STRUCTURAL EQUATION MODELING OF THE COVID-19 INCIDENCE RATE ASSOCIATED WITH THE DEATH RATE AND THE IMPACT OF SOCIOECONOMIC FACTORS IN ASEAN COUNTRIES

The pandemic crisis caused by COVID-19 jeopardizes our progressive social, ecological, and economic development. Many studies have focused on its impact on sustainable development in emerging economies and rapid regional development. We attempted to create a study from a distinct point of view. We investigate how socioeconomic factors contribute to regional variation in the incidence and death rate of COVID-19 risk and attempt to develop statistical modeling with a focus on the Association of Southeast Asian Nations (ASEAN). We conducted quantitative research using structural equation modeling based on structure variance, also known as partial least square path modeling. Our findings suggest that socioeconomic factors have a significant impact on the incidence and mortality rate of COVID-19. COVID-19 infection rates are highest in developed countries in general. As discovered in this study, socioeconomic factors have a positive effect on the incidence rate but a negative effect on the death rate. The findings provide policy recommendations for controlling and preventing COVID-19 transmission and lowering the death rate in the ASEAN community. In general, the paper adds to our limited understanding of the socioeconomic consequences of the COVID-19 pandemic.

By early August 2021, globally, over 220 million cases had been announced by WHO, resulting in over 4.5 million deaths [4]. Despite extraordinary public health interventions such as social distancing, contact tracing, and full and partial lockdown, the disease burden has risen, with significant variation between countries and regions and many countries experiencing multiple waves [1,5]. The COVID-19 crisis exposed significant inequalities between countries worldwide [6]. As of 8 August 2021, the ASEAN countries had reported over 9.5 million cases of infection and over 212,730 deaths [7].
Understanding how pandemics differ in different countries will benefit in future pandemic response. Current research from high-and middle-income nations reveals that demographic characteristics, comorbidities, healthcare resources, and response rigor all contribute significantly to the incidence of COVID-19-related infections [1,8,9]. The effects of socioeconomic factors, on the other hand, remain unknown [10]. Liu et al. (2020) [10] and Zhang et al. (2021) [1] discovered an unexpectedly positive effect of the HDI on the confirmed rate and death rate. Higher HDI countries have more people with multiple chronic diseases, lower smoking rates, and a higher annual gross salary.
Despite the fact that the most developed countries have high numbers of cases, mortality rates in these countries are low. This could be a result of major geographical variations in the availability and accessibility of health care resources [8].
We developed a method for modeling structural equations (SEM) based on structure variance, also known as partial least squares path modeling (PLSPM). SEM is a type of statistical model that can be used to simultaneously model latent variables with complex relationships [11].
The SEM methodology is proposed in this study due to the high correlation between socioeconomic indicators such as income per capita, human development index, nurses and 3 STRUCTURAL EQUATION MODELING OF COVID-19 INCIDENCE RATE midwives, hospital bed availability, and population density. The socioeconomic indicators were obtained from the United Nation's report [12].
The paper is divided into four sections: Section 2 discusses the materials and techniques, Section 3 discusses the empirical findings, and Section 4 concludes with a discussion.

Study area
The Association of Southeast Asian Nations, or ASEAN, is comprised of eleven member countries: Indonesia, Myanmar, Laos, Malaysia, the Philippines, Singapore, Thailand, Vietnam, Timor-Leste, Brunei, and Cambodia. ASEAN is home to approximately 8.5% of the global population [13].    Figure 2 shows the cumulative incidence and death rates for COVID-19 in ASEAN countries. The spatial pattern of the death rate seems to not correspond to the incidence rate. To avoid the problem of underreporting, the data is transformed into weekly data. The data consists of the number of confirmed cases and deaths.

Socioeconomic Data
The United Nations Report 2020 provided socioeconomic data such as the number of people, population density, per capita income, human development index (HDI), nurses, midwives, and hospital beds. The data used in this study is presented in Table 1. The spatial patterns of the socioeconomic variables are presented in Figure 3.

Method
To model the socioeconomic factor on incidence and death rates at the same time, we use structural equation modeling (SEM) based on structural variance. Partial least square SEM (PLSPM) is a term used to describe the structure variance of SEM [14][15][16]. PLSPM is a strategy for causal modeling that prioritizes maximizing the variance explained by the dependent latent construct over generating a theoretical covariance matrix [15,16]. PLSPM can be used to estimate the complex relationship between exogenous and endogenous variables without imposing stringent assumptions about normality or sample size [17]. When working with small samples and asymmetric distributions, PLSPM is particularly effective because it evaluates latent variable scores directly. It can also make factor identification easier by introducing a flexible residual covariance structure. It can also make accurate predictions in the presence of interdependent observations [16]. PLSPM consists of two models: (i) a measurement model (outer model) and (ii) a structural model (inner model) [17]. PLSPM specifically follows a four-stage process. The first stage is to define the structural model. The second stage is the selection and specification of measurement models. At the third stage, the model is estimated the parameters model [15,18,19].
Stage 1: Specifying the structural model The following structure Figure 4 illustrates the research hypothesis: where 1 denotes incidence rate, 2 death rate, 1 is socioeconomic factor with 11 and 12 denote the effect of socioeconomic factor on incidence and death rates respectively with 21 is the effect of incidence rate to death rate. Stage 2: Selecting and specifying the measurement models Socioeconomic factor consists of five indicators including income per capita (USD) (x1), human development index (x2), nurses and midwives (nurse/10,000 people) (x3), hospital bed (bed/10,000 people) (x4), and population density (inhabitants/km 2 ) (x5). The path diagram of measurement model is presented in Figure 5.

Stage 3: Model estimation and validation
The ordinary least square method is used to estimate the standardized loading factors and the regression coefficients as follows: Reflective measurement: Model validation includes item validity and reliability [20]. Convergent validity is typically used to determine item validity, while composite reliability and average variance extraction are typically used to determine reliability.

Convergent validity
Convergent validity quantifies the degree to which an indicator and its latent factor are correlated.
If the indicator's loading factor is greater than 0.50, it has convergent validity [21].

Composite reliability
Composite reliability (occasionally referred to as "construct reliability") is an approximate measure of internal consistency, similar to Cronbach's alpha [22]. It can be thought of as the ratio of the total variance in true scores to the total variance in scale scores [23].
where denotes the number of indicators that are used to measure the construct, with 1 denotes the th standardized loading factors of first construct. Internal consistency is satisfied when the composite reliability of the items is greater than 0.70 [24].

Average variance extracted (AVE)
AVE is designed to measure the total variance of indicators that can be accounted by its latent factor (construct). The AVE is formulated as : where denotes the number of indicators that are used to measure the construct, with 1 denotes the th standardieed loading factors of first construct. AVE is also used to measure internal reliability and discriminant validity [25]. Items meeting the A given construct should have 9 STRUCTURAL EQUATION MODELING OF COVID-19 INCIDENCE RATE a minimum reliability of 0.5 in order to be considered reliable [21]. The square root of AVE should be larger than the correlation between the constructs to satisfy the discriminant validity. We perform PLS-PM using the plspm package in R [26]. Table 2 summarizes the descriptive statistics for the variables of the interest.  The Pearson correlation coefficients for covariates other than hospital bed were greater than 0.70 ( Figure 6), indicating a strong linear relationship between the socioeconomic indicators. As a result, several covariates have a VIF greater than 10 (Table 3), indicating a significant collinearity issue.

Descriptive Statistics
We employ structural equation modeling to address this condition by developing socioeconomic factors that account for socioeconomic indicators.

Structural equation modeling of socioeconomic factor against incidence and death rates
Structural equation modeling is used to investigate the impact of socioeconomic factors on incidence and mortality. Five reflective indicators make up a socioeconomic factor. These indicators include the income per capita (USD), the human development index (HDI), the number of nurses and midwives (per 10,000 inhabitants), the number of hospital beds (per 10,000 inhabitants), and the population density (inhabitants/km 2 ).

Reliability and validity tests
Two approaches are used to evaluate the reliability and validity of the reflective indicators. First, determine the composite reliability [21], which must exceed 0.70 (or at least not be less than 0.60).
Additionally, we present another measure of reliability, namely average variance extracted (AVE).
The AVE must be greater than or equal to 0.500. AVE is also frequently used to assess discriminant validity. Second, examine the standardized loading factor to determine the item's validity for each indicator. It must be greater than or equal to 0.70 [21]. If one or more items are identified with a standardized loading factor of less than 0.6, the composite reliability is first evaluated. If the 11 STRUCTURAL EQUATION MODELING OF COVID-19 INCIDENCE RATE composite reliability value for the item is less than 0.600, the item should be excluded from the analysis. However, if the composite reliability value remains greater than 0.700, items with a value of less than 0.400 may be retained. As shown in Table 4, the reflective construction exhibits a relatively high level of internal reliability, with a composite reliability value of 0.897, an AVE value of 0.781, and standardized loading factors ranging from 0.404 to 0.948. Based on these results, all items are declared reliable and valid.  The estimation of structural equation model parameters is shown in Table 5, and a detailed path diagram with parameter estimation is shown in Figure 7.  highest socioeconomic score, as well as the highest incidence and death rate. Malaysia, despite its strong socio-economic conditions, has not been fully able to control COVID-19. Thailand also scores highly in terms of socioeconomic status, with a low death rate but a high incidence rate.
Finally, our findings demonstrate that socioeconomic factors have a positive impact on the incidence rate of COVID-19 while having a negative impact on the death rate of COVID-19. The confirmed rate has an impact on the death rate in a positive way. Additionally, we discovered that the most important indicators of socioeconomic factors are nurses and midwives, income per capita (USD), and the human development index (HDI). The findings of this study indicate that developed countries have a higher confirmed rate than developing countries or low-middle-income countries. However, developed countries have significant resources to prevent high mortality rates by providing very adequate health facilities.

DISCUSSION AND CONCLUSION
The COVID-19 pandemic is the century's most extraordinary health problem and the greatest threat to humanity since the Second World War [27]. It has a significant adverse effect on public health and has socioeconomic consequences. It is rapidly spreading throughout the world, as Wuhan, China reported an outbreak on 30 December 2019 [28]. On 1 March 2020, the WHO declared COVID-19 to be a pandemic. As of 20 August 2021, the coronavirus COVID-19 has been transmitted into over 200 countries and territories worldwide (Worldometer 2021). COVID-19 has had a negative impact on the economic structure of every country worldwide, including Southeast Asian countries. The slowdown in economic growth occurred in the second and third quarters of 2020. Almost all sectors of the economy experienced a slowdown in growth. The debt of countries is increasing as a result of the enormous socio-economic consequences of COVID-19. Developed countries, in general, have the highest rates of COVID-19 infection. This study discovered that socioeconomic factors have a positive effect on the incidence rate but a negative effect on the death rate. Income per capita and the number of nurses and midwives are the two most important indicators of socioeconomic factors. This is because developed countries have higher population densities and mobility, which allow for faster transmission of the casting virus. However, they have adequate health care facilities, which helps to reduce the death rate from COVID-19.