangular fuzzy numbers and fuzzy initial direct relation matrix was developed by taking the average of these fuzzy numbers. The fuzzy initial direct relation matrix is shown in Table A6 in  Appendix 3.

The crisp direct relation matrix from the average fuzzy initial direct relation matrix was computed using Eqn (1)–(7). The total relation matrix was derived from the crisp relation matrix using Eqns (8) and (9). The total relation matrix is shown in Table A7 in  Appendix 3.

The degree of influential impact $(D)$ and the degree of influenced impact $(R)$ were calculated from the total relation matrix using Eqns (10) and (11). Finally, the prominence vector and relation vector were obtained. Prominence vector and relation vector along with the ranking were represented in Tables 6 and 7 respectively. To develop the cause and effect diagram, the relation vector was used. All the indicators having positive relation values $(D−R)$ were grouped into cause group and indicators with negative relation values were grouped into an effect group.

In this section, the findings from the implementation of fuzzy DEMATEL have been discussed. The $(D+R)$ scores indicate the relative importance of the resilient PSC performance indicators. Therefore, a higher score of the indicators represents higher priority in the indicator's ranking system. Table 6 represents the prominence vector of the resilience PIs obtained from the fuzzy DEMATEL. According to the prominence vector, the top five performance indicators are “Collaboration (PI₁)”, “SC risk orientation (PI₂)”, “Visibility (PI₉)”, “Flexibility (PI₇)” and “Agility (PI₁₂)” which were ranked based on $(D+R)$ scores. “Available resource (PI₁₁)”, “Adaptability (PI₁₀)”, “SC network design (PI₃)”, “Velocity (PI₅)” and “Robustness (PI₄)” are the last five PIs in the prominence ranking although they have significant impact on the resilience of PSC.

The cause and effect relationships of these PIs of resilience PSC can be understood from the relation vector (Table 7). The $(D−R)$ score represents the relation characteristics of the indicators. The positive $(D−R)$ score indicates the cause group and negative $(D−R)$ score indicates the effect group. According to the relation vector, “IT capability (PI₈)”, “Flexibility (PI₇)”, “SC network design (PI₃)”, “Available resource (PI₁₁)”, “SC risk orientation (PI₂)” and “Velocity (PI₅)” were grouped into cause category since they have positive $(D−R)$ scores. On the other hand, “Sustainability (PI₆)”, “Agility (PI₁₂)”, “Collaboration (PI₁)”, “Creating robust supply chain (PI₄)”, “Visibility (PI₉)” and “Adaptability (PI₁₀₎” were grouped into effect category. All these six indicators are influenced by other performance indicators. Figure 2 shows the cause and effect relation of the resilient supply chain performance indicators.

“Information sharing (PI₈)” has got the first position in the cause category. This indicator has significant impact on establishing resilient PSC. Lack of proper information sharing causes inefficiency of coordinating action and disruptions in overall supply chain (Yu et al., 2020). It is very important to exchange the right information, for increasing resilience within the supply chain (Faisal et al., 2006). Colicchia et al. (2019) also found information sharing as one of the significant performance indicators of supply chain. To maximize resilience in the PSC, it is important to make a community of active partners within the current, complex, indeterminate supply chain environment and proper information should flow between all partners of this particular group (Setak et al., 2018; Tohidi et al., 2017). The top management should take action to concern about their information sharing to build resilient PSC. The second indicator in the causal group is “Flexibility (PI₇)”. Lack of flexibility causes fluctuation in lead times and delays in product quality and information (Giannoccaro et al., 2003). The numerous styles of flexibility may boost the resilience of PSC, together with versatile travel, versatile work game plans (Pettit et al., 2013). “SC network design (PI₃)” is another important causal indicator for resilient PSC. Poor network design is responsible for production of short deliveries, poor quality issues and delays. It is important for a proper understanding of the architecture of the supply chain network to make supply chain resilient (Giannoccaro et al., 2018). Rajesh (2019) performed a study on resilient supply chain and showed that network design has a significant influence on the performance of the resilient supply chain. “Available resources (PI₁₁)” is the key to manage the resilience in the supply chain. When resources are available, it makes it easier to mitigate any disruption. “SC risk orientation (PI₁₁)” is another causal indicator in case of PSC. However, risk orientation in the pharmaceutical supply chain is very difficult. An organization should take necessary action to identify and understand supply chain risks and take proper mitigation actions.

The effect category includes six indicators that can be highly affected by the factors in the cause group. “Sustainability (PI₆)”, and “Agility (PI₁₂)” hold the first and second position of the effect group respectively and they are highly interrelated. Lack of information sharing, limited resources, product complexity and environmental uncertainty affect these two factors. Different policies and management strategies should be taken by supply and operations managers to increase SC risk orientation and increase flexibility. Another crucial indicator is supply chain “Visibility (PI₉)”. It can be affected by supply chain network design, flexibility, supply chain adaptability and collaboration in supply chain. “Adaptability (PI₅)” is the last indicator in the effect group. The stability of the supply chain depends on the flexibility of the system to respond to transitory problems and market fluctuations (Chowdhury and Quaddus, 2016). Supply chain and operations managers should formulate strategy to improve the supply chain resilience by setting target for these performance indicators and following proper action plan.

System dynamics (SD) model has been developed for predicting the PSC resilience after identifying the cause group and effect group from the fuzzy DEMATEL. The system dynamics (SD) model was constructed in two phases. In the first phase, cause group and effect group were translated into partial causal loop diagram (CLD) and stock and flow diagram was developed from the causal loop diagram in the second phase. These phases are described as follows:

A causal loop diagram (CLD) of PSC resilience PIs was developed based on the cause and effect group derived from the fuzzy DEMATEL and some other auxiliary variables. These auxiliary variables were selected with the help of experts in Delphi phase. These variable were selected based on their prior experiences. Auxiliary variables are related to the main variables and the values of auxiliary variables are required to predict the values of stocks and flows. The values of auxiliary variables were collected from the case company. The eight most influential resilience PIs from the prominence vector (Table 7) were selected for developing the causal loop diagram. Table 8 presents the list of these resilience PIs and auxiliary variables.

For developing the CLD, Vensim PLE version 32 software has been used. The causal loop diagram is shown in Figure 3.

At the next step, stock and flow diagram was developed with the information from the CLD. Figure 4 shows the stock and flow diagram for predicting the resilience level of PSC. Here, the “SC Resilience” is the main stock. The other stocks in the model are “SC Agility”, “Sustainability”, “Lack of collaboration” and “Lack of visibility”. Flows and the auxiliary variables have been created according to the information from the CLD.

The equation was set for each flow in the model by consulting with the experts. Then, this model was implemented to the case pharmaceutical company for predicting the resilience level of the supply chain. At first, the initial values of each stock were given after consulting with the case pharmaceutical company's top management. Then, the target values of each auxiliary variables were given in the model. This data was given by the supply chain department of the case company and they set the values of each variable ranges between 0 and 1 considering the current condition of each variable. Then, the simulation was run for 100 months for observing the resilience level for the case pharmaceutical company. The equations and initial values used for different variables in the model are given below:

Figure 5 shows the predicted conditions of supply chain agility, sustainability, collaboration and supply chain visibility. It is noted that if the case pharmaceutical company takes action according to their target, then their supply chain sustainability will improve rapidly. However, supply chain agility will improve very slowly. Besides this, lack of collaboration and lack of visibility will be minimized within the next 100 months.

Figure 6 shows the predicted resilience level of the case pharmaceutical company for the next 100 months. At present, the resilience level of the supply chain of the company is 20 which means very low. The curve indicates that the level of supply chain resilience will be improved very quickly after a certain time if the company can maintain their strategy and action plans.

Sensitivity analysis is usually performed to examine the reliability of the decisions made by the evaluators (Tanino, 1999). Sensitivity analysis is conducted by keeping equal weights and adjusts more weights to any evaluator to realize the effects of using various combinations of decision criteria weight (Goodridge, 2016). In this study, the results obtained from fuzzy DEMATEL have been verified by performing sensitivity analysis. Five scenarios were assumed to conduct this sensitivity analysis. Different weights of evaluators were given based on their working experience in different scenarios which is shown in Table 9.

Initially, each evaluator was given equal weight (Scenario-1). After that, the weights of the evaluators were changed based on their working experiences to analyze how much the prominence rank, and the cause and effect relations vary. Since Evaluator 1 and Evaluator 2 were more experienced than all other evaluators, their weights were assigned higher than other evaluators in the next three scenarios. Then the cumulative average initial matrix was derived shown in appendix D4. Then, the fuzzy DEMATEL was applied on the same data in each scenario to observe the reliability of the results. Figure 7 shows the results obtained in different scenarios. Prominence rank of the resilient PSC performance indicators was stable in different scenarios although the weights of the evaluators were varied. Accordingly, the cause and effect relations among the resilient PSC performance indicators were varied a little with insignificant impact in each scenario. Cause group and effect group hold the same members in each scenario. Therefore, it is clear that the obtained results are not much sensitive to the variation of weights among the evaluators which indicates the high reliability of the results of the current research. As a result, the understanding of evaluators about the resilient pharmaceutical supply chain is adequate for this study.

The major contribution of this work is the identification and prioritization of PIs of resilient PSC, which are significant to increase resilience as well as to improve overall performance. Another contribution of this work is a proposed SD model that can be used to measure resilience of PSC by simulating performance level for a certain time. After the recognition of fundamental knowledge and the strengthening of resilient supply chain PIs and the result of simulation, the business manager will be able to enhance resilient supply chain performance by introducing constructive supply chain policies. This research will help policymakers to identify the most relevant metrics of success and recommend ways to improve the outcomes in PSC. Some of the policies are also recommended for helping managers to improve PSC resilience, given as below:

1. Improving the resilience of a PSC requires understanding and selecting the appropriate PIs of the resilient supply chain to optimize the efforts. This research will be helpful for the supply chain professionals and researchers to select the appropriate indicators and understand the relationships among them to improve the resilience of a supply chain.

2. The pharmaceutical sector must formulate an effective strategy to improve the resilience of the supply chain. However, the effectiveness of the strategy must be assessed and the resilience level of the supply chain must be predicted on a regular basis. The fuzzy-DEMAEL part of this study will help the supply chain managers to formulate an effective strategy and the proposed SD model will help to predict the resilience level of the PSC.

Establishing resilience in the supply chain has become one of the major challenges for the pharmaceutical industries. However, a structured framework is required to identify the PIs for developing resilient PSC successfully. Motivated by the drawbacks of previous research as well as the current pandemic situation, this study presents an integrated approach (Delphi based fuzzy DEMATEL and SD model) for assessing the contextual relationships among the indicators and predicting the resilience level for a certain time period.

There are three major findings in this study. First, from the Delphi method, 12 most relevant PIs for resilient PSC were identified. Second, this study applied Fuzzy DEMATEL to determine the causal relationships among the identified PIs. This study contributes to the existing supply chain literature by exploring the relationships among the PIs of resilient PSC and developing an SD model to predict the resilience level. Here, the values of the elements of the developed SD model were set according to the assumptions of the top management of the case company. The model can be more robust if the real value for each element of the model is set.

Although this research indicates new insights into resilient PSC indicators, the research inevitably has limitations regarding data collection and validation. As in this research work, only 12 PIs are considered. More indicators can be analyzed to develop the model using fuzzy DEMATEL and system dynamics in further analysis. As the model is developed and verified on the basis of expert's opinions, biases can occur. The results of this study can be compared with other methods such as fuzzy TISM, grey based TISM etc. While the focus of this research is mainly on the PSC of Bangladesh, the proposed model can be applied to other industries, such as beverage, construction, manufacturing, service and so on, for identifying resilient supply chain performance indicators.