{"title":"中东和北非与冠状病毒大流行相关的死亡风险因素的估计","authors":"Sami Khedhiri","doi":"10.1111/jebm.12538","DOIUrl":null,"url":null,"abstract":"<p>Several issues related to the coronavirus pandemic have not yet been fully and unequivocally identified despite the attempts made in the literature to explain why COVID-19 case-fatality rates vary among countries and why in some developed countries, fatality rates were high. For instance, a study<span><sup>1</sup></span> investigated the clinical risk factors of COVID-19 fatality, which might include obesity and diabetes.<span><sup>2</sup></span> Other researchers studied the impact of demographic factors such as age, socioeconomic factors, environmental factors, and all these indicators combined.<span><sup>3</sup></span> Countries worldwide reported different case-fatality rates (CFR), a measure defined as the proportion of cases of COVID-19 that were fatal within a specified time. However, these differences in mortality rates might not be attributed to just the above-mentioned factors. There were other explanations, including the difference in the number of people tested and the characteristics of the healthcare system. Countries with fewer resources might have a higher mortality risk because their hospitals became overwhelmed with the increased number of infections.</p><p>The current study applied statistical methods to investigate the association between COVID-19 deaths and potential clinical, demographic, and socioeconomic risk factors. In addition, the variation of case-fatality rates across the countries and over time was also studied. There was ample research published in the literature about this issue,<span><sup>4</sup></span> with evidence from different countries and regions. Our contribution was to examine how the results for the MENA (Middle East and North Africa) region compared with results from other populations and whether the association of pandemic mortality and risk factors was confounded with population genotypes and racial differences, as these factors had not been sufficiently emphasized in the literature.</p><p>The MENA region includes 22 countries and makes up 6% of the world's population and more than 50% of the world's total oil reserves. Like most countries, the Middle East and North Africa have had their share of human and economic losses because of the COVID-19 outbreak, and as of April 2022, it was estimated that nearly 20 million people had been infected and 300 thousand had died from the coronavirus in the region.<span><sup>5</sup></span> Although governments in the MENA countries at first reacted swiftly to contain the coronavirus by implementing strict health protocols and developing policy and institutional plans to support households and firms, which helped to limit the first wave of the pandemic, however, after relaxing health restrictions in summer 2020, the situation quickly diverged and cases and death tolls rapidly increased. To compare this situation with the European management of the pandemic, a study<span><sup>6</sup></span> found that in the first phase of the pandemic, the inefficiency of the health systems was relatively high in Western Europe, both during the relaxation phase and in the second wave. The study found that European countries were severely affected at the pandemic's beginning. However, unlike the MENA countries, the Europeans were able to take adequate measures, and they succeeded in improving the efficiency of their healthcare systems. The MENA countries differed quite notably in their per-capita GDP, per-capita health expenditure, and health system characteristics. For example, in the wealthier Gulf countries, the per-capita GDP in Qatar is nearly 14 times higher than in Egypt or Tunisia, and it is more than 37 times greater than in Syria. However, Qatar has less than half the number of physicians per 1000 people compared to Israel. Also, according to recent World Bank data, the percentage of seniors (aged 65 years or more) in the UAE is only one-third of the percentage of seniors in neighboring Saudi Arabia or Kuwait. When we look at the clinical factors, the World Bank data shows that there is more than twice as much diabetes prevalence in Saudi Arabia as in Iran. These notes clearly illustrated the remarkable differences between the countries, which were geographically located in the same region, and the objective of this paper was to investigate whether this could explain why case-fatality rates varied significantly among them.</p><p>Several papers emerged recently to analyze the effects of the pandemic outbreak, to model and forecast the infections and deaths, and to study the effects of the pandemic in the MENA region.<span><sup>7</sup></span> Also, novel dynamic measures of case-fatality variations<span><sup>8</sup></span> were suggested. However, there have been only limited studies to investigate the association between risk factors and the pandemic fatalities in the region. For instance, it was found that the mortality rate in Kuwait was higher in older patients with comorbidities such as hypertension and cardiovascular diseases in Kuwait,<span><sup>9</sup></span> and a study related to Turkish COVID-19 patients showed that age, COPD, and smoking represented risk factors for mortality. The current paper was among the first to study the association between potential risk factors and COVID-19 lethality in the Middle East and North Africa based on statistical modeling with longitudinal data and to deal with the issue of among-country differences in the coronavirus fatality rates, which could clarify the regional discrepancies in pandemic mortality.</p><p>Publicly available data on daily COVID-19 cases and deaths for 18 MENA countries were collected from the Statista website. These included Algeria, Bahrain, Egypt, Iran, Iraq, Israel, Kuwait, Lebanon, Libya, Morocco, Oman, Palestine, Qatar, Saudi Arabia, Syria, Tunisia, Turkey, and the United Arab Emirates (UAE). Due to issues with data reliability, the other countries of the region were not included in the study. The data covered the period from March 24, 2020, to April 21, 2021, for 394 daily time series observations. Also, data from the World Bank were collected to retrieve the following variables: the per-capita GDP (<i>gdp</i>) of each country, the number of hospital beds per 1000 persons (<i>hospt</i>), the number of physicians per 1000 people (<i>doct</i>), the percentage of diabetes prevalence (<i>diab</i>), the percentage of senior citizens aged 65 years or more (<i>senior</i>), the percentage of smokers (<i>smoke</i>), and the per-capita health expenditure (<i>health</i>). The statistical analysis investigated these variables to verify if they were associated with pandemic mortality and if they constituted significant factors for across-country variations in fatality rates.</p><p>Where <i>Y</i> was the response variable with a lognormal distribution, <math>\n <semantics>\n <mrow>\n <mi>β</mi>\n <mspace></mspace>\n </mrow>\n <annotation>$\\beta \\;$</annotation>\n </semantics></math>was the vector of fixed-effect parameters, α was the vector of random effects parameters, and <i>X</i> and <i>Z</i> were the design matrices for the fixed and random effects, respectively. <i>U</i> contained the residual components. The model assumed that each observation was independent. However, there might be some interdependence in the response variable, which was given by the case-fatality measure, in relation to some factors, namely the study variables that would be investigated in this paper. To deal with this issue, a random effect was added into the model that allowed to assume a different baseline response value for each factor. The study model the individual differences in relation to each factor by assuming different random intercepts for each response. Such a model was called a mixed model since it contained the usual fixed effects as seen in linear regression, and one or more random effects, essentially giving some structure to the error term characterizing variation due to some factor level.</p><p>In the next step of the analysis, Equation (1) would be estimated with the penalized quasi-likelihood method by applying the Laplace approximation in a quasi-likelihood formulation of the model. It was noticed that the transformed mortality rate was not a discrete count, and thus using penalized quasi-likelihood would produce desirable unbiased statistical estimates in a linear mixed model regression with panel data. It should also be reminded that penalization was a method used to remove stability issues for the parameter estimates, which usually arise when the likelihood function was flat, therefore, when it became difficult to compute the maximum likelihood estimates using standard approaches. The response variable (<i>Y</i>) was formed by adding one to the ratio of the number of deaths divided by the number of confirmed cases. First, using probability plots in R, the distribution of case-fatality rates was checked, and the results showed that it was not normal. Next, the lognormal distribution was applied, and the results showed that the distribution provided the best fit of the response variable. Also, fatality rate variations between countries were presented by running the dependent variable (<math>\n <semantics>\n <mrow>\n <mi>r</mi>\n <mi>a</mi>\n <mi>t</mi>\n <mspace></mspace>\n <msub>\n <mi>e</mi>\n <mrow>\n <mi>i</mi>\n <mi>t</mi>\n </mrow>\n </msub>\n <mo>=</mo>\n <mspace></mspace>\n <mn>1</mn>\n <mo>+</mo>\n <mspace></mspace>\n <mi>c</mi>\n <mi>f</mi>\n <msub>\n <mi>r</mi>\n <mrow>\n <mi>i</mi>\n <mi>t</mi>\n </mrow>\n </msub>\n <mrow>\n <mo>)</mo>\n </mrow>\n </mrow>\n <annotation>$rat\\;{e_{it}} = \\;1 + {\\rm{\\;}}cf{r_{it}})$</annotation>\n </semantics></math> on fixed factors which included country-id, time, and an interaction term (county-id × time). The results displayed in the center left panel of Table 1 shows statistical evidence of between-country differences in the pandemic lethality rates for the MENA countries and prove that CFR measure varies significantly over time and across countries.</p><p>The random effect parameter was given by (<math>\n <semantics>\n <msub>\n <mi>α</mi>\n <mrow>\n <mi>O</mi>\n <mi>t</mi>\n </mrow>\n </msub>\n <annotation>${\\alpha _{Ot}}$</annotation>\n </semantics></math>) and the <math>\n <semantics>\n <mrow>\n <msup>\n <mi>β</mi>\n <mo>′</mo>\n </msup>\n <mi>s</mi>\n </mrow>\n <annotation>$\\beta ^{\\prime}s$</annotation>\n </semantics></math> represent fixed effects as explained in model Equation (1). The statistical results proved that the model which was most supported by the data should include the percentage of seniors and the diabetic prevalence or their interaction, plus either the per-capita GDP or the per-capita health expenditure, but not both because of high collinearity between the two variables. The upper panels of Table 1 list the results of both regression models. It should be noted, however, that the results were not significant when we included seniors, diabetes, and their interactions all together. The study findings could be interpreted by noticing that model 2 results found strong evidence that countries with higher per-capita health expenditure had, on average, lower fatality rates due to COVID-19. Also, there was strong statistical evidence that in the MENA region, the pandemic was more lethal in countries with higher percentages of diabetes prevalence for seniors. Model 3 results also showed evidence that countries with higher per-capita health expenditure have, on average, lower fatality rates due to COVID-19. In addition, higher percentage of seniors was strongly associated with more deadliness of the pandemic, but diabetes prevalence, if not confounded with seniors, was not a strong predictor of mortality. Furthermore, as indicated in the center left panel of Table 1, it found evidence of significant variation across countries in terms of their COVID-19 case-fatality rates, and thus it was suggested that data analysis techniques based on functional principal components (FPC) might be used to identify the type of these variations. This analysis was performed with R coding using the <i>FPCA</i> function of the <i>fdapace</i> package. The B splines-based smoother was used to convert the panel CFR trajectories into functions with the smoothing parameter selected by the generalized cross-validation technique, and the underlying functions to COVID-19 case-fatality measures could be estimated and represented. The lower panel of Table 1 shows that the first 3 components account for most of the explained mortality rate variation. In fact, the fraction of explained variation reached 90% with only 3 eigen components and 6 components needed to achieve a 99% level of sample variance.</p><p>The study highlighted that the presence of diabetes alone was not a strong risk factor for increased COVID-19 mortality; however, it was shown that the confounding of demographic characteristics (age) and diabetes represented major risk factors. This proved that senior patients who had diabetes were at a higher risk of dying from the disease in the Middle East region. This result of confounding effect also aligned with a similar finding in a study related to US asthma patients,<span><sup>10</sup></span> where the study showed that asthma alone was not a significant clinical factor. However, they warned that older coronavirus patients with asthma were at increased risk of hospitalization due to COVID-19. Some of our findings aligned with previous research results for the demographic variables. For instance, the statistical analysis showed no impact of smoking on increased mortality from the coronavirus; a similar result was concluded in a study,<span><sup>11</sup></span> which suggested that smoking increased the risk of severe disease in hospitalized COVID-19 patients but showed no significant association between smoking and increased mortality despite some differences in the results between the Chinese and US studies. However, the results differed from other research findings regarding the effects of smoking on COVID-19 severity and eventually death risk. In fact, a recent study based on European ancestry participants<span><sup>12</sup></span> showed that smoking increased susceptibility to sepsis and severe COVID-19, and conducted<span><sup>13</sup></span> a study using the UK Biobank cohort and concluded that results from two analytical approaches supported a causal effect of smoking on the risk of severe coronavirus, which might potentially lead to death. These results were regional-specific, and they were related to European COVID-19 patients. The current study was specific to the MENA countries and found no significant association between smoking and potential death from the pandemic. Furthermore, the paper showed different results regarding the effects of some healthcare services on COVID-19 lethality compared to other European studies. For example, a study in France<span><sup>14</sup></span> revealed that COVID-19 mortality rate was associated with the physician's density, and a Yale study<span><sup>15</sup></span> found that the lack of ICU hospital beds was associated with excess COVID-19 deaths. However, in the current study, there was no significant association between pandemic mortality and the number of hospital beds or the number of physicians in the MENA region. Also, to test the significance of economic factors, it was shown that the per-capita GDP and the per-capita healthcare expenditure, which was used as continuous variables in mixed-effect models, were important factors for COVID-19 fatality. Thus, our statistical analysis determines that in the MENA countries with lower per-capita GDP and/or lower per-capita health care expenditure, the risk of dying from the coronavirus disease is significantly higher than in the wealthier MENA countries.</p><p>The differences in the findings between the current paper and other studies, which are related to European and Asian populations, point to the importance of considering regional-specific factors and the role of ethnicity, racial differences, and population genotypes and their association with COVID-19 deadliness. Accounting for these factors in future research will add exciting knowledge in identifying the clinical stages for the progression of the disease to become lethal.</p><p>No funding information to declare.</p><p>There is no conflict of interest to declare.</p><p>This research uses publicly available data, so ethical approval is not applicable.</p>","PeriodicalId":16090,"journal":{"name":"Journal of Evidence‐Based Medicine","volume":null,"pages":null},"PeriodicalIF":3.6000,"publicationDate":"2023-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1111/jebm.12538","citationCount":"0","resultStr":"{\"title\":\"Estimation of death risk factors associated with the coronavirus pandemic in the Middle East and North Africa\",\"authors\":\"Sami Khedhiri\",\"doi\":\"10.1111/jebm.12538\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p>Several issues related to the coronavirus pandemic have not yet been fully and unequivocally identified despite the attempts made in the literature to explain why COVID-19 case-fatality rates vary among countries and why in some developed countries, fatality rates were high. For instance, a study<span><sup>1</sup></span> investigated the clinical risk factors of COVID-19 fatality, which might include obesity and diabetes.<span><sup>2</sup></span> Other researchers studied the impact of demographic factors such as age, socioeconomic factors, environmental factors, and all these indicators combined.<span><sup>3</sup></span> Countries worldwide reported different case-fatality rates (CFR), a measure defined as the proportion of cases of COVID-19 that were fatal within a specified time. However, these differences in mortality rates might not be attributed to just the above-mentioned factors. There were other explanations, including the difference in the number of people tested and the characteristics of the healthcare system. Countries with fewer resources might have a higher mortality risk because their hospitals became overwhelmed with the increased number of infections.</p><p>The current study applied statistical methods to investigate the association between COVID-19 deaths and potential clinical, demographic, and socioeconomic risk factors. In addition, the variation of case-fatality rates across the countries and over time was also studied. There was ample research published in the literature about this issue,<span><sup>4</sup></span> with evidence from different countries and regions. Our contribution was to examine how the results for the MENA (Middle East and North Africa) region compared with results from other populations and whether the association of pandemic mortality and risk factors was confounded with population genotypes and racial differences, as these factors had not been sufficiently emphasized in the literature.</p><p>The MENA region includes 22 countries and makes up 6% of the world's population and more than 50% of the world's total oil reserves. Like most countries, the Middle East and North Africa have had their share of human and economic losses because of the COVID-19 outbreak, and as of April 2022, it was estimated that nearly 20 million people had been infected and 300 thousand had died from the coronavirus in the region.<span><sup>5</sup></span> Although governments in the MENA countries at first reacted swiftly to contain the coronavirus by implementing strict health protocols and developing policy and institutional plans to support households and firms, which helped to limit the first wave of the pandemic, however, after relaxing health restrictions in summer 2020, the situation quickly diverged and cases and death tolls rapidly increased. To compare this situation with the European management of the pandemic, a study<span><sup>6</sup></span> found that in the first phase of the pandemic, the inefficiency of the health systems was relatively high in Western Europe, both during the relaxation phase and in the second wave. The study found that European countries were severely affected at the pandemic's beginning. However, unlike the MENA countries, the Europeans were able to take adequate measures, and they succeeded in improving the efficiency of their healthcare systems. The MENA countries differed quite notably in their per-capita GDP, per-capita health expenditure, and health system characteristics. For example, in the wealthier Gulf countries, the per-capita GDP in Qatar is nearly 14 times higher than in Egypt or Tunisia, and it is more than 37 times greater than in Syria. However, Qatar has less than half the number of physicians per 1000 people compared to Israel. Also, according to recent World Bank data, the percentage of seniors (aged 65 years or more) in the UAE is only one-third of the percentage of seniors in neighboring Saudi Arabia or Kuwait. When we look at the clinical factors, the World Bank data shows that there is more than twice as much diabetes prevalence in Saudi Arabia as in Iran. These notes clearly illustrated the remarkable differences between the countries, which were geographically located in the same region, and the objective of this paper was to investigate whether this could explain why case-fatality rates varied significantly among them.</p><p>Several papers emerged recently to analyze the effects of the pandemic outbreak, to model and forecast the infections and deaths, and to study the effects of the pandemic in the MENA region.<span><sup>7</sup></span> Also, novel dynamic measures of case-fatality variations<span><sup>8</sup></span> were suggested. However, there have been only limited studies to investigate the association between risk factors and the pandemic fatalities in the region. For instance, it was found that the mortality rate in Kuwait was higher in older patients with comorbidities such as hypertension and cardiovascular diseases in Kuwait,<span><sup>9</sup></span> and a study related to Turkish COVID-19 patients showed that age, COPD, and smoking represented risk factors for mortality. The current paper was among the first to study the association between potential risk factors and COVID-19 lethality in the Middle East and North Africa based on statistical modeling with longitudinal data and to deal with the issue of among-country differences in the coronavirus fatality rates, which could clarify the regional discrepancies in pandemic mortality.</p><p>Publicly available data on daily COVID-19 cases and deaths for 18 MENA countries were collected from the Statista website. These included Algeria, Bahrain, Egypt, Iran, Iraq, Israel, Kuwait, Lebanon, Libya, Morocco, Oman, Palestine, Qatar, Saudi Arabia, Syria, Tunisia, Turkey, and the United Arab Emirates (UAE). Due to issues with data reliability, the other countries of the region were not included in the study. The data covered the period from March 24, 2020, to April 21, 2021, for 394 daily time series observations. Also, data from the World Bank were collected to retrieve the following variables: the per-capita GDP (<i>gdp</i>) of each country, the number of hospital beds per 1000 persons (<i>hospt</i>), the number of physicians per 1000 people (<i>doct</i>), the percentage of diabetes prevalence (<i>diab</i>), the percentage of senior citizens aged 65 years or more (<i>senior</i>), the percentage of smokers (<i>smoke</i>), and the per-capita health expenditure (<i>health</i>). The statistical analysis investigated these variables to verify if they were associated with pandemic mortality and if they constituted significant factors for across-country variations in fatality rates.</p><p>Where <i>Y</i> was the response variable with a lognormal distribution, <math>\\n <semantics>\\n <mrow>\\n <mi>β</mi>\\n <mspace></mspace>\\n </mrow>\\n <annotation>$\\\\beta \\\\;$</annotation>\\n </semantics></math>was the vector of fixed-effect parameters, α was the vector of random effects parameters, and <i>X</i> and <i>Z</i> were the design matrices for the fixed and random effects, respectively. <i>U</i> contained the residual components. The model assumed that each observation was independent. However, there might be some interdependence in the response variable, which was given by the case-fatality measure, in relation to some factors, namely the study variables that would be investigated in this paper. To deal with this issue, a random effect was added into the model that allowed to assume a different baseline response value for each factor. The study model the individual differences in relation to each factor by assuming different random intercepts for each response. Such a model was called a mixed model since it contained the usual fixed effects as seen in linear regression, and one or more random effects, essentially giving some structure to the error term characterizing variation due to some factor level.</p><p>In the next step of the analysis, Equation (1) would be estimated with the penalized quasi-likelihood method by applying the Laplace approximation in a quasi-likelihood formulation of the model. It was noticed that the transformed mortality rate was not a discrete count, and thus using penalized quasi-likelihood would produce desirable unbiased statistical estimates in a linear mixed model regression with panel data. It should also be reminded that penalization was a method used to remove stability issues for the parameter estimates, which usually arise when the likelihood function was flat, therefore, when it became difficult to compute the maximum likelihood estimates using standard approaches. The response variable (<i>Y</i>) was formed by adding one to the ratio of the number of deaths divided by the number of confirmed cases. First, using probability plots in R, the distribution of case-fatality rates was checked, and the results showed that it was not normal. Next, the lognormal distribution was applied, and the results showed that the distribution provided the best fit of the response variable. Also, fatality rate variations between countries were presented by running the dependent variable (<math>\\n <semantics>\\n <mrow>\\n <mi>r</mi>\\n <mi>a</mi>\\n <mi>t</mi>\\n <mspace></mspace>\\n <msub>\\n <mi>e</mi>\\n <mrow>\\n <mi>i</mi>\\n <mi>t</mi>\\n </mrow>\\n </msub>\\n <mo>=</mo>\\n <mspace></mspace>\\n <mn>1</mn>\\n <mo>+</mo>\\n <mspace></mspace>\\n <mi>c</mi>\\n <mi>f</mi>\\n <msub>\\n <mi>r</mi>\\n <mrow>\\n <mi>i</mi>\\n <mi>t</mi>\\n </mrow>\\n </msub>\\n <mrow>\\n <mo>)</mo>\\n </mrow>\\n </mrow>\\n <annotation>$rat\\\\;{e_{it}} = \\\\;1 + {\\\\rm{\\\\;}}cf{r_{it}})$</annotation>\\n </semantics></math> on fixed factors which included country-id, time, and an interaction term (county-id × time). The results displayed in the center left panel of Table 1 shows statistical evidence of between-country differences in the pandemic lethality rates for the MENA countries and prove that CFR measure varies significantly over time and across countries.</p><p>The random effect parameter was given by (<math>\\n <semantics>\\n <msub>\\n <mi>α</mi>\\n <mrow>\\n <mi>O</mi>\\n <mi>t</mi>\\n </mrow>\\n </msub>\\n <annotation>${\\\\alpha _{Ot}}$</annotation>\\n </semantics></math>) and the <math>\\n <semantics>\\n <mrow>\\n <msup>\\n <mi>β</mi>\\n <mo>′</mo>\\n </msup>\\n <mi>s</mi>\\n </mrow>\\n <annotation>$\\\\beta ^{\\\\prime}s$</annotation>\\n </semantics></math> represent fixed effects as explained in model Equation (1). The statistical results proved that the model which was most supported by the data should include the percentage of seniors and the diabetic prevalence or their interaction, plus either the per-capita GDP or the per-capita health expenditure, but not both because of high collinearity between the two variables. The upper panels of Table 1 list the results of both regression models. It should be noted, however, that the results were not significant when we included seniors, diabetes, and their interactions all together. The study findings could be interpreted by noticing that model 2 results found strong evidence that countries with higher per-capita health expenditure had, on average, lower fatality rates due to COVID-19. Also, there was strong statistical evidence that in the MENA region, the pandemic was more lethal in countries with higher percentages of diabetes prevalence for seniors. Model 3 results also showed evidence that countries with higher per-capita health expenditure have, on average, lower fatality rates due to COVID-19. In addition, higher percentage of seniors was strongly associated with more deadliness of the pandemic, but diabetes prevalence, if not confounded with seniors, was not a strong predictor of mortality. Furthermore, as indicated in the center left panel of Table 1, it found evidence of significant variation across countries in terms of their COVID-19 case-fatality rates, and thus it was suggested that data analysis techniques based on functional principal components (FPC) might be used to identify the type of these variations. This analysis was performed with R coding using the <i>FPCA</i> function of the <i>fdapace</i> package. The B splines-based smoother was used to convert the panel CFR trajectories into functions with the smoothing parameter selected by the generalized cross-validation technique, and the underlying functions to COVID-19 case-fatality measures could be estimated and represented. The lower panel of Table 1 shows that the first 3 components account for most of the explained mortality rate variation. In fact, the fraction of explained variation reached 90% with only 3 eigen components and 6 components needed to achieve a 99% level of sample variance.</p><p>The study highlighted that the presence of diabetes alone was not a strong risk factor for increased COVID-19 mortality; however, it was shown that the confounding of demographic characteristics (age) and diabetes represented major risk factors. This proved that senior patients who had diabetes were at a higher risk of dying from the disease in the Middle East region. This result of confounding effect also aligned with a similar finding in a study related to US asthma patients,<span><sup>10</sup></span> where the study showed that asthma alone was not a significant clinical factor. However, they warned that older coronavirus patients with asthma were at increased risk of hospitalization due to COVID-19. Some of our findings aligned with previous research results for the demographic variables. For instance, the statistical analysis showed no impact of smoking on increased mortality from the coronavirus; a similar result was concluded in a study,<span><sup>11</sup></span> which suggested that smoking increased the risk of severe disease in hospitalized COVID-19 patients but showed no significant association between smoking and increased mortality despite some differences in the results between the Chinese and US studies. However, the results differed from other research findings regarding the effects of smoking on COVID-19 severity and eventually death risk. In fact, a recent study based on European ancestry participants<span><sup>12</sup></span> showed that smoking increased susceptibility to sepsis and severe COVID-19, and conducted<span><sup>13</sup></span> a study using the UK Biobank cohort and concluded that results from two analytical approaches supported a causal effect of smoking on the risk of severe coronavirus, which might potentially lead to death. These results were regional-specific, and they were related to European COVID-19 patients. The current study was specific to the MENA countries and found no significant association between smoking and potential death from the pandemic. Furthermore, the paper showed different results regarding the effects of some healthcare services on COVID-19 lethality compared to other European studies. For example, a study in France<span><sup>14</sup></span> revealed that COVID-19 mortality rate was associated with the physician's density, and a Yale study<span><sup>15</sup></span> found that the lack of ICU hospital beds was associated with excess COVID-19 deaths. However, in the current study, there was no significant association between pandemic mortality and the number of hospital beds or the number of physicians in the MENA region. Also, to test the significance of economic factors, it was shown that the per-capita GDP and the per-capita healthcare expenditure, which was used as continuous variables in mixed-effect models, were important factors for COVID-19 fatality. Thus, our statistical analysis determines that in the MENA countries with lower per-capita GDP and/or lower per-capita health care expenditure, the risk of dying from the coronavirus disease is significantly higher than in the wealthier MENA countries.</p><p>The differences in the findings between the current paper and other studies, which are related to European and Asian populations, point to the importance of considering regional-specific factors and the role of ethnicity, racial differences, and population genotypes and their association with COVID-19 deadliness. Accounting for these factors in future research will add exciting knowledge in identifying the clinical stages for the progression of the disease to become lethal.</p><p>No funding information to declare.</p><p>There is no conflict of interest to declare.</p><p>This research uses publicly available data, so ethical approval is not applicable.</p>\",\"PeriodicalId\":16090,\"journal\":{\"name\":\"Journal of Evidence‐Based Medicine\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":3.6000,\"publicationDate\":\"2023-06-15\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"https://onlinelibrary.wiley.com/doi/epdf/10.1111/jebm.12538\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Journal of Evidence‐Based Medicine\",\"FirstCategoryId\":\"3\",\"ListUrlMain\":\"https://onlinelibrary.wiley.com/doi/10.1111/jebm.12538\",\"RegionNum\":2,\"RegionCategory\":\"医学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"MEDICINE, GENERAL & INTERNAL\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Evidence‐Based Medicine","FirstCategoryId":"3","ListUrlMain":"https://onlinelibrary.wiley.com/doi/10.1111/jebm.12538","RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"MEDICINE, GENERAL & INTERNAL","Score":null,"Total":0}
Estimation of death risk factors associated with the coronavirus pandemic in the Middle East and North Africa
Several issues related to the coronavirus pandemic have not yet been fully and unequivocally identified despite the attempts made in the literature to explain why COVID-19 case-fatality rates vary among countries and why in some developed countries, fatality rates were high. For instance, a study1 investigated the clinical risk factors of COVID-19 fatality, which might include obesity and diabetes.2 Other researchers studied the impact of demographic factors such as age, socioeconomic factors, environmental factors, and all these indicators combined.3 Countries worldwide reported different case-fatality rates (CFR), a measure defined as the proportion of cases of COVID-19 that were fatal within a specified time. However, these differences in mortality rates might not be attributed to just the above-mentioned factors. There were other explanations, including the difference in the number of people tested and the characteristics of the healthcare system. Countries with fewer resources might have a higher mortality risk because their hospitals became overwhelmed with the increased number of infections.
The current study applied statistical methods to investigate the association between COVID-19 deaths and potential clinical, demographic, and socioeconomic risk factors. In addition, the variation of case-fatality rates across the countries and over time was also studied. There was ample research published in the literature about this issue,4 with evidence from different countries and regions. Our contribution was to examine how the results for the MENA (Middle East and North Africa) region compared with results from other populations and whether the association of pandemic mortality and risk factors was confounded with population genotypes and racial differences, as these factors had not been sufficiently emphasized in the literature.
The MENA region includes 22 countries and makes up 6% of the world's population and more than 50% of the world's total oil reserves. Like most countries, the Middle East and North Africa have had their share of human and economic losses because of the COVID-19 outbreak, and as of April 2022, it was estimated that nearly 20 million people had been infected and 300 thousand had died from the coronavirus in the region.5 Although governments in the MENA countries at first reacted swiftly to contain the coronavirus by implementing strict health protocols and developing policy and institutional plans to support households and firms, which helped to limit the first wave of the pandemic, however, after relaxing health restrictions in summer 2020, the situation quickly diverged and cases and death tolls rapidly increased. To compare this situation with the European management of the pandemic, a study6 found that in the first phase of the pandemic, the inefficiency of the health systems was relatively high in Western Europe, both during the relaxation phase and in the second wave. The study found that European countries were severely affected at the pandemic's beginning. However, unlike the MENA countries, the Europeans were able to take adequate measures, and they succeeded in improving the efficiency of their healthcare systems. The MENA countries differed quite notably in their per-capita GDP, per-capita health expenditure, and health system characteristics. For example, in the wealthier Gulf countries, the per-capita GDP in Qatar is nearly 14 times higher than in Egypt or Tunisia, and it is more than 37 times greater than in Syria. However, Qatar has less than half the number of physicians per 1000 people compared to Israel. Also, according to recent World Bank data, the percentage of seniors (aged 65 years or more) in the UAE is only one-third of the percentage of seniors in neighboring Saudi Arabia or Kuwait. When we look at the clinical factors, the World Bank data shows that there is more than twice as much diabetes prevalence in Saudi Arabia as in Iran. These notes clearly illustrated the remarkable differences between the countries, which were geographically located in the same region, and the objective of this paper was to investigate whether this could explain why case-fatality rates varied significantly among them.
Several papers emerged recently to analyze the effects of the pandemic outbreak, to model and forecast the infections and deaths, and to study the effects of the pandemic in the MENA region.7 Also, novel dynamic measures of case-fatality variations8 were suggested. However, there have been only limited studies to investigate the association between risk factors and the pandemic fatalities in the region. For instance, it was found that the mortality rate in Kuwait was higher in older patients with comorbidities such as hypertension and cardiovascular diseases in Kuwait,9 and a study related to Turkish COVID-19 patients showed that age, COPD, and smoking represented risk factors for mortality. The current paper was among the first to study the association between potential risk factors and COVID-19 lethality in the Middle East and North Africa based on statistical modeling with longitudinal data and to deal with the issue of among-country differences in the coronavirus fatality rates, which could clarify the regional discrepancies in pandemic mortality.
Publicly available data on daily COVID-19 cases and deaths for 18 MENA countries were collected from the Statista website. These included Algeria, Bahrain, Egypt, Iran, Iraq, Israel, Kuwait, Lebanon, Libya, Morocco, Oman, Palestine, Qatar, Saudi Arabia, Syria, Tunisia, Turkey, and the United Arab Emirates (UAE). Due to issues with data reliability, the other countries of the region were not included in the study. The data covered the period from March 24, 2020, to April 21, 2021, for 394 daily time series observations. Also, data from the World Bank were collected to retrieve the following variables: the per-capita GDP (gdp) of each country, the number of hospital beds per 1000 persons (hospt), the number of physicians per 1000 people (doct), the percentage of diabetes prevalence (diab), the percentage of senior citizens aged 65 years or more (senior), the percentage of smokers (smoke), and the per-capita health expenditure (health). The statistical analysis investigated these variables to verify if they were associated with pandemic mortality and if they constituted significant factors for across-country variations in fatality rates.
Where Y was the response variable with a lognormal distribution, was the vector of fixed-effect parameters, α was the vector of random effects parameters, and X and Z were the design matrices for the fixed and random effects, respectively. U contained the residual components. The model assumed that each observation was independent. However, there might be some interdependence in the response variable, which was given by the case-fatality measure, in relation to some factors, namely the study variables that would be investigated in this paper. To deal with this issue, a random effect was added into the model that allowed to assume a different baseline response value for each factor. The study model the individual differences in relation to each factor by assuming different random intercepts for each response. Such a model was called a mixed model since it contained the usual fixed effects as seen in linear regression, and one or more random effects, essentially giving some structure to the error term characterizing variation due to some factor level.
In the next step of the analysis, Equation (1) would be estimated with the penalized quasi-likelihood method by applying the Laplace approximation in a quasi-likelihood formulation of the model. It was noticed that the transformed mortality rate was not a discrete count, and thus using penalized quasi-likelihood would produce desirable unbiased statistical estimates in a linear mixed model regression with panel data. It should also be reminded that penalization was a method used to remove stability issues for the parameter estimates, which usually arise when the likelihood function was flat, therefore, when it became difficult to compute the maximum likelihood estimates using standard approaches. The response variable (Y) was formed by adding one to the ratio of the number of deaths divided by the number of confirmed cases. First, using probability plots in R, the distribution of case-fatality rates was checked, and the results showed that it was not normal. Next, the lognormal distribution was applied, and the results showed that the distribution provided the best fit of the response variable. Also, fatality rate variations between countries were presented by running the dependent variable ( on fixed factors which included country-id, time, and an interaction term (county-id × time). The results displayed in the center left panel of Table 1 shows statistical evidence of between-country differences in the pandemic lethality rates for the MENA countries and prove that CFR measure varies significantly over time and across countries.
The random effect parameter was given by () and the represent fixed effects as explained in model Equation (1). The statistical results proved that the model which was most supported by the data should include the percentage of seniors and the diabetic prevalence or their interaction, plus either the per-capita GDP or the per-capita health expenditure, but not both because of high collinearity between the two variables. The upper panels of Table 1 list the results of both regression models. It should be noted, however, that the results were not significant when we included seniors, diabetes, and their interactions all together. The study findings could be interpreted by noticing that model 2 results found strong evidence that countries with higher per-capita health expenditure had, on average, lower fatality rates due to COVID-19. Also, there was strong statistical evidence that in the MENA region, the pandemic was more lethal in countries with higher percentages of diabetes prevalence for seniors. Model 3 results also showed evidence that countries with higher per-capita health expenditure have, on average, lower fatality rates due to COVID-19. In addition, higher percentage of seniors was strongly associated with more deadliness of the pandemic, but diabetes prevalence, if not confounded with seniors, was not a strong predictor of mortality. Furthermore, as indicated in the center left panel of Table 1, it found evidence of significant variation across countries in terms of their COVID-19 case-fatality rates, and thus it was suggested that data analysis techniques based on functional principal components (FPC) might be used to identify the type of these variations. This analysis was performed with R coding using the FPCA function of the fdapace package. The B splines-based smoother was used to convert the panel CFR trajectories into functions with the smoothing parameter selected by the generalized cross-validation technique, and the underlying functions to COVID-19 case-fatality measures could be estimated and represented. The lower panel of Table 1 shows that the first 3 components account for most of the explained mortality rate variation. In fact, the fraction of explained variation reached 90% with only 3 eigen components and 6 components needed to achieve a 99% level of sample variance.
The study highlighted that the presence of diabetes alone was not a strong risk factor for increased COVID-19 mortality; however, it was shown that the confounding of demographic characteristics (age) and diabetes represented major risk factors. This proved that senior patients who had diabetes were at a higher risk of dying from the disease in the Middle East region. This result of confounding effect also aligned with a similar finding in a study related to US asthma patients,10 where the study showed that asthma alone was not a significant clinical factor. However, they warned that older coronavirus patients with asthma were at increased risk of hospitalization due to COVID-19. Some of our findings aligned with previous research results for the demographic variables. For instance, the statistical analysis showed no impact of smoking on increased mortality from the coronavirus; a similar result was concluded in a study,11 which suggested that smoking increased the risk of severe disease in hospitalized COVID-19 patients but showed no significant association between smoking and increased mortality despite some differences in the results between the Chinese and US studies. However, the results differed from other research findings regarding the effects of smoking on COVID-19 severity and eventually death risk. In fact, a recent study based on European ancestry participants12 showed that smoking increased susceptibility to sepsis and severe COVID-19, and conducted13 a study using the UK Biobank cohort and concluded that results from two analytical approaches supported a causal effect of smoking on the risk of severe coronavirus, which might potentially lead to death. These results were regional-specific, and they were related to European COVID-19 patients. The current study was specific to the MENA countries and found no significant association between smoking and potential death from the pandemic. Furthermore, the paper showed different results regarding the effects of some healthcare services on COVID-19 lethality compared to other European studies. For example, a study in France14 revealed that COVID-19 mortality rate was associated with the physician's density, and a Yale study15 found that the lack of ICU hospital beds was associated with excess COVID-19 deaths. However, in the current study, there was no significant association between pandemic mortality and the number of hospital beds or the number of physicians in the MENA region. Also, to test the significance of economic factors, it was shown that the per-capita GDP and the per-capita healthcare expenditure, which was used as continuous variables in mixed-effect models, were important factors for COVID-19 fatality. Thus, our statistical analysis determines that in the MENA countries with lower per-capita GDP and/or lower per-capita health care expenditure, the risk of dying from the coronavirus disease is significantly higher than in the wealthier MENA countries.
The differences in the findings between the current paper and other studies, which are related to European and Asian populations, point to the importance of considering regional-specific factors and the role of ethnicity, racial differences, and population genotypes and their association with COVID-19 deadliness. Accounting for these factors in future research will add exciting knowledge in identifying the clinical stages for the progression of the disease to become lethal.
No funding information to declare.
There is no conflict of interest to declare.
This research uses publicly available data, so ethical approval is not applicable.
期刊介绍:
The Journal of Evidence-Based Medicine (EMB) is an esteemed international healthcare and medical decision-making journal, dedicated to publishing groundbreaking research outcomes in evidence-based decision-making, research, practice, and education. Serving as the official English-language journal of the Cochrane China Centre and West China Hospital of Sichuan University, we eagerly welcome editorials, commentaries, and systematic reviews encompassing various topics such as clinical trials, policy, drug and patient safety, education, and knowledge translation.