key: cord-1046212-zve89gga authors: Alsalem, M. A.; Alsattar, Hassan; Albahri, A. S.; Mohammed, R. T.; Albahri, O. S.; Zaidan, A. A.; Alnoor, Alhamzah; Alamoodi, A. H.; Qahtan, Sarah; Zaidan, B. B.; Aickelin, Uwe; Alazab, Mamoun; Jumaah, F. M. title: Based on T-spherical Fuzzy Environment: A Combination of FWZIC and FDOSM for Prioritising COVID-19 Vaccine Dose Recipients date: 2021-08-26 journal: J Infect Public Health DOI: 10.1016/j.jiph.2021.08.026 sha: e7523cdb05873d21db99d3882332ab82c1291c35 doc_id: 1046212 cord_uid: zve89gga The problem complexity of multi-criteria decision-making (MCDM) has been raised in the distribution of coronavirus disease 2019 (COVID-19) vaccines, which required solid and robust MCDM methods. Compared with other MCDM methods, the fuzzy-weighted zero-inconsistency (FWZIC) method and fuzzy decision by opinion score method (FDOSM) have demonstrated their solidity in solving different MCDM challenges. However, the fuzzy sets used in these methods have neglected the refusal concept and limited the restrictions on their constants. To end this, considering the advantage of the T-spherical fuzzy sets (T-SFSs) in handling the uncertainty in the data and obtaining information with more degree of freedom, this study has extended FWZIC and FDOSM methods into the T-SFSs environment (called T-SFWZIC and T-SFDOSM) to be used in the distribution of COVID-19 vaccines. The methodology was formulated on the basis of decision matrix adoption and development phases. The first phase described the adopted decision matrix used in the COVID-19 vaccine distribution. The second phase presented the sequential formulation steps of T-SFWZIC used for weighting the distribution criteria followed by T-SFDOSM utilised for prioritising the vaccine recipients. Results revealed the following: (1) T-SFWZIC effectively weighted the vaccine distribution criteria based on several parameters including T = 2, T = 4, T = 6, T = 8, and T = 10. Amongst all parameters, the age criterion received the highest weight, whereas the geographic locations severity criterion has the lowest weight. (2) According to the T parameters, a considerable variance has occurred on the vaccine recipient orders, indicating that the existence of T values affected the vaccine distribution. (3) In the individual context of T-SFDOSM, no unique prioritisation was observed based on the obtained opinions of each expert. (4) The group context of T-SFDOSM used in the prioritisation of vaccine recipients was considered the final distribution result as it unified the differences found in an individual context. The evaluation was performed based on systematic ranking assessment and sensitivity analysis. This evaluation showed that the prioritisation results based on each T parameter were subject to a systematic ranking that is supported by high correlation results over all discussed scenarios of changing criteria weights values. Countries worldwide faced the greatest challenge last year brought by the coronavirus disease 2019 (COVID-19) pandemic [1] [2] [3] [4] [5] [6] [7] [8] , and the need for a vaccine has become more important than ever [9] . Thus, many companies have succeeded in developing vaccines to stop the disease [10] . However, the limited available doses might not cover the current needs of all populations and could lead to anxiety amongst such populations [11] . Considering these limited vaccine doses, some fear that the allocation will not be fair at the global (between countries) and local (amongst different groups of society) levels where the evaluation of vaccine distribution has become a complex problem and the state of vaccine progress is unclear [12] . Therefore, governments must follow a priority mechanism for allocating COVID-19 vaccine doses amongst the population and avoid randomisation of vaccine distribution [13] . Equity and fairness considerations are high priorities in healthcare policy discussions and have become an important global responsibility [14] . To support the community with a The first phase of the proposed methodology is presented to discuss the decision matrix used in the process of COVID-19 vaccine distribution. Reference [13] formulated the decision matrix based on three sequential steps including criteria identification, vaccine recipients as alternatives and data generation of alternatives to provide an artificial dataset of vaccine recipients cases (Table 1) In this decision matrix, the mechanism of vaccine distribution is achieved to serve the vaccine recipients who represent the alternatives based on five criteria, namely, vaccine recipient memberships, chronic disease conditions, age, geographic locations severity and disabilities. After that, an adequate augmented dataset is adopted from [13] . In this adopted dataset, 300 cases of vaccine recipients were generated as proof of concept. Although the generalisation and inclusion of more than 300 cases are possible, the insights from the generated cases usually can satisfy the concepts of the presented work, from which the results can then meet the desired goals. A coding scheme using the exception-handling model was developed in Python to generate the augmented dataset of the 300 cases based on the five discussed criteria. The most suitable probabilities and certain assumptions about COVID-19 vaccine alternatives were generated. In that date set, the rule-based control scheme was based on expert opinions with precise descriptions for the criteria. After generating the dataset, a panel of three experts subjectively validated it to increase the veracity of the data to the best extent possible and cover most recipients' situations. The three expert panellists were identified and selected from related study areas (i.e. molecular biology, immunology, biomedical engineering, medical biotechnology and clinical microbiology). According to the same expert panel, C3 and C4 (age and geographic locations severity) have ranges of measures and are considered benefit criteria. Moreover, other criteria belong to the categorical type. Lastly, this decision matrix is introduced to the next phase (development) to start with the distribution process. This phase presents the development of the proposed vaccine distribution methodology based on new extensions of MCDM methods. T-SFWZIC was used to achieve the criteria weighting, whereas T-SFDOSM was used for ranking the vaccine recipients. The following subsections describe each method separately along with the relevant mathematical expressions. The T-SFWZIC method is an extension of the original FWZIC [85] , which has five steps for criteria weight determination ( Figure 1 ). The following show the complete details of the five steps: Step 1: Criteria Definition This step has two processes. The first process is the exploration and presentation of the predefined set of evaluation criteria, and the second process is the classification and categorisation of all the collected criteria. As discussed before, the criteria used in the process of COVID-19 vaccine distribution are identified in Section 2.1. Furthermore, the defined and selected criteria were evaluated by the same panel of experts (those in phase 2.1), which is explained in the next step. Step 2: Structured expert judgement (SEJ) To evaluate and define the level of importance for the criteria identified, the same panel of three experts was utilised. After exploring and identifying the list of prospective experts, the selection and nomination commenced, and the SEJ panel was established. Lastly, an evaluation form was developed to obtain the consensus of all the SEJ panellists for each criterion, followed by the conversion of the linguistic scale to its equivalent numerical scale. a) Identify experts: Anyone who has knowledge about a subject cannot be considered an expert. Instead, an 'expert for a given subject' is used here to designate a person whose present or past field involves the subject in question and who is regarded by others as knowledgeable about the subject. These individuals are occasionally designated in the literature as 'domain' or 'substantive' experts to distinguish them from 'normative experts', that is, experts in statistics and subjective probability. In the current study, the expert selection method was based on a bibliometric analysis of all authors and co-authors of studies that have listed vaccine distribution criteria. b) Select experts: After identifying the set of experts, the experts who will be involved in the study were selected. In general, the largest number of experts consistent with the level of resources should be used. In this study, three experts were chosen for a given subject. All potential experts named during the expert identification phase were contacted via email to determine whether they were interested and whether they J o u r n a l P r e -p r o o f considered themselves potential experts for the panel. After the list of candidate experts was established, the three experts collaborated as expert judgement panellists. c) Develop the evaluation form: The development of an evaluation form is a crucial step because this instrument is used to obtain expert consensus. Before finalising the evaluation form, the questionnaire underwent reliability and validity testing, and all the three experts selected in the previous step reviewed it. d) Define the level of importance scale: In this step, the selected group of three experts defined the level of importance/significance of each criterion using a five-point Likert scale. No theoretical reason exists to rule out different lengths of the response scale [86] . The options reflect an underlying continuum rather than a finite number of possible attitudes. Various lengths ranging from 2 to 11 points or higher are used in surveys. Five has become the norm in Likert scales probably because this number strikes a balance between the conflicting goals of offering sufficient choices (as providing only two or three options means measuring only the direction rather than the strength of opinion) and making things manageable for respondents (few people have a clear idea of the difference between the eighth and ninth points in an 11-point agree-disagree scale). Research confirmed that data from Likert items (and those from similar rating scales) become significantly less accurate when the number of scale points decreases to below five or increases to above seven. However, these studies provided no reasons for preferring five-point scales to seven-point scales. e) Convert linguistic scale to equivalent numerical scale: As mentioned, all preference values are identified in the subjective form, which cannot be used for further analysis unless they are converted into numerical values. Thus, in this step, the level of importance/significance of each criterion recorded by each expert on the linguistic Likert scale was converted to an equivalent numerical scale, as shown in Table 2 . A Likert scale assumes that the vaccine distribution criteria have different important levels that should be assigned by an expert. The importance level is assigned using a linguistic scale that facilitates the process of the evaluation criteria. The importance levels range from 'not important' to 'very important'. However, when an additional analysis needs to be conducted on the scores obtained by experts, extracting any useful information from linguistic scores is difficult unless converted into numerical values. Thus, an equivalent numerical value has been provided along with each linguistic term where measuring the importance level of the vaccine distribution criteria is possible. The previous step clarifies how the experts were selected and how their preferences were indicated. In this step, the EDM is constructed. The main parts of the EDM are the vaccine distribution criteria and the alternatives, as shown in Table 3 . According to Table 3 , a crossover is made between the vaccine distribution criteria and the SEJ panel. Each criterion (Cj) in the attribute intersects with each selective expert ( ), where the expert has scored a suitable level of J o u r n a l P r e -p r o o f importance for each criterion. The EDM is the base for further analysis steps in the proposed method, which are illustrated in the next sub-sections. The membership function and the subsequent defuzzification process of T-SFS are applied to the EDM data where the data are transformed to a T-SF EDM to increase their precision and ease of use in further analysis. However, in MCDM, the problem is uncertain and imprecise because assigning a precise preference rate to any criterion is difficult. The advantage of using the fuzzy method is the use of vague numbers instead of crisp numbers to determine the relative value of attributes (criteria) and address the issue of imprecise and uncertain problems [87] [88] [89] . The T-SFS is an objective having the form of [77, 90] and as defined in Eqs. (1) and (2). where : → [0,1] is the membership function, whereas : → [0,1] is the non-membership function of element ∈ , and : → [0,1] is the hesitate function to and must fulfil the restriction seen in Eq. (2). 0 < ( ( )) + ( ( )) + ( ( )) ≤ 1, (2) where T ≥ 1. The degree of hesitancy is presented in Eq. (3) [90] . The applied arithmetic operation using T-SFS utilised the following equations. T-SFS summation and aggregation operations can be seen in Eq. (4) [91] . The division operation was performed using Eqs. (3) and (5). However, Eq. (5) was adopted from [92] , which is used in the spherical fuzzy set. Thus, in this study, the square within this operation has been converted to the power t to fulfil the T-SFS structure. Eq. (6) shows the equation of T-SFS division on crisp value [83] . The value of each linguistic term with T-SFS is shown in Table 4 . where > 0. The defuzzied (crisp) value of a T-SFS fuzzy number is defined as follows in Eq (7) [77]: J o u r n a l P r e -p r o o f Table 4 indicates that all linguistic variables are converted into T-SFS, assuming that the fuzzy number is the variable for each criterion for Expert K. In other words, Expert K (a vaccine distribution expert) was asked to identify the importance level of the vaccine distribution criteria within the variables measured using a linguistic scale. Step 5: Computation of the final values of the weight coefficients of the evaluation criteria Based on the fuzzification data for the criteria in the previous step, the final values of the weight coefficients of the evaluation criteria ( 1, 2, . . . , ) are calculated in this step as follows. a) The ratio of fuzzification data is computed using Eqs. (3), (4) and (5); the preceding equations are used with T-SFS, where Eq. (8) symbolises the process as shown in Table 5 . ( 1/ 1) where ( 1/ 1) represents the fuzzy number of Imp (E1/C1). The mean values are computed to find the final fuzzy values of the weight coefficients of the evaluation criteria ( 1 , 2 , . . . ,̃) . The T-SF EDM is used to compute the final weight value of each criterion using Eq. (6), where Eq (9) symbolises the process. c) Defuzzification is performed to find the final weight; Eq. (7) is used as the defuzzification method. To calculate the final values of the weight coefficients, the weight for the importance of each criterion should be assigned given the sum of the weights of all the criteria for the rescaling purpose applied in this stage as well. [94] , which is used in the proposed COVID-19 vaccine distribution methodology (Figure 1 ). The following section provides information about the first stage of T-SFDOSM, which is the data transformation unit. After this, the second stage, data processing, is presented. According to [95] , the transformation of the DM into an opinion matrix is achieved using the following steps. Step 1: Select the ideal solution of each sub-criterion used in the DM of COVID-19 vaccine distribution. Therefore, the ideal solution is defined as shown in Eq. (10). where max is the ideal value for benefit criteria (i.e. C3 and C4), min is the ideal solution for cost criteria (no cost criteria are identified in the COVID-19 vaccine distribution) and is the ideal value for critical/categorical criteria (i.e. C1, C2 and C5) when the ideal value lies between the max and min. The critical value is determined by the decision-maker. Step 2: Reference comparison is made between the ideal solution and other values for each of the criteria used in the COVID-19 vaccine distribution criteria. A five-point Likert scale is used. The ideal solution selection step is followed by comparing the ideal solution with the value of vaccine recipients in the same criterion, as shown in Eq. (11). where ⊗ represents the reference comparison between the ideal solution and the value of alternatives in the same criterion. The final output of this block indicating the linguistic term is the opinion matrix that is ready to be transformed into a fuzzy opinion matrix by using T-SFS, as expressed in Eq. (12). Stage Two: Data-processing unit The opinion matrix of each Likert scale refers to the output of the transformation unit. The final block begins by transferring the opinion matrix into a fuzzy opinion DM by converting the linguistic terms of the opinion matrix into T-SFS using Table 6 . Eq. (13) multiplies the weights with each criterion value; this concept can calculate the effectiveness of weights in T-SFDOSM used in COVID-19 distribution thoroughly. Then, the defuzzification process of each alternative is computed using Eq. (7). After that, vaccine recipients can be prioritised. Each vaccine recipient will be assigned a value, and they will be ordered based on the best value. The vaccine recipient with the highest score will have the highest priority. Considering the variations in the distribution ranking of the COVID-19 vaccine amongst decision-makers, aggregated decisions obtained from various evaluators are necessary to unify the distribution ranking. Thus, this study utilised the GDM context with T-SFDOSM to unify all the variations in the distribution ranking of the decision-makers and arrive at the final distribution ranking. Furthermore, the arithmetic mean was used to arrive at the final score of GDM, as expressed in Eq. (14). The highest score value is the best vaccinator. In this case, the decision makers' opinions were combined after arriving at the final distribution ranking of vaccine recipients. = ⨁ * , (14) where ⨁ = AM; S* = The final rank for each expert. This section presents the evaluation and differentiation results of COVID-19 vaccine recipients to formulate the vaccine distribution mechanism. The section is divided into two subsections. Section 3.1 provides the results of the T-SFWZIC method and the constructed criteria weights; in particular, the three experts' judgment is converted using mathematical calculations to show the overall weights within this section. Section 3.1 displays the distribution results of the COVID-19 recipients based on the individual decision-making T-SFDOSM and GDM T-SFDOSM and are then presented. This section provides the weight determination results of the COVID-19 vaccine distribution criteria using the T-SFWZIC method developed in Section 2.2.1. After performing the involved steps, the T-SFWZIC method process resulted in GDM contexts weights (obtained from the three experts) without any inconsistency following the method philosophy. In addition, the obtained weights applied for T values (i.e. T = 2, T = 4, T = 6, T = 8 and T = 10) and Table 7 illustrates the final weight results of the five criteria for vaccine distribution. According to step 4, the process of the NS membership function is employed to transform crisp values into equivalent fuzzy numbers. After that, the process of transformation and the fuzzification of the experts' opinions on the significance of the five criteria are also presented. The ratio values of the criteria are computed according to Eqs. (3) and (6), followed by computing the mean of the experts' preference for each criterion to determine the fuzzy weight. Then, Eq. (7) is employed to determine the final weight for each of the five criteria as explained in step 5. Finally, the computed ratio and fuzzy value of the final weights of the five criteria are calculated. the second important criteria, whereas geographic locations severity (C4) received the lowest weight as the fifth important criteria. In addition, chronic disease conditions (C2) received the third important criteria for the T = 2 value and received the fourth important criteria for T values (4, 6, 8 and 10). Finally, disabilities (C5) received the fourth importance criteria for the T = 2 value, the third importance criteria for T values (4 and 6) and the second important criteria for T values (8 and 10). These final benchmarking results can be achieved by using the T-SFDOSM method as described in the next section; practically, these weight values need to be provided for the T-SFDOSM to compute the benchmarking results of the 300 vaccine recipients. The results and discussions presented in this section pertaining to the distribution of the COVID-19 vaccine are based on individual (Section 2.2.2.1) and GDM contexts (Section 2.2.2.2). The results of the opinion matrix and fuzzy opinion matrix used in the distribution of the COVID-19 vaccine are obtained. By using the five scales, the three decision-makers provided their opinions on converting the DM into the opinion matrix. According to Eq. (9), the decision-makers determined the ideal solution value based on the COVID-19 vaccine distribution criteria (i.e. vaccine recipient memberships, chronic disease conditions, age, geographic locations severity and disabilities). The opinion matrix was created by comparing the ideal solution with other values per criterion or each alternative using linguistic terms. The opinion matrix of each decision-maker is converted into a fuzzy opinion matrix. Thereafter, the fuzzy opinion matrix of the 300 vaccine recipients from the decision-maker is obtained (Table 1) . Moreover, the T-SFDOSM method was applied to the resulting T-SF opinion matrices to achieve the distribution of the COVID-19 vaccine. Table 8 presents the results of the COVID-19 vaccine distribution based on the individual T-SFDOSM decision-making context for the three decision-makers resulted from T = 2, T = 4, T = 6, T = 8 and T=10 with a sample of 10 vaccine recipients. The remaining is presented in Table A1 in the Appendix. As mentioned previously in Section 2.2.2, the highest alternative must have the highest score, and the lowest alternative must have the lowest score value. However, to provide additional analyses for the individual T-SFDOSM final rank results, Table 9 shows the best fourth alternatives (VR) obtained from the three experts for all T values. As shown in Table 9 , we aim to analyse the effect of variation in T value on the individual T-SFDOSM ranking results. For this purpose, we presented the best four alternatives (VR) for various values of T, and the ranking results are given for the three experts. Table 9 shows that varying T has a limited effect on ranking for the best four alternatives of each expert. For example, for the first expert with all T values, the best alternative is VR281 followed by VR221 as the second rank, VR93 as the third rank and VR274 as the fourth rank. In the same context, the results are also similar for the second and third experts. However, the little effectiveness for T values on the best four alternatives does not provide the precise conclusion on the overall 300 alternatives. Therefore, to discuss the real effectiveness of T values on T-SFDOSM individual ranking results, we calculate the overall variations that occurred in the ranking orders for the individual ranking for each expert that presented in Table A1 in the APPENDIX. The results showed that, for expert 1, 228 out of 300 alternatives (76%) were changed and received the different rank orders, whereas 72 alternatives (24%) received the same ranking order and not changed when T values are applied (T = 2, T = 4, T = 6, T = 8 and T = 10). Moreover, for expert 2, 229 out of 300 alternatives (76.3%) were changed and received different rank orders, whereas 71 alternatives (23.6%) received the same ranking order and have not been changed. Finally, for expert 3, 246 out of 300 alternatives (82%) were changed and also received different rank orders, whereas 54 alternatives (18%) received the same ranking order and have not to be changed. Although little variance has been observed for the best four ranking orders amongst alternatives (Table 9 ), these orders do not reflect the complete picture of how T values affected the ranking results. Therefore, as a conclusion for the above discussion, we found that a big variance has occurred on the ranking orders and score values based on T values, indicating the existence of T values' effectiveness on vaccine distribution. From another perspective, we found that the ranking results changed amongst the three experts. Therefore, this case shows the significance of variation in experts' preferences in decision analysis amongst experts. For example, Table 9 and Table A1 (APPENDIX) show that, for expert 1 in the case of T = 2, the VR281 is the best alternative rank and obtained the score of 0.758775, whereas for experts 2 and 3, the VR221 is the first alternative rank and obtained the score 0.731969. After reviewing the scores and ranking orders results for the individual T-SFDOSM, we found differences amongst the three experts obtained for the vaccine recipients. Overall, no unique prioritisation result was observed based on the opinions provided by the three experts. Given this variance, GDM, considering all the experts' opinions, is essential to provide final and unique prioritisation. Furthermore, GDM is necessary to solve the problem of variations in the final rank. Therefore, we present the results of the GDM context for all T values in Table A2 in the APPENDIX. As mentioned in Section 2.2.2.2, the final results of the three decision-makers were aggregated by using Eq. (14), and the final GDM raking for COVID-19 vaccine distribution was obtained. In addition, Table 10 shows the results of the COVID-19 vaccine distribution based on the GDM T-SFDOSM for the three decision-makers resulted from T = 2, T = 4, T = 6, T = 8 and T = 10 with a sample of 10 vaccine recipients. (Table 7 which indicates that age weight received higher priority for all T values based on the three experts) played a major role in the decision-making process and provided the alternative with a high priority. Hence, the remaining criteria varied somewhat in terms of importance. From another perspective, VR180 is almost located in the middle of ranking results, that is, rank 146 when T = 2 and obtained a score value of 0.44272; rank 151 for T = 4, T = 6 and T = 8 and obtained score values of 0.46522, 0.485626 and 0.494799, respectively; rank 150 for T = 10 and obtained a score value of 0.498175. The criteria specifications of VR180 related to C1, C2, C3, C4 and C5, as he is a recipient membership (employee postal), are not affected by chronic disease, 29 years old, from red geographical location and not affected with disabilities. A satisfactory ranking result had been assigned to alternative VR180, specifically the vaccine distribution criteria specifications are relatively averagely important and earned a middle priority. The lowest-ranked recipients were the alternatives VR22, VR166, VR205, VR229, VR269 and VR285, and they obtained the same ranking order (rank 293) and same scores for all T values. They received scores 0.174299, 0.277296, 0.351137, 0.400839 and 0.433967for T = 2, T = 4 T = 6, T = 8 and T = 10, respectively. The closeness of the criteria specifications for these alternatives is the reason for admitting them in the same order of priority and for obtaining the same score. For example, the criteria specifications of VR22 related to C1, C2, C3, C4 and C5, as he is not from vaccine recipient memberships, are not affected by chronic disease, 13 years old, from green geographical location and having disabilities, respectively. In conclusion for those in the worst ranked, all of their profile data do not have vaccine recipient memberships and are not affected by any chronic condition, younger age, from green or yellow geographic locations severity and slightly affected by disabilities. From another perspective, in line with the discussion analyses presented previously for individual T-SFDOSM of how the T values were affected the first four ranking results (Table 9) , Table 11 presents the best four alternatives based on GDM T-SFDOSM. Table 11 shows that for T = 4, T = 6, T = 8 and T = 10, the best alternative is VR221 followed by VR281 as the second-best rank, VR274 as the third-best rank and VR93 as the fourth-best rank. In the same context for ranking results when T = 2, the best three rank alternatives are similar to other T values, namely, VR221, VR281 and VR274. The only different result is that VR206 has the best fourth rank according to T = 2. To discuss the effect of T values on GDM T-SFDOSM, we also calculate the variations that occurred in the ranking orders for the GDM ranking results ( Table A2 in the APPENDIX) when T values are applied (T = 2, T = 4, T = 6, T = 8 and T = 10). In these contexts, 268 out of 300 alternatives (89.3%) were changed and received different rank orders when these T values are applied, whereas 32 alternatives (10.7%) received the same rank order and have not been changed. Therefore, as a conclusion of how T values affect GDM T-SFDOSM ranking orders, the big variance also has been occurred in line with the individual T-SFDOSM. Thus, T values play a key role in the overall ranking for the COVID-19 vaccine distribution for individual and GDM T-SFDOSM and should be considered. Finally, the rank of COVID-19 vaccine distribution is in line when comparing the GDM results with the opinion matrices. Thus, this rank is considered the final ranking results for COVID-19 vaccine distribution, which will be evaluated in detail in the next section. According to the literature review studies, the systematic ranking and sensitivity analysis assessments have been most widely used in the evaluation of the MCDM results. Thus, the efficiency of the proposed work for COVID-19 vaccine distribution is evaluated and tested through those assessments. Firstly, the systematic ranking of the vaccine recipients' ranking results is evaluated. Secondly, the effect of changing the criteria weight on the ranking result is examined and analysed over different scenarios. In the first evaluation process, to assess the prioritisation results for COVID-19 vaccine distribution and substantiate the obtained COVID-19 vaccine distribution GDM results, the prioritised vaccine recipients were divided into different groups following their prioritisation order. In this section, the systematic ranking evaluation process for the COVID-19 vaccine distribution results is discussed. To substantiate the COVID-19 vaccine distribution GDM results obtained, the validation process was performed by dividing the vaccine recipients into different groups. This process has been followed in various MCDM studies [96] [97] [98] . The number of groups or the number of vaccine recipients within each group does not affect the validation result [99] [100] [101] . To validate the group COVID-19 vaccine distribution results, several procedures are performed as follows: (1) All opinion matrices were aggregated to produce a unified opinion matrix. (2) The vaccine recipients within the unified opinion matrix were sorted/ordered according to GDM results. (2) The vaccine recipients were divided into six equal groups. (3) The mean ( ̅ ) for each group is calculated thereafter (Eq. (15)). The comparison process was based on the result of the mean in each group. The lowest mean values of each group lead to valid results because the decision-makers have assigned the lowest linguistic terms to the ideal solution of each criterion, which is the philosophy of FDOSM. Thus, the first group is assumed to have the lowest mean to check the result validity and is compared thereafter with the second group, and so on. The second group's mean result must be higher than that of the first group. The same applies to the third, fourth, fifth and sixth groups. If the evaluations are consistent with the assumption, then the results are valid. Table 12 presents the validation results for the group results obtained using the proposed T-SFDOSM. As shown in Table 12 , the initial observation of the ranking results of the six groups shows that all groups are systematically distributed across all the five scenarios (T1, T2, T3, T4 and T5) as the ranking results of the second group start from the end of the ranking results of the first group and so on for the other groups. In all the scenarios, the mean value of the first group was smaller than the mean results for the following group 2. Moreover, this consequent process was carried whilst considering that a group mean is smaller than the mean of the next corresponding group in each scenario. When the latter is achieved across all the groups, the systematic ranking is deemed valid. Judging by all the mean values in all the scenarios across all the groups, no group nor scenario was against the rule, and thus, all the scenarios across all the groups are valid. The statistical validation results indicate that the T-SFDOSM results of COVID-19 vaccine distribution extended by the groups are valid and have been systematically ranked. In this second evaluation process, the sensitivity of the proposed T-SFWZIC method against the changing criteria weight is analysed. Thus, the sensitivity analysis predicts the effect of changing criteria weights on the systematic ranking results of the vaccine distribution results. To analyse the sensitivity, firstly, the most important criterion should be identified for each T value. In this study, out of the five criteria, C3 = age was the most important for all T values as presented in Table 7 . To examine the effect of changing criteria weights, nine different scenarios for each T value generated from the relativity of criteria weight were computed using Eq. (16) [102] . The relative change for each criterion over the most important one (age) with respect to each T value was computed using the elasticity coefficient ( ) as shown in Table 13 . where for T value:  is the higher significant contribution,  represents the original weight values computed using T-SFWZIC,  0 is the total of original weights for the changing criteria weight values, and  Δ is the range of change applied on the five criteria weight values, which represents the limit values of the most significant criterion in this study (age), as follows: As shown in Table 13 , the T Value for each criterion has changed the weight values according to Eq. 16. For all (α_c) with respect to T values (T = 2, T = 4, T = 6, T = 8 and T = 10), the age (C3) received the highest weight as the first important criteria, whereas geographic locations severity (C4) received the lowest weight as the fifth important criteria. Then, the interval range of Δ for T values is used to generate nine new weighting values for each criterion by dividing it into nine equal relative values based on the number of scenarios, as shown in Table 14 . Based on Table 14 , these ninth new weight values for each T value are employed to assess the sensitivity of the 300 vaccine recipients' prioritisation towards changing criteria weights. The aim is to determine how target T-SFWZIC weights are affected based on changes for the nine scenarios for each T value. Figure 2 illustrated the influences of changing the criteria weight over the first 10 ranks only for T = 2. Figures A1, A2, A3 and A4 in the Appendix illustrate the influences of changing the criteria weight over the first 10 ranks of T = 4, T = 6, T = 8 and T = 10, respectively. Incontrovertibly, the criteria weights play a vital role in changing the priority of each vaccine recipient; these nine-scenario results for all T values support the research assertion about the significant contribution of these five criteria. Notably, although this change is logical and likely, maintaining the results in most of the nine scenarios proved the efficiency of the proposed integration methods in handling such sensitive cases with a large-scale dataset and producing supportive results for the outcomes of systematic ranking. (Table 7 shows the weights). The sensitive analysis comparisons can be discussed from two points of view as follows: (A) Effectiveness of the first three ranks: the comparison with respect to the first three ranking alternatives needs to be discussed because those vaccine recipients received important orders. For T = 2, scenarios S3, S4, S5, S6 and S7 have the same ranking results as T-SFWZIC which are obtained by the first three alternatives (V221, V281 and V274), whereas other scenarios (S1, S2, S8 and S9) were relatively different. For T = 4, T = 6 and T = 8, scenarios S3 to S9 have the same ranking results as T-SFWZIC, which are obtained by the first three alternatives (V221, V281 and V274), whereas only scenarios S1 and S2 were relatively different. For T = 10, only scenarios S3 to S7 have the same ranking results as T-SFWZIC, which is obtained by the first three alternatives (V221, V281 and V274), whereas only scenarios S1 and S2 were relatively different. When comparing the above new results with the first three ranks that were obtained from T-SFWZIC weights, the results revealed that no big differences exist that have been changing the first three ranking results for the sensitively of T values. However, the first three ranks cannot provide the full sensitive analyses for the overall changes that occurred in the ranking results. Therefore, the overall effect should be discussed. (B) Effectiveness of overall ranks: after reviewing the overall ranking results, we found that the changing behaviour of the nine scenarios with respect to each T value has widely occurred. Moreover, how exactly the overall new ranking results are affecting the previous ranking results obtained from T-SFWZIC weights should be measured. We can measure this effectiveness by calculating the changing occurred in the orders between both ranks, and then, we calculate the changing percentage between both ranking orders. In other words, for example, for T = 2, the number of changes that occurred in the ranking orders obtained from T-SFWZIC weights after applying S1 weights is 296 (98.67%), whereas only four orders are not changed and have the same orders. Table 15 explains the overall effectiveness analyses that occurred on the ranking results between the ninth scenario and T-SFWZIC weights. Table 15 presents the final sensitive analyses for all scenarios with respect to all T values. The highest mean value is obtained by T = 2 and T = 10 (89.74%). The lowest mean value is obtained by T = 8 (87.19%). These interesting results indicate that the rank stability is almost highly sensitive with all T values, and then, ranking obtained by T-SFWZIC weight is affected by the nine scenarios. Surely, these widely changing results in the weights numbers are defiantly changing the overall ranking results. This concept is already reported and considered one of the four MCDM issues which is 'important criteria'. If we reviewed these issue concepts, then we can realise that the 'important criteria' has been sensitively recognised and proven here for the presented study which is vaccine distribution. At this step, sensitivity analysis is conducted to investigate the priority ranking stability; however, the sensitivity of the priority ranks of T values for the nine scenarios is influenced by the criteria weights changing. Furthermore, the overall rank for all vaccine recipients changed except for some priority ranks (the first three ranks). This fact is probably caused by some important issues of criteria importance and has been demonstrated for T-SFWZIC weights. Finally, the Spearman correlation coefficient (SCC) was employed to evaluate the relationship between the results of the 15 scenarios statistically [102] . Figure 3 shows the high-level correlation amongst the nine scenarios for all 300 vaccine recipients for T = 2. Figures A5, A6 , A7 and A8 shows the remaining correlations for other T values. In conclusion, the high SCC mean value corresponds to T = 10 (0.940); however, all the T values are relatively similar to one another based on correlation analyses. Thus, this high correlation value indicates a significant correlation of the rank outcomes, which in turn supports the systematic ranking results amongst T values. This study contributes to the body of knowledge of the MCDM methods by proposing new formulations of FWZIC and FDOSM based on the T-SFSs environment. The reason for such formulations was to perform both methods with no restrictions on their constants and obtain more degree of freedom in handling the uncertainty in the data. To achieve the study objective, the proposed methodology was presented in two phases, namely, decision matrix adoption and development ( Figure 1 ). The result was an inductive methodology based on the detailed weighting and prioritisation steps presented within each MCDM method. The evaluation process was performed based on systematic ranking and sensitivity analyses, which proved the robustness of the proposed work. Notably, the sensitivity analysis results show that the weight importance posts a considerable issue for the distribution of the COVID-19 vaccine. Thus, assigning the importance weights for the distribution criteria used in the prioritisation of vaccine recipients is very important. However, this study has two main limitations. Firstly, T-SFWZIC and T-SFDOSM methods were formulated considering only one T-SFSs aggregation operator in addition to using one defuzzification technique only to produce the final weighting and ranking results. Secondly, the importance measurement reflected on each DM's preferences was not considered in the proposed methods. Several future directions can be achieved as follows: (1) Presenting and processing a large-scale dataset of COVID-19 vaccine recipients considering all probabilities frequently augmented for each alternative and distribution criteria. (2) Performing the proposed MCDM methods based on two levels: firstly, each vaccine recipient membership will be prioritised, and secondly, each alternative within each membership will be prioritised followed by accumulating them effectively. (3) Several fuzzy types, such as interval type-2 hesitant [103] , intuitionistic and interval-valued [104] and neutrosophic [102] , can be adopted in the FDOSM and/or FWZIC to effectively overcome the uncertainty limitation. [7] A. J o u r n a l P r e -p r o o f 9. Figure A2 Sensitivity analysis of first 10 vaccine receipts ranks in 9 scenarios (T=6). 10. Figure A3 Sensitivity analysis of first 10 vaccine receipts ranks in 9 scenarios (T=8). J o u r n a l P r e -p r o o f 11. Figure A4 Sensitivity analysis of first 10 vaccine receipts ranks in 9 scenarios (T=10). 12. Figure A5 Correlation of ranks among 9 scenarios for all 300 vaccine recipients for T=4. Figure A6 Correlation of ranks among 9 scenarios for all 300 vaccine recipients for T=6. 14. Figure A7 Correlation of ranks among 9 scenarios for all 300 vaccine recipients for T=8. Role of biological Data Mining and Machine Learning Techniques in Detecting and Diagnosing the Novel Coronavirus (COVID-19): A Systematic Review Systematic review of artificial intelligence techniques in the detection and classification of COVID-19 medical images in terms of evaluation and benchmarking: Taxonomy analysis, challenges, future solutions and methodological aspects Multi-Biological Laboratory Examination Framework for the Prioritization of Patients with COVID-19 Based on Integrated AHP and Group VIKOR Methods Helping doctors hasten COVID-19 treatment: Towards a rescue framework for the transfusion of best convalescent plasma to the most critical patients based on biological requirements via ml and novel MCDM methods Sentiment analysis and its applications in fighting COVID-19 and infectious diseases: A systematic review Detection-based Prioritisation: Framework of Multi-laboratory Characteristics for Asymptomatic COVID-19 Carriers Based on Integrated Entropy-TOPSIS Methods Generalized MULTIMOORA method and Dombi prioritized weighted aggregation operators based on T-spherical fuzzy sets and their applications Interval Valued T-Spherical Fuzzy Soft Average Aggregation Operators and Their Applications in Multiple-Criteria Decision Making Emergency decision support modeling for COVID-19 based on spherical fuzzy information A spherical fuzzy TODIM approach for green occupational health and safety equipment supplier selection A New Extension of Pythagorean Fuzzy Opinion Score Method based on Power Bonferroni Mean Operator for Evaluating and Benchmarking the Sign Language Recognition Systems Fuzzy decision by opinion score method Based on real time remote health monitoring systems: a new approach for prioritization "large scales data" patients with chronic heart diseases using body sensors and communication technology Multidimensional benchmarking of the active queue management methods of network congestion control based on extension of fuzzy decision by opinion score method A new standardisation and selection framework for real-time image dehazing algorithms from multi-foggy scenes based on fuzzy Delphi and hybrid multi-criteria decision analysis methods Novel technique for reorganisation of opinion order to interval levels for solving several instances representing prioritisation in patients with multiple chronic diseases A Uniform Intelligent Prioritisation for Solving Diverse and Big Data Generated From Multiple Chronic Diseases Patients Based on Hybrid Decision-Making and Voting Method A Novel Multi-Perspective Benchmarking Framework for Selecting Image Dehazing Intelligent Algorithms Based on BWM and Group VIKOR Techniques A novel fuzzy hybrid neutrosophic decision-making approach for the resilient supplier selection problem Correlation coefficients of dual type-2 hesitant fuzzy sets and their applications in clustering analysis Intuitionistic fuzzy interval-valued linguistic entropic combined weighted averaging operator for linguistic group decision making The authors are grateful to the Universiti Pendidikan Sultan Idris, Malaysia for funding this study under UPSI SIG Grant No. 2020-0150-109-01.