Pandemic response management framework based on efficiency of COVID-19 control and treatment

Aims: The existing response management system for pandemic disease fell short of controlling COVID-19. This study evaluates the response management relative efficiency of 58 countries in two stages, using two models. Materials & methods: Data envelopment analysis was applied for efficiency analysis. Results: 89.6% of countries were inefficient in pandemic control and 79% were inefficient in treatment measures. Sensitivity analysis underlines resources as a critical factor. Further examination points to absence of a robust and uniform mitigation measure against the pandemic in most countries. Conclusions: Preventing spread is not only the first line of defense; it is the only line of defense. The lack of a global public health database support system and uniform response compounded inefficiency. A robust pandemic response management framework is developed based on practices of key performers. Action plans are proposed, with a recommendation for a global public health pandemic database monitoring and support system as the nucleus.


Data description
The process of analyzing COVID-19 pandemic response management is categorized into two stages. Stage 1 considers COVID-19 contagion control efficiency, analyzes countries' performance in terms of minimizing the spread of the virus and identifies countries that adopted efficient pandemic control measures. Two inputs and one output are considered in this stage. Factors that have been described as critical to the spread are utilized. Population density and COVID-19 confirmed cases are considered as outputs. Population density is the measurement of population per unit area; it is one of the factors known to influence transmissibility of COVID-19, with a moderate risk of infection for people working in areas with high population density [12,13]. The average of 13 International Health Regulations (IHR) core capacity scores is an indicator representing the core capacities that have been achieved by a country at a given point. These 13 indicators have been identified to have connection to  and are integral to the preparedness and vulnerability of countries in relation to COVID-19 [14]. They are: legislation and financing, IHR coordination and national focal point functions, zoonotic events and the human-animal health interface, food safety, laboratory, surveillance, human resources, national health emergency framework, health service provision, risk communication, point of entry, chemical events and radiation emergencies. Stage 1 of the analysis is further classified into two phases: stage 1A considers the first 3 months after announcement of the pandemic, and stage 1B examines the subsequent 3 months to see which countries improve or maintain efficient contagion control and the ways in which they achieve these improvements. It is important to note that the number of tests conducted was considered for inclusion; however, a lack of data and the known inconsistencies and unreliability of existing data for most countries led to its exclusion. Nonetheless, the test trends were examined at later stages of the analysis.
Stage 2 of the analysis evaluates the treatment efficiency and management of the pandemic. The number of confirmed COVID-19 cases is an important input in this stage, because it represents the pressure exerted on the healthcare system and constitutes the primary input of the pandemic treatment. Relevant resources, such as number of ventilators, number of testing and amount of PPE were considered for inclusion; however, due to unavailability of data, they were discarded as variables. Other relevant resources with available data (number of physicians per 1000 population and number of hospital beds per 1000 population) were utilized. The percentage of the population with age >65 is an important factor because fatality of the virus is more prevalent in the elderly population and individuals with preexisting health conditions [15,16]. The number of physicians per 1000 population and the number of hospitals per 1000 population are important parameters used to evaluate efficiency of healthcare systems and adequacy of the system capacity [17,18]. This stage uses data from 6 months after the pandemic announcement. Variables for efficiency evaluation are as follows. Stage 1 (COVID-19 contagion control efficiency): inputs are population density [19] and average of 13 IHR core capacity scores [20]; outputs are COVID-19 confirmed cases [20]. Stage 2 (COVID-19 treatment efficiency): inputs are COVID-19 confirmed cases [21], number of physicians per 1000 population [27], number of hospital beds per 1000 population [19] and percentage of population with age >65 years [19]; outputs are COVID-19 related deaths [21] and COVID-19 recovered cases [21].

Data envelopment analysis
The variables used to model efficiency of COVID-19 control and treatment, present a complex system. Therefore, a robust technique that can handle multiple inputs and outputs in addition to negative outputs (e.g., positive COVID-19 cases and mortality) is required. Data envelopment analysis (DEA) is a performance evaluation technique capable of handling multiple inputs and outputs [22], with abundant empirical applications in healthcare systems and strategies [17,[23][24][25]. DEA has been applied to analyze effects and efficiency of pandemics such as HIV/AIDS [26]. The efficiency of schistosomiasis control programs in Jiangsu Province, China was also analyzed using DEA [27].
DEA is a nonparametric method of efficiency evaluation introduced by Charnes et al. [28] under constant return to scale (CRS) to evaluate efficiency of systems known as decision-making units (DMUs). It was later modified by Banker et al. [29] with variable return to scale (VRS). Subsequently, various models have been developed, including direction distance function [30] and target setting model [22]. DMUs are generic, taking the form of countries, systems or companies that need evaluation with a set of homogeneous parameters. It constructs a bestpractice frontier from the sample observations and measures the radial distance of other observations relative to the frontier [31]. This study utilizes DEA to evaluate the performance of countries in terms of their COVID-19 pandemic management. The DEA efficiency scores show the performance level of each country relative to other countries for the evaluated period. DEA compares the homogeneous units among themselves and accepts the best observation as the efficient frontier, then other observations are benchmarked against that frontier [17].
Efficient pandemic contagion control requires utilization of resources to minimize the spread of the pandemic, in addition to new protocol implementation. Furthermore, efficient pandemic treatment practice with the number of infections and available resources necessitates minimizing the fatality rate and maximizing the number of patients treated. In this context, the DEA model adequately handles such parameters (desirable and undesirable outputs) and objectively evaluates efficiency by accounting for the asymmetry between both types of outputs [32] and alleviating the possibility of biased results due to converting undesirable outputs to their inverse (ratio) [33].
When considering a multiple input and output system [28] the production possibility set (PPS) is defined as: where X j = (x 1 j , · · · , x mj ) and Y j = (y 1 j , · · · , y sj ) represents the observed m-inputs and s-outputs of j = 1, . . . , n DMUs. Chambers introduced a directional distance efficiency measure by projecting units (x 0 , y 0 ) to a preassigned coordinate g = (−g − x , g + y ) = 0 m+s , g − x ∈ R m and g + y ∈ R s in a direction β [34]. Equation 1 illustrates the linear program associated to the estimation.

Max
The optimal solution of Equation 1 corresponds to the CRS efficiency β * CRS . If β * CRS = 0, the unit under evaluation is technically efficient, whereas β * CRS > 0 signifies an inefficient unit. Correspondingly, the VRS model is achieved by adding n j =1 λ j = 1 , as shown in Equation 2. The optimal solution of Equation 2 is VRS efficient if β * VRS = 0 and inefficient if β * VRS > 0. Consequently, the scale efficiency from the directional model is achieved as follows: Along with positive output of a system, undesirable outputs are sometimes observed, such as hazardous waste in an environmental context or mortality/fatality in healthcare. Most efficiency evaluation models do not account for the asymmetry between both types of outputs, which leads to erroneous efficiency estimation. Incorporation of the characteristics of undesirable outputs into DEA efficiency estimation relies on a directional measure that handles desirable and undesirable outputs differently [32].
The PPS is redefined as follows: the initial output vector of i = 1, 2, . . . , s .y ∈ R s ++ is divided into desirable and undesirable y = (y d , y u ), with y d ∈ R q ++ respectively. This is expressed into the following reference PPS CRS = (x , y d , y u ) | x ≥ x λ, y d ≤ y λ, y u = y λ, y ≥ 0 , designating undesirable outputs as weakly disposable [35]. To prevent the inconsistencies in the method of [30], the method of [36] is used to define directional efficiency, resulting in an increase in desirable outputs and a decrease in undesirable outputs from the same inputs. Therefore the directional efficiency measure corresponds to the solution of Equation 3.
The optimal solution of Equation 3 is β * CRS , if β * CRS = 0, with λ = 1, λ j = 0( j = 0), the unit under evaluation is directionally efficient. Otherwise, β * CRS > 0 signifies an inefficient unit. Given the different frontier estimating methods, this application can be difficult to understand for non-experts on frontier based models. Figure 1 presents a flow diagram illustrating the development and implementation of the model.

Results
The COVID-19 data of 15 July 2020 were the latest data extracted. Across the period, confirmed cases increased in all measures. The USA recorded the highest number of confirmed cases in both the first 3 months and the subsequent 3 months. There was an increase in the average confirmed cases among the countries considered. China recorded the minimum confirmed cases, with a 97% decrease compared with the preceding 3 months (see Supplementary Table 1: descriptive statistics).
To ensure a balanced dataset, 58 countries were considered in stage 1 and 57 countries in stage 2. Figure 2 presents the efficiency scores for contagion control of stage 1A and stage 1B. 89.6% of the countries evaluated were inefficient, with an average efficiency of 45.6% in stage 1A. The average contagion control efficiency improved to 64.3% with about 87.9% of the countries still inefficient in stage 1. China and South Korea showed a remarkable improvement, with 99.7 and 95.2% efficiency improvement respectively, in the second phase of contagion control ( Figure 2). Other significant improvements included Denmark (64%), Switzerland (63.1%), Austria (59%), Japan (57.8%), Bahrain (52.3%), Portugal (50.7%) and Morocco (54.9%). Australia, Argentina, Afghanistan, Kazakhstan and Peru were consistently efficient. Countries such as Oman, Guatemala, Mexico, Columbia and Bangladesh performed worse in stage 1B, with negative efficiency improvements of 15.3, 9.2, 7.7, 3.3 and 2%, respectively. Pakistan, the USA, Brazil and Chile showed no improvement in the second phase despite their significant inefficiency in the first phase. Supplementary Table 2 illustrates the numerical contagion efficiency scores. Figure 3 presents a summary analysis of contagion control efficiency. Changes in efficiency of the most and least efficient countries are illustrated in Supplementary Figure 1.
The second stage of the analysis looks at countries' efforts toward treating the virus during the evaluated period. Consideration of the efficacy and efficiency of the drugs used for treatment is beyond the scope of this study, which focuses on identifying the countries that have done a relatively good job of minimizing COVID-19-related deaths and maximizing recovered cases. Figure 4 presents the results of COVID-19 treatment (model 1) and sensitivity analysis using only COVID-19 confirmed cases as input (model 2). Model 1 indicates that 79% of the countries considered were inefficient in treating the virus, with an average efficiency score of 62.1%. A robustness check of the result, performed using sensitivity analysis by considering only confirmed cases as inputs, shows 96.5% of the countries were inefficient in treating the virus, with an average efficiency score of 51%. Supplementary Table 3 illustrates the numerical treatment efficiency scores.

Discussion
The efficiency analysis of control and treatment of COVID-19 across 58 countries for the first 6 months of the pandemic provides insight on response management performance of different countries. Countries like Austria, Bahrain,  China, Denmark, Germany, Ghana, Ireland, Italy, Morocco, Qatar, Singapore, Switzerland, Turkey and the United Arab Emirates showed a consistent efficiency in both treatment efficiency analysis models. The United Kingdom, Netherlands, Belgium, France, Guatemala and Honduras (among others) were consistently inefficient in both treatment efficiency analysis models. The USA, Brazil, Russia, Pakistan, Bangladesh, India and South Africa showed above 35% decrease in efficiency in the second model.
Countries that have zero or negative changes in efficiency of contagion control between stage 1A and stage 1B exhibit an inefficient COVID-19 treatment. Therefore, for most countries, it is important to note that preventing the spread of the virus is not only the first line of defense; it is the only line of defense. In addition, sensitivity analysis highlights the significance of resources such as number of physicians and hospitals as critical factors toward defeating the pandemic. This is supported by the significant drop in efficiency in countries such as Bangladesh, Brazil, India, Nigeria, Pakistan and South Africa.

Pandemic response management framework & action plan
The gross inefficiency of COVID-19 contagion control across the 58 countries evaluated in this study is indicative of the absence of a robust pandemic response management framework capable of controlling a pandemic of such magnitude. Practices of the best and worst performing countries are examined to propose a robust pandemic response management framework. Countries with high stage 1A scores and a significant positive difference in efficiency scores are analyzed (Table 1). Countries with negative and zero difference in stage 1 efficiency scores are also examined ( Table 2). Actions of these key countries were used to develop the pandemic response management framework illustrated in Figure 5.
Clear, uniform and regular public communication has proved effective in informing the population on the severity and importance of adhering to new protocols. Furthermore, upscaling vigilance coupled with the proposed pandemic response management framework could be more effective.
The mandatory lockdowns that have been imposed are not a sustainable approach, due to their economic and health effects.
Step 10 of the framework suggests gradual lifting of restrictions with precautions. The following can be incorporated as restrictions on traveling and other aspects of human life are lifted: • The use of infrared thermal imaging scanning; • The use of QR codes for all international travelers entering a country; the traveler will be asked to scan a QR code that takes them to an online declaration form containing contact information and determining whether they have COVID-19 symptoms. In addition, it can be used in hospitals to track confirmed cases [62];

China
Efficiency stage 1B    • As knowledge and research increases, technology such as artificial intelligence can aid in faster decision-making and tracking of COVID-19 cases. It can be used in various applications, including: • Developing advanced diagnostic tests and vaccines; • Predicting vulnerable regions, people and countries in which measurements should be taken rapidly; • Providing data on the number of resources needed in certain hospitals, such as number of beds and ventilators [63].
The lack of a global public health database support system compounded the complication and inefficiency of developing a robust and uniform response to COVID-19. Global collaboration and high-quality data sharing are needed to fight COVID-19 [64] and any similar pandemic. It is recommended that a global public health pandemic database monitoring and support system be established and supported by all countries, because a pandemic knows no border. Figure 6 summarizes an action plan for decision-makers based on the framework and considering the level of criticality of a pandemic. The action plan includes:  • Early control: the first initiative is to identify that a pandemic has started. The main goal at this stage is to minimize the spread; • Implementation of travel restrictions: one of the early measures for controlling the pandemic is to restrict travel.
This step is necessary to isolate the uninfected regions, as well as limiting the probability of an asymptomatic person traveling. In addition, other countries will benefit from travel restrictions that slow the global spread of the pandemic, especially at a stage where it is not contained at its sources [65]; • Implementation of social distancing/mandatory lockdown: social distancing or mandatory lockdown aims at reducing community spread of the pandemic. In terms of effectiveness, mandatory lockdown is a strict measure that restricts people from leaving their homes, apart from through necessity and at certain determined times. In addition, mandatory lockdown enables drastic reductions in social contact [65]; • Randomized testing: randomized testing at the population-wide level will help understanding of the country's epidemiological status and of transmission within the population setting, as well as estimation of secondary attack rates. Randomized testing within random households will help to characterize secondary cases, analyze the range of clinical presentations and the expected likelihood of infection, and expose asymptomatic infections [66]; • Expansion of the Epidemic Intelligence Service (EIS) workforce: one of the ways a country can measure and control the spread of the pandemic is through the use of EIS technology. The main goal of EIS is to rapidly provide guidance when selecting and implementing interventions to prevent the spread of the pandemic when it arises; 10.2217/fvl-2020-0368 Future Virol. (Epub ahead of print) future science group • Expansion of testing capacity: another step is to ensure that testing capacities can be expanded in infected regions. This can provide necessary information to further support decisions on the appropriate timing, response and type of precautionary measures to be implemented [67]; • Mandatory closure of nonessential businesses: this step includes the closure of nonessential businesses to the public as well as nonessential on-site business operations; • Mandatory quarantine of uninfected people: it is essential to encourage the public to limit unnecessary contact because the safety measures will not help in identifying asymptomatic individuals. It inhibits asymptomatic individuals from further infecting others, which subsequently impacts the testing policies and strains the healthcare system due to limited capacity [68]; • Review of the country's status: before lifting restrictions, the country's situation and performance should be evaluated in terms of resources (medical supply, healthcare staff and number of tests). The number of infected cases and population should be taken into account. Safety measures should be established, and strict mandatory regulation should be applied to maintain the results gained from the previous stages. Lessons learned from the rapid action will be considered in taking subsequent steps. The spread of the virus, the preparedness of public health and curative services to contain all new cases, the ability to minimize the risk of resurgence, and population awareness are other factors to consider [69]. • Gradual lifting of restrictions with precautions: the three Rs -readiness, responses and resilience/recoveryrepresent the systematic approach for lifting lockdowns taken in times of crisis. Readiness consists of coordination Step 1: Step 2: Step 3: Step 4: Step 5: Step 6: Step 7: Step 8: Step 9: Step 10: Figure 6. Action plan for decision-makers.
of emergency task forces, training and skills building capacity, and increasing preparedness for health resources and services. Responses include legislation and laws for managing the reopening at the provincial level; public engagement and involvement of stakeholders, public awareness and education through effective communication, and activating the role of the judicial police are essential factors for responsiveness. The last step, resilience and recovery, involves taking advantage of the existing database by documenting the lessons learned. It includes health resilience and surveillance assessments and public policy and priority-setting based on setting criteria for lifting the lockdown, beginning with vital public sectors such as health and food security and followed by other sectors in a gradual approach that provides enough time to control the virus after reopening and detecting any new or suspected cases and their contacts [69].

Conclusion
COVID-19 has made a significant impact on human life. The particular response strategy implemented has an enormous impact on the outcome for the country. In this study, DEA models were used to estimate the efficiency of contagion control for 58 countries and treatment efficiency of 57 countries affected by the COVID-19 pandemic.
The results show significant inefficiency in contagion control, hence the large number of confirmed cases and consequent rise in related deaths. 89.6% of countries were inefficient in the first phase; this figure increases in the second phase to reach 96.5%. Sensitivity analyses underline the importance of resources in fighting the pandemic, thus resource augmentation for strategic purposes is recommended.
Further examination of efficient countries shows that mask wearing, social distancing, quick isolation and testing are key practices for an efficient response. Furthermore, the results of the study are consistent with observational studies such as that of Khorram-Manesh et al. [70] that emphasize continuous assessment, communication and complete physical distancing among the initial key strategies. The proposed pandemic response management framework minimizes the potential for overwhelming spread of the virus and the chances of viral resurgence. The recommended action plan helps decision-makers to implement the framework at different levels of criticality. It is evident that collective and spontaneous measures across countries will also minimize the impact of the pandemic. Therefore establishment of a global public health pandemic monitoring and support system will help to organize a global effort toward defeating possible future pandemics.
The study has some limitations. The authors acknowledge the absence of data on the number of COVID-19 tests during the evaluated period; the absence/inconsistency in data on COVID-19 testing and the possibility of repetition within the dataset hindered the use of this indicator as an input variable. However, this limitation does not affect the credibility of the analysis, because further examination identified countries with reliable data on testing to have adequate testing capacity. However, a micro-analysis at national level should consider testing as an input after rigorous statistical checks.

Future perspective
Integration of innovative technology in the early stages of the pandemic was limited. Future studies should analyze strategic utilization of innovative technologies such as artificial intelligence (AI)/machine learning in the response system. In addition, future studies can support the proposed framework by integrating AI/machine learning at stages that require tracking, predicting, and proper screening process. It could also account for the statistical limitation of repetitive data in indicators such as number of testing or unreported cases.

Executive summary
• The study involves a comprehensive relative efficiency analysis of COVID-19 response management systems based on contagion control and treatment in 58 countries. • It includes a comprehensive review of the COVID-19 response management strategies of efficient and inefficient countries. • A robust pandemic response management framework is developed to address the shortfall of existing pandemic response management systems. • Action plans are proposed with a recommendation for a global public health pandemic database monitoring and support system as the nucleus.

Supplementary data
To view the supplementary data that accompany this paper please visit the journal website at: www.futuremedicine.com/doi/suppl/10.2217/fvl-2020-0368 future science group 10.2217/fvl-2020-0368