|
| |
| REVIEW ARTICLE |
|
| Ahead of print publication |
|
|
COVID-19: An evidence-based and disaster response methodology
Syed Muhammad Faisal Iradat, Muhammad Zain Uddin, Syed Irfan Nabi, Zaheeruddin Asif
Department of Computer Science, School of Mathematics and Computer Science, Institute of Business Administration, Karachi, Pakistan
| Date of Submission | 22-Sep-2022 |
| Date of Decision | 30-Nov-2022 |
| Date of Acceptance | 12-Dec-2022 |
| Date of Web Publication | 10-Apr-2023 |
Correspondence Address:
Syed Muhammad Faisal Iradat,
IBA Main Campus, University Road, University of Karachi, Karachi
Pakistan
 Source of Support: None, Conflict of Interest: None DOI: 10.4103/njbcs.njbcs_49_22
The world is gradually getting out of the grip of COVID-19 pandemic, although there are still high number of cases in some countries. Most of the initial attempts to predict and project the course of pandemic were hypothetical or based on historical data, as no current and specific data were available in the early days of pandemic. Most governments followed the policy of “flattening the curve” in order to avoid overwhelming their health systems. Most of the world also followed the policy of forced lockdowns to stop the spread of the virus. However, these policies produced did not produce consistent results across the globe. To investigate the impact of various policy measures on the reported outcomes, this research analyzed the actual COVID-19 data up till May 29, 2021, and the associated outcomes. Using global COVID-19 death rate as a base, the death rates of various countries were analyzed to gauge the efficacy of lockdown measures through probabilistic estimates and relative lack of uncertainty. Brier score was calculated to find the accuracy of probabilistic estimates. The data show high divergence in infection, death, and growth rates of the virus in different countries. The research also includes comparing the effects of virus in year 2020 and 2021, and the effect of vaccination. It can be seen that the collective world response was not commensurate with the actual risks involved. The paper concludes by emphasizing the need for specific evidence-based governance and disaster response management to face similar challenges in the future.
Keywords: Accuracy of probabilistic estimates, Brier score, correlation, COVID-19, lockdown, sensible policymaking, smart lockdown, vaccination
How to cite this URL:
Faisal Iradat SM, Uddin MZ, Nabi SI, Asif Z. COVID-19: An evidence-based and disaster response methodology. Niger J Basic Clin Sci [Epub ahead of print] [cited 2023 Jun 10]. Available from: https://www.njbcs.net/preprintarticle.asp?id=373998 |
| Introduction | |  |
Pakistan reported its first novel corona virus disease 2019 (COVID-19) infection case on February 26, 2020. The USA and Italy reported their first COVID-19 death in the same month, with a population of 212 million people[1] and gross domestic product (GDP) 1 / 79th of that of the USA,[2] the government of Pakistan seemed to fear that it will be unable to cope with the pandemic, resulting in large number of deaths. Following the footsteps of many other countries affected by COVID-19, Pakistan decided to impose a complete lockdown on March 26, 2020. The government adopted harsh precautionary social distancing procedures based on incomplete data and hypothetical projections based on past pandemics. These policies were not based on the established principles of evidence-based medicine.[3] The initial predicted infection and death rates were said to grow exponentially,[4] which later proved to be untrue. At the time, there was no treatment available for the novel virus and the earliest availability of vaccine was calculated to be at least 12-18 months away. The focus of the governments around the world, therefore, shifted to containing the spread and keeping the infection rate “below the curve”. Not many people challenged this approach to lockdown around the world as the true picture was not known. One of the challenges at the time was the lack of actual data. However, Giesecke estimated that over a year the total number of deaths with or without lockdown would have been similar.[5]
Due to the corona virus the real GDP of Pakistan was projected to shrink by 1≺3% in fiscal year 2020.[6] The government, therefore, started to provide extensive economic relief including stimulus packages, cash stipends, tax cuts and refunds, and deferred interest payments. Some experts warned that the economic fallout of these measures could be worse in the years to come. All these facts indicate that the governments around the world were following each other, creating a policy feedback loop. In this paper, using the evidence based medicine principles, we present some empirical data, to analyze the dynamics of lockdowns, social distancing, virus spread, and death and vaccination rates. We find that approaches like smart lockdowns as compared to a nationwide lockdown produced better outcomes in terms of arresting death and virus spread rates and maintaining low impact on economy. We speculate that it may be because the former restricted economic activities only in specific areas where the cases were high and did not cause economic loss for whole of the country. Further, by scoring the probability of no change in daily deaths rate, we attempt to compute the relative lack of uncertainty component of the Brier score[7] to construct a simple effective forecasting strategy. Finally, we conclude our discussion of the facts presented and recommend rethinking of the national strategy to balance strict measures to contain COVID-19 with the economic fallout of the same.
| Methods | | |
Premier on facts and data
There are two issues to be noted here. The first one is the actual data on infections and deaths, which we have used to select the countries for analysis. The second one is the resulting lockdown and other measures to curtail the pandemic, which we have used to formulate our research questions.
The forecast research method
We assume that actual COVID-19 data provide a more realistic estimate than the hypothetical data sets. To devise a forecasting mechanism for predicting the probability of change in lockdown strategy, we adopt the base rate method where we assign equal probabilities to possible outcomes. After computing the base rate probability of change in lockdown strategy for each country, we determine relative lack of uncertainty. The relative lack of uncertainty would mean simply to continue with the base rate method used as a forecasting mechanism. Otherwise in case of presence of uncertainty the base rate method may not be supported for forecasting. The following equations are used for calculating the base rate and relative lack of uncertainty.
Base rate = Increase/total times (1)
Relatives lack of uncertainity = base rate x (1- base rate) (2)
In this case, establishing the adequacy of the probabilistic estimates is very important. Brier score gives accuracy of probabilistic estimates.[7] Thus, to find if the forecast will perform well we compute the relative lack of uncertainty component of the Brier score.[7] With actual COVID-19 data,[8] we analyze the probability of change of strategy (that is, making lockdown more stringent) for the selected countries grounded on the overall base death rate of the world. To keep things simple, we code each probability of change of strategy by a country to binary outcomes (coded outcome). The “coded outcome” translates into decision “No Change,” or “Increase” depending on the overall base death rate of the world due to COVID-19. As of May 29, 2021,[8] the overall base death rate of the world due to COVID-19 was ≈ 0.07% (0.0007), as given in [Table 1]. It has been calculated from the overall daily deaths as a fraction of the number of people tested to be positive based on the globally available data on COVID-19.[8] | Table 1: Daily world death rate due to coronavirus disease 2019 (COVID-19)
Click here to view |
Further, to develop a better insight into the relationship between variables, e.g., Total Deaths with New Tests and, etc., from the dataset,[9] correlation analysis was carried out. The following equation was used for determining correlation between the presented variables. As an example the equation below determines the correlation coefficient, r(Total Deaths to New Tests), of Total Deaths to New Tests. Similarly, the other coefficients were determined and plotted a heatmap.
Sampling
Despite being a global pandemic, the infection, recovery and death rates are not uniformly distributed across all countries. Almost 90 percent of all COVID-19 deaths are in North America and Europe while the rest of the world accounts for only 10 percent[8]; a situation not as critical compared to countries experiencing larger outbreaks of local transmission, for instance, the USA and Italy. In the most recent situation report published by the World Health Organization (WHO),[10] there are clusters of cases in Pakistan. Many geographical, socio-economic, cultural, and healthcare facts could possibly be the reason for this discrepancy, but this discussion is beyond the scope of the current research. Many theories are put forward by researchers to explain this divergence in corona virus behavior across the countries. Few of the prominent theory suggest that the use of Bacillus Calmette–Guérin (BCG)[11] vaccine and antimalarial drugs[12] has created immunity in the receiving populations.
Overall rate
As per [Table 1], the overall base death rate of the world is found to be 0.0007 and that the relative lack of uncertainty is calculated to be 0.00074. Moreover, the data in [Table 2], [Table 3], [Table 4], [Table 5], [Table 6], [Table 7] have helped in calculating the base rates and relative lack of uncertainty for the six countries (part of research), shown in [Table 8]. During the starting days of spread of virus, the base rates were quite high but over the time of nearly 1.5 years, the data are providing a better picture. The overall base rate of USA is found to be 0.0007 which is similar to the overall base rate of the world, its relative lack of uncertainty is 0.00069 which assures the decisions regarding the closure or opening of the country can be made on the basis of this data. Only Italy (0.0014) is found to be having its overall base rate higher than that of the world with a relative lack of uncertainty around 0.00138. The other four countries are found to have their base rates less than that of the world, as Pakistan, Bangladesh, Czechia, and Nigeria have 0.0004, 0.0006, 0.0003, and 0.0004, with the relative lack of uncertainty of 0.00038, 0.00056, 0.00034, 0.00044, respectively. The overall data points toward the thought of applying better strategies instead of nationwide lockdowns as they have drastically affected the economical conditions of all the countries, especially the underdeveloped countries like Pakistan and Nigeria, and even the other four countries as well. We propose that strategies like smart lockdown, which aim to isolate the people of a particular area having high spread of virus, to be more impactful in dealing with the virus and less harmful for the economy. This strategy was applied in different parts of Pakistan over the year after the country endured huge losses due to the nationwide lockdown, and proved to be helpful.
Year-wise comparison
As now we have the data comprising the years 2020 and 2021, so we can look at the damage caused by the virus in a year-wise comparison. As per data of [Table 1], in 2020, the overall base death rate of the world was 0.0010 with relative lack of uncertainty around 0.00104 while that in 2021 those values are 0.0001 and 0.00009, respectively. This reflects that the world is learning to deal with the effect of virus and that it is reducing the virus's effect with time. In [Table 9], similar trend is seen in the base rates of all the countries (part of research) with USA (0.00100 and 0.00100) and Italy (0.00197 and 0.00197) having base rates and relative lack of uncertainty in 2020, respectively, while that in 2021 they reduced to (6.1807E-05 and 6.1803E-05) for USA and (0.00011 and 0.00011) for Italy. Similarly, Pakistan (0.00053 and 0.00053), Bangladesh (0.00084 and 0.00084), Czechia (0.0004 and 0.00040), and Nigeria (0.00066 and 0.00066) had these base rates and relative lack of uncertainty in 2020, respectively, while in 2021 they almost became negligible (0.00010 and 0.00010), (5.0522E-05 and 5.0519E-05), (0.00010 and 0.00010), and (3.7879E-05 and 3.7874E-05) for these countries, in the same order as mentioned above. This yearly comparison makes the claim made above regarding world learning from the damages of the virus and effectively tackling it, more stronger, and further proves that the world is also succeeding in herd immunity for the virus. | Table 9: Base rate and relative lack of uncertainty computations (Year-wise Comparison)
Click here to view |
Vaccination and its effects
The world saw the destruction wreaked by COVID-19 as over 3 million people lost their lives due to the virus.[8] Since, isolation and lockdowns can never be considered as the permanent solutions of dealing with this virus, vaccine is seen as the ultimate solution. COVID-19 vaccines manufactured by different pharmaceutical companies came out at the end of 2020 and were made available to different countries on different dates. So, to understand the effects of vaccination, we have also drawn comparison of the six countries (part of research) with the data of before and after the vaccination time period as per [Table 10]. In USA, mass vaccination started on 14 Dec “2020 and that in Italy started on 27 Dec” 2020. The base rates and relative lack of uncertainty of USA and Italy before vaccination was (0.00104 and 0.00104) and (0.00199 and 0.00199), respectively, while post vaccination the numbers massively reduced to (6.9494E-05 and 6.9490E-05) for USA and (0.00012 and 0.00012) for Italy Similarly, the base rates and relative lack of uncertainty for Pakistan, Bangladesh, Czechia, and Nigeria prior to vaccination were (0.00044 and 0.00044), (0.00074 and 0.00074), (0.00039 and 0.00039), and (0.00055 and 0.00055) while these numbers (in same order) reduced to (0.00012 and 0.00012), (5.7906E-05 and 5.7903E-05), (9.1038E-05 and 9.1030E-05), and (9.5645E-05 and 9.5644E-05). It is worth noting that all four of these countries started their mass vaccination campaigns in different months of 2021. This comparison emphasizes the need for vaccination and indicates why vaccination is necessary and needs to be made mandatory for everyone. It also reflects that post-vaccination, the base death rates have drastically come down which reinforces the need for vaccination and deconstructs the irrelevant conspiracies and propaganda against vaccination globally. The base rate and relative lack of uncertainty analysis is further validated through correlational analysis. In [Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5], [Figure 6] the correlational analysis of each country in study shows a negative correlation, which means that post immunization results in decrease of infections. | Table 10: Base rate and relative lack of uncertainty computations (Pre-Vaccination vs Post-Vaccination)
Click here to view |
| Results and Discussion | | |
Empirical evidence and selection
To form an informed opinion about the present situation of the outbreak that may have some relevance to Pakistan, we studied the data of five countries USA, Italy, Bangladesh, Czech Republic, and Nigeria. The selection was based on the following primary criteria:
- Coverage of Bacillus Calmette–Guérin (BCG) vaccine administered at birth.
- Countries where a major cause of illness is malaria and quinine is used for its treatment.
- Italy and the USA were selected because of high rate of infection and deaths.
Our hypothesis for selection of countries is that the countries where BCG vaccines and/or antimalarial drugs are administered for different diseases the chances of developing acute pulmonary diseases will be lesser and thus COVID-19 infections will be less fatal. It is based on the assumption reported in literature that in countries where BCG is administered at birth, the chances of acute pulmonary diseases are highly unlikely.[11] [Table 11] shows the current and past BCG policies and practices in selected countries under study. | Table 11: Coverage of Bacillus Calmette–Guérin (BCG), source: BCG world atlas
Click here to view |
Further, the data of the same set of countries for malaria and use of quinine for its treatment are also used to study the effects on acute pulmonary diseases.[12] [Table 12] shows cases of malaria, its treatment, and prevention for the same set of countries.
It is worth to mention here a study conducted on mice by,[13] on studying the effects of quinine on influenza virus infection, which showed that it is a possible treatment for virus infection.
Consolidating the data from [Table 11] and [Table 12] with the data on COVID-19 (number of tests carried out, infections and deaths), it is worth noting that in countries where BCG is or was administered for Tuberculosis (TB) immunization or quinine was administered for treating malaria the total number of deaths is considerably less, as shown in [Table 13]. It is worth mentioning that this is stated as a fact and no causal relationship is proposed.[14],[15],[16]
It is also worth noting that the Czech Republic carried out a large number of COVID-19 tests compared to Pakistan, Bangladesh and Nigeria, yet their total number of deaths still stands low. One may argue that the analysis does not take the population into account, however, our analysis relies on the number of COVID-19 infections reported based on the total number of test carried out. Taking this into account it can be seen that the countries where BCG immunization for TB and quinine for treatment of malaria is administered, the people are at lower risk of suffering from acute pulmonary diseases including COVID-19. According to the data in [Table 13], despite the fact that the effects of COVID-19 infections may be largely asymptomatic, the total number of deaths in these countries are still lower compared to other countries (such as Italy and USA) where people seem to be at higher risks of suffering from acute pulmonary diseases. We categorize these countries (Italy and USA) as high risk as they have never been immunized for pulmonary diseases or treated for malaria which indirectly also treats viral infections such as COVID-19. Continuing our discussion in context of the outbreak in Pakistan, we argue the situation is not the same as the rest of the world, and it calls for a complete rethink of our response and future strategy.
Resulting forecasts and discussion
The values with respective codes are shown in [Table 14]. In [Table 2], [Table 3], [Table 4], [Table 5], [Table 6], [Table 7], a binary code is assigned based on the decision “No Change”/”Increase” to daily new deaths of the selected set of countries. Plugging the equations (1) and (2) in the datasets of [Table 2], [Table 3], [Table 4], [Table 5], [Table 6], [Table 7] of each country, the base rates and relative lack of uncertainty are determined, as shown in [Table 8]. Analyzing the base rate probability of change of strategy with relative lack of uncertainty, the forecast suggesting lack of uncertainty would mean to simply go with the base rate. Thus, in case of Pakistan, there is no probability of change and thus it may not be wise to panic and wrongly predict an exponential increase in death rates due to COVID-19 out of fear only. On the other hand, in case of USA and Italy since the relative lack of uncertainty was quite low but the base rates were higher 71% and 56% in May 2020, more severe lockdowns and social distancing measures had to be put into place for minimizing an already increased death rate due to COVID-19.
Limitations and future research direction
One limitation of this study is that it does not analyze the latent factors that might have caused the non-uniformity of death rate across the globe. However, for selecting countries for comparative study, based on existing theories, only two factors: BCG vaccination and antimalarial drugs, are considered to have contributed toward reduced death rate. Neither establishing causal relationship is in the scope of this research nor was any such relationship proposed.
In addition, this research did not take into consideration the age as a factor of probability of death. It might be of interest to analyze the median age, and life expectancy with age distribution of COVID-19 deaths to get better understanding of the vulnerable population. One of the interesting aspect of future research would be to understand the behavioral and decision making aspects and of implementing a complete cutoff and lockdown all over the world so quickly.
| Conclusion | | |
COVID-19 restrictions have been or are being lifted throughout the world, including those social distancing, lockdowns, and the mandatory use of face masks. In retrospect, the initial world (over) reaction can be attributed to the overblown estimates of certain models, whose logic seems questionable now. The reportedly extremely high infection rate of COVID-19 combined with excessive number of estimated deaths based on hypothetical projections created a global wave of economically harsh measures. It is understandable as due to its novel nature, it was difficult to predict COVID-19 future course without sufficient data. Now that we have some actual data, much more realistic and evidence-based picture can be projected. Based on actual data, it can be seen that the effects of COVID-19 are not uniformly distributed across the globe and over time. The world is learning to deal with the virus and the reduced numbers in the tables prove this. Moreover, the comparison based on years 2020 and 2021 also reflects that with time the effect of virus decreased. The number of deaths caused by COVID-19 in some countries including Pakistan is an order of magnitude less than the initial projections. In fact the death rate, whether taken separately or as percentage of infected population is not as catastrophic in many countries including Pakistan as initially projected.
The analysis shows that a better measure of seriousness of COVID-19 pandemic is the death rate and not the infection rate. Therefore, the countries do not need to increase the strictness with increasing numbers, rather they can confidently relax the lockdown measures while analyzing the actual death-rate and projecting the realistic trend. Thus, we recommend using the good quality data, and applying wise approaches like smart lockdown to tackle any the virus spread in the future. This would result in more sensible policymaking supporting economic activities and helping the poor, especially in developing countries. Lastly, the analysis of pre-vaccination and post-vaccination shows significant decrease in the death rates and thus, reflects that vaccine is effectively tackling the threat caused by corona virus.
Financial support and sponsorship
Nil.
Conflicts of interest
There are no conflicts of interest.
| References | | |
| 1. | |
| 2. | |
| 3. | Sackett DL. Evidence-based medicine. Semin Perinatol 1997;21:3-5.  |
| 4. | Liu Z, Magal P, Seydi O, Webb G. Predicting the cumulative number of cases for the COVID-19 epidemic in China from early data. medRxiv 2020;2020.03.11.20034314. doi: 10.1101/2020.03.11.20034314. |
| 5. | Giesecke J. The invisible pandemic. Lancet 2020;395:e98. |
| 6. | |
| 7. | Stephenson DB, Coelho CA, Jolliffe IT. Two extra components in the Brier score decomposition. Weather and Forecasting 2008;23:752-7. |
| 8. | |
| 9. | |
| 10. | |
| 11. | Shet A, Ray D, Malavige N, Santosham M, Bar-Zeev N. Differential COVID-19- attributable mortality and BCG vaccine use in countries. medRxiv 2020;2020.04.01.20049478. doi: 10.1101/2020.04.01.20049478. |
| 12. | Devaux CA, Rolain J-M, Colson P, Raoult D. New insights on the antiviral effects of chloroquine against coronavirus: What to expect for COVID-19? Int J Antimicrob Agents 2020;55:105938. |
| 13. | Seeler AO, Graessle O, Ott WH. Effect of quinine on influenza virus infections in mice. J Infect Dis 1946;79:156–8. |
| 14. | |
| 15. | |
| 16. | |
[Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5], [Figure 6]
[Table 1], [Table 2], [Table 3], [Table 4], [Table 5], [Table 6], [Table 7], [Table 8], [Table 9], [Table 10], [Table 11], [Table 12], [Table 13], [Table 14]
|
