Loading [MathJax]/jax/output/SVG/jax.js
Research article

Primary school children's health and its association with physical fitness development and health-related factors

  • The health status (HS) of children is influenced by a variety of factors, including physical fitness (PF) or social and environmental characteristics. We present a 4-year longitudinal study carried out with 263 primary school children. PF was assessed yearly using the German Motor Performance Test 6–18. Demographic data, leisure time behavior and socioeconomic factors were collected using questionnaires for children and parents. Based on parents' ratings in year 4, children were categorized as either “very good health status” (VGHS) or “good health status or below” (GHSB). Children with VGHS (73%) showed a larger improvement of global PF (p < 0.001), a significantly higher proportion of being/playing outside (p < 0.001), significantly lower proportions of overweight (p < 0.001), of media availability in the bedroom (p = 0.011) and of daily media consumption > 2 h (p = 0.033) compared to children with GHSB. Regarding socio-economic factors, children with VGHS revealed significantly fewer parents with lower education (p = 0.002), lower physical activity levels (p = 0.030) and lower migration background (p < 0.001). Physical fitness (p = 0.019) and outdoors exercising (p = 0.050) were the only variables to provide significantly higher chances of perceiving one's own health as very good when tested within a complex model including all the variables studied in this work. Considering the little focus on PF in the current Austrian physical education curriculum and the favorable environmental features of the Tyrolean region, more emphasis should be given to promoting didactical and pedagogical approaches that allow schoolers to be active in the nature.

    Citation: Gerhard Ruedl, Armando Cocca, Katharina C. Wirnitzer, Derrick Tanous, Clemens Drenowatz, Martin Niedermeier. Primary school children's health and its association with physical fitness development and health-related factors[J]. AIMS Public Health, 2024, 11(1): 1-18. doi: 10.3934/publichealth.2024001

    Related Papers:

    [1] Araya Ranok, Chanida Kupradit . Effect of whey protein and riceberry flour on quality and antioxidant activity under gastrointestinal transit of gluten-free cookies. AIMS Agriculture and Food, 2020, 5(3): 434-448. doi: 10.3934/agrfood.2020.3.434
    [2] Siti Hajar Othman, Nazirah Mohd Rosli, Norhazirah Nordin, Masturina Abdul Aziz . Formulation of crispy chicken burger patty batter: Properties and storage qualities. AIMS Agriculture and Food, 2022, 7(2): 426-443. doi: 10.3934/agrfood.2022027
    [3] Nionela V. Kononenko, Aleksandra A. Sharova, Larisa I. Fedoreyeva . Tissue damage to wheat seedlings (Triticum aestivum) under salt exposure. AIMS Agriculture and Food, 2020, 5(3): 395-407. doi: 10.3934/agrfood.2020.3.395
    [4] Nicole Roberta Giuggioli, Thais Mendes da Silva, Simone Dimitrova . Geographical indication (GI) branded quality: a study case on the homogeneity of the Carota Novella di Ispica Region. AIMS Agriculture and Food, 2021, 6(2): 538-550. doi: 10.3934/agrfood.2021031
    [5] Mutshidzi Matidza, Tsietsie Ephraim Kgatla, Mpho Edward Mashau . Leavening capacity, physicochemical and textural properties of wheat dough enriched with non-commercial unripe banana flours. AIMS Agriculture and Food, 2023, 8(4): 978-994. doi: 10.3934/agrfood.2023052
    [6] Luh Suriati, I Made Supartha Utama, Bambang Admadi Harsojuwono, Ida Bagus Wayan Gunam, I Made Adnyana, Ahmad Fudholi . Nano-ecogel to maintain the physicochemical characteristics of fresh-cut mangosteen. AIMS Agriculture and Food, 2021, 6(4): 988-999. doi: 10.3934/agrfood.2021059
    [7] Elena Y. Osipenko, Yuliya Y. Denisovich, Galina A. Gavrilova, Ekaterina Y. Vodolagina . The use of bioactive components of plant raw materials from the far eastern region for flour confectionery production. AIMS Agriculture and Food, 2019, 4(1): 73-87. doi: 10.3934/agrfood.2019.1.73
    [8] Md. Shaha Nur Kabir, Kamal Rasool, Wang-Hee Lee, Seong-In Cho, Sun-Ok Chung . Influence of delayed cooling on the quality of tomatoes (Solanum lycopersicum L.) stored in a controlled chamber. AIMS Agriculture and Food, 2020, 5(2): 272-285. doi: 10.3934/agrfood.2020.2.272
    [9] Mariyana Strateva, Deyan Stratev, Georgi Zhelyazkov . Freezing influence on the histological structure of Mediterranean mussel (Mytilus galloprovincialis). AIMS Agriculture and Food, 2023, 8(2): 278-291. doi: 10.3934/agrfood.2023015
    [10] Ibrahim Fusseini, Stephen K. Torkpo, Kwame Afreh-Nuamah . Assessment of levels of temperature conditions on the effectiveness of multiple-layer hermetic bag for bio-rational management of aflatoxin in stored maize. AIMS Agriculture and Food, 2016, 1(3): 342-353. doi: 10.3934/agrfood.2016.3.342
  • The health status (HS) of children is influenced by a variety of factors, including physical fitness (PF) or social and environmental characteristics. We present a 4-year longitudinal study carried out with 263 primary school children. PF was assessed yearly using the German Motor Performance Test 6–18. Demographic data, leisure time behavior and socioeconomic factors were collected using questionnaires for children and parents. Based on parents' ratings in year 4, children were categorized as either “very good health status” (VGHS) or “good health status or below” (GHSB). Children with VGHS (73%) showed a larger improvement of global PF (p < 0.001), a significantly higher proportion of being/playing outside (p < 0.001), significantly lower proportions of overweight (p < 0.001), of media availability in the bedroom (p = 0.011) and of daily media consumption > 2 h (p = 0.033) compared to children with GHSB. Regarding socio-economic factors, children with VGHS revealed significantly fewer parents with lower education (p = 0.002), lower physical activity levels (p = 0.030) and lower migration background (p < 0.001). Physical fitness (p = 0.019) and outdoors exercising (p = 0.050) were the only variables to provide significantly higher chances of perceiving one's own health as very good when tested within a complex model including all the variables studied in this work. Considering the little focus on PF in the current Austrian physical education curriculum and the favorable environmental features of the Tyrolean region, more emphasis should be given to promoting didactical and pedagogical approaches that allow schoolers to be active in the nature.



    A first episode of pneumonia with unknown cause was detected in Wuhan at the end of 2019 and reported to the World Health Organization (WHO) Country Office in China on 31 December 2019 [1]. This was the first case of the novel coronavirus (SARS-CoV-2) that quickly spread all over the world in the past few months. The WHO declared the outbreak a Public Health Emergency of International Concern on 30th January 2020; on 11th March 2020 WHO further determined that COVID-19 could be characterized as a pandemic [1].

    Several studies [2][4] have concentrated on the biological and epidemiological factors governing COVID-19 transmission, while few others [5] have investigated the potential impact of socio-economic characteristics on governing the extent of COVID-19 diffusion in the population. Societal and economic factors can be of critical importance for accuracy of models of the outbreak because of the economic [6],[7] and health impacts of the drastic measures that have been put in place in an effort to slow the spread of the disease in those same countries (e.g social distancing, quarantine, lockdowns, testing, and reallocation of hospital resources) [8], [9] and worldwide.

    In this paper, we take a reverse perspective and analyze how socio-economic determinants pre-dating the pandemic (data taken not later than 2019) relate to the number of reported cases, deaths, and the ratio detahs/cases of COVID-19 via machine learning methods. Our focus is on understanding the connection between epidemiological variables of the COVID-19 pandemic and the (i) level of health care infrastructure, (ii) general health of the population, (iii) economic factors, (iv) demographic structure, (v) environmental health, (vi) societal characteristics, and (vii) religious characteristics of a country. We hypothesize that different countries have different specific socio-economic features and incidence of the disease and therefore the implementation of government measures must be thoughtful and data evidence-driven and heterogeneous across countries and across resources in order to effectively combat COVID-19.

    We analyze 32 interpretable models, including (i) regression models with both independent and proximity dependent outcomes; and (ii) variable selection through LASSO. After this step, we build two indices of variable importance for each of the determinants to estimate their overall association with the the number of cases, deaths, and deaths/cases rate. An Absolute Importance Index (AII) and a Signed Importance Index (SII) are constructed. The AII determines the presence of the variable in the top-10 absolute correlation ranking, while the SII takes in consideration the sign of the correlation. By focusing only on those findings that are common to a majority, the findings are less sensitive to the limitations of any single model considered.

    Our analysis determines that the socio-economic status of a country follows some sort of Action-Reaction Principle, as it is not only heavily influenced by the pandemic (we did not address this in the study, but it is an established fact in the literature [10][12]), but it is also a distinct factor of the pandemic and must be taken into consideration by governments. The level of mobility, the quality of the health care system, the economic status of a country, and the features of a society are associated with the number of cases and deaths due to COVID-19.

    Our work suggests that government resources (both in the form of equipment and staff) must not be allocated blindly, and highlights that different countries might benefit from different measures based on the specific country socio-economic status.

    The remaining part of the manuscript is organized as follows. Section 2 is dedicated to a brief description of the datasets used, the description of the epidemiological variables that we are considering, and the description of the socio-economic determinants involved in our study. Moreover, it includes a summary of the methods used. More details on this are present in the Supplementary Material (Appendixes S1, S2, S3, S4). Section 3 is dedicated to the results of our analysis, while Section 4 to the discussion of the results, and Section 5 to the conclusions.

    Remark. A complete literature review is not feasible, given that the global effort of researchers around the world has produced a massive amount of results on COVID-19. We apologize for all citations that are missing.

    This section is dedicated to (i) the description of the datasets used for the input variables, the outcome variables, and the geographical weighting; and (ii) the methods used for the statistical analysis (linear regression, LASSO, and MICE). We refer to the Supporting Information for more details about the Data Sources (S1), technical results about our statistical methods (S2), and the tables of Descriptive Statistics (S3), and Importance indices (S4).

    Multiple datasets have been combined for the analysis (See Appendix S1 for the specific information on the data sources, and relative websites). The datasets were then organized in three groups based on their role in the models: (i) the outcome variables Y, namely epidemiology variables, such as the number of cases, the number of deaths, and the ratio between them; (ii) the predictors X which include the socio-economic determinants; and (iii) the weighting matrix A, which includes geographic information.

    The total number of reported cases and deaths attributed to COVID-19 as of 2nd May 2020 were obtained from Our World In Data [13] and used as outcome variables of our models.

    As the only predictors of our models, 44 SE variables were chosen for our analyses based on their potential explanatory power and to facilitate comparisons with other published works [5]. Data were obtained from publicly available databases [14][20] for a total of 199 countries/regions, 32 of which only had data for all 44 variables of interest (see Appendix S1 for more details). Given the fact that the years for which data were available varied by country/region, we chose to use the most recent data available for each country, (oldest being 2010, most recent being 2019). SE factors were divided into 7 categories based upon the common theme to which each of our 44 variables most closely aligned. These categories are similar to some that have been used previously [5]: (i) Capacity of a country to deal with COVID-19 cases (Healthcare Infrastructure); (ii) Statistics indicative of the health of the population of a country (Health Statistics); (iii) Economic situation and tourism/mobility in a country (Economic Health); (iv) Demographic structure of a country, in particular the age structure and the spatial distribution of the population (Demographic Structure); (v) Societal characteristics such as the level of education, the possibility of access to technology, and features of government (Societal Characteristics); (vi) Pollution level and ecological footprint (Environmental Health); and (vii) Religious practices in the population (Religious Characteristics). The division of variables into each of these categories can be seen in Appendix S1.

    Latitude and longitude coordinates for capital cities were obtained for each country from the CEPII (Centre d'Études Prospectives et d'Informations Internationales) GeoDist database. Coordinates for 11 countries were not found in the GeoDist database and were obtained from Google.

    Using these coordinates, we calculated the pairwise distances between cities using the Spherical Law of Cosines. Given two cities C1, C2 on the surface of the Earth with latitude and longitude coordinates (α1, β1) and (α2, β2), and assuming a constant radius of the Earth R = 6378 kms, we have that the distance between C1 and C2 is given by the following formula

    d(C1,C2)=arccos(sin(α1)*sin(α2)+cos(α1)*cos(α2)*cos(β2β1))*R.

    The matrix A has then been computed such that the entries were given by

    Aij=exp{d(Ci,Cj)}exp{d(,)}

    for i, j = 1,..., 199. Here d(·,·) a normalization factor computed as the average over the distances between every city Ci and every city Cj. We considered the ellipticity of the Earth to have a minor influence on our results and we we considered the spherical approximation appropriate. Our strategy relates to the theoretical work [21].

    Multiple countries did not report values for some of the SE determinants, so we decided to do imputation, in order to be able to include all countries (both those with and without any missing information). To perform imputation, we used the R package mice which performs imputation via Multivariate Imputation by Chained Equations (MICE). The method assumes that the probability that a value is missing depends only on observed values and not on unobserved values [22], [23]. We assumed this throughout all our analyses.

    The MICE algorithm [22], [23] produces a series of regression models, where each variable with missing data is modeled conditional upon the other variables in the data, and so according to its distribution. In the first step, MICE performs a simple imputation for every missing value in the dataset. The imputations for one of the variables are set back to missing, and this variable is considered as the dependent variable in a regression model with all the other variables used as independent variables, under the assumptions valid in generalized linear models [24]. The missing values for the variable playing the role of dependent variable at this step are then replaced with predictions (imputations) from the regression model. This procedure is then repeated for every variable and in an iterative fashion. At the end of the iteration process, MICE outputs one imputed dataset, and after the process stabilized, the distribution of the parameters governing the imputations is produced. The algorithm is independent on the order in which the variables are imputed. For a summary of the method, we refer to Appendix S2.

    We considered 5 different outcome variables: (i) Y1 = # cases; (ii) Y2 = # deaths; (iii) 1 = # cases/total population; (iv) 2 = # deaths/total population; and (v) Y0 = # deaths/# cases. To determine how total population impacted each of the outcome variables 1, 2, and Y0 (those dependent variables scaled by population size), we ran each of these models with two different sets of explanatory variables: (i) All variables (|X| = 44); and (ii) All but total population (POP) count (|X| = 43).

    Although the complete dataset contained 199 countries/regions, missing values resulted in preliminary linear regression models excluding 135 of those countries/regions [25]. As mentioned, we imputed the missing values using the mice package in R [22] and re-ran our models, so as to include all 199 countries. Models for automatic variable selection, such as LASSO [26], [27] were also considered. We used R Studio Version 1.2.5042 and libraries readxl, readr, gdata, mice, glment, caret for the computations.

    To measure the importance of the variables across our models, we built two indices. An Absolute Importance Index (AII) and a Signed Importance Index (SII). The AII counts the number of time a variable appears in the top-10 correlated variables based on absolute correlation. On the other side, the SII counts the presence of a variable in the top-10 correlated variables with sign. For example if the variable X appears 12 times in the top-10 highest correlated variables with Y, 10 times positively correlated, then AII = 12, while SII = 8. We computed such indices globally and for each single type of outcome variable. Please refer to Appendix S4 for more details.

    The results of the models produced with the imputed dataset are reported below in Table 1. Multiple and adjusted R2 values that describe the fit of each model can be found in Table 1 as well. We will describe the specific results about the importance of the variables grouping them by the 7 classes as in Appendix S1, but by joining Health Infrastructures and Health Statistics in one single subsection for convenience of explanation.

    Table 1.  Multiple and adjusted R2 for all linear regression models performed on the MICE-imputed dataset, containing data for 199 countries/regions. Models that included total population contained 44 socio-economic variables while those excluding total population contained 43. All models were run with and without the weighting matrix A based on geographical distance between the largest cities in each country/region. Abbreviations: Pop. Tot. = Population Total; Geo. Weight. = Weighting by Geographical Distance Matrix A.
    Outcome Variable With Pop. Tot. No Geo. Weight.
    With Geo. Weight.
    Mult. R2 Adj. R2 Mult. R2 Adj. R2
    # cases (Y1) Y 0.8532 0.8113 0.5626 0.4376
    # deaths (Y2) Y 0.7918 0.7323 0.6206 0.5122
    2*# cases/Pop. Tot. (1) Y 0.5135 0.3745 0.5225 0.3860
    N 0.5135 0.3785 0.5205 0.3874
    2*# deaths/Pop. Tot. (2) Y 0.4238 0.2592 0.5249 0.3892
    N 0.4235 0.2636 0.5229 0.3905
    2*# deaths/# cases (Y0) Y 0.2997 0.0996 0.4942 0.3597
    N 0.2973 0.1023 0.4891 0.3474

     | Show Table
    DownLoad: CSV

    In all the figures, the results are grouped based on the category. Half of the models included weighting with a geographical distance matrix (see Section 2 for more details). The number of times each variable appeared among the top-10 highest correlated variables in these models was tallied with its own sign (sign + if positively correlated, sign - if negatively correlated). The magnitude and direction of each bar represents the signed percentage of this tally. The letters in the figures represent the way the 44 socio-economic variables are divided into categories. A: Health Infrastructure, B: Health Statistics, C: Environmental Health, D: Economic Health, E: Demographic Structure, F: Societal Characteristics, and G: Religious Characteristics. Tables of detailed raw, signed, and weighted tallies can be found in Appendix S4.

    The number of physicians, essential health coverage index, and death rate were among the top-10 variables in 25%, 25%, 15.63% of our models, respectively (Panels A,B in Figures 16; Appendix S4). The number of physicians correlated positively with Y1, 1, Y2, and 2 and negatively with Y0. Access to essential health services consistently correlated positively in our models, appearing in 50% of Y1, Y2, and Y0 models, though completely absent from the top-10 variables in 1 and 2 models. Similarly, crude death rate correlated positively in 25–50% of Y1, Y2, and Y0 models, but was never selected among the top-10 variables in 1 and 2 models.

    Number of nurses and midwives, number of hospital beds, prevalence of diabetes, and crude birth rate consistently correlated negatively in 37.5%, 50%, 21.88%, and 43.75% of our models, respectively (Panels A,B in Figures 16; Appendix S4). Number of hospital beds appeared in all categories of models; it was least common in Y1 at 25%, most common in Y2 at 75%, and was in 50% of the models in the remaining categories. Number of nurses and midwives were identified as important in 100% of 1 models and 50% of 2 models but were notably absent from all other models. Birth rate appeared among our top-10 variables in 25% of Y1 and 1 models and 50% of 2 models. Prevalence of diabetes appeared in 25–75% of all model categories, with a notable absence from Y1 models.

    Figure 1.  Indices of importance SII of socio-economic variables in models of number of COVID-19 cases (Y1). We analyzed the effects of 44 socio-economic determinants in a total of 4 models, with number of COVID-19 cases as the outcome variable.
    Figure 2.  Indices of importance SII of socio-economic variables in models of number of COVID-19 cases/population total (1). We analyzed the effects of 44 socio-economic determinants in a total of 8 models, with number of COVID-19 cases/population as the outcome variable (4 with total population included among the determinants, and 4 with total population removed).
    Figure 3.  Indices of importance SII of socio-economic variables in models of number of COVID-19 deaths (Y2). We analyzed the effects of 44 socio-economic determinants in a total of 4 models, with number of COVID-19 deaths as the outcome variable.
    Figure 4.  Indices of importance SII of socio-economic variables in models of number of COVID-19 deaths/population total (2). We analyzed the effects of 44 socio-economic determinants in a total of 8 models, with number of COVID-19 deaths/population as the outcome variable (4 with total population included among the determinants, and 4 with total population removed).
    Figure 5.  Indices of importance SII of socio-economic variables in models of number of COVID-19 deaths/cases (Y0). We analyzed the effects of 44 socio-economic determinants in a total of 8 models, with number of COVID-19 deaths/cases as the outcome variable (4 with total population included among the determinants, and 4 with total population removed).
    Figure 6.  Indices of importance SII of socio-economic variables in all models of COVID-19 cases, deaths, and deaths/cases. We analyzed the effects of 44 socio-economic determinants in a total of 32 models, with number of COVID-19 cases, cases/population total, deaths, deaths/population total, and deaths/cases as outcome variables.

    Domestic government health expenditure was the only measure of economic health that was identified among the top-10 variables of all categories of models and consistently correlated positively with the outcome variables in all those models (Panel D in Figures 16; Appendix S4). In contrast, employment to population ratio and income distribution consistently correlated negatively with the outcome variables in all models and were identified as important determinants in 43.75% of all our models (Panel D in Figures 16; Appendix S4). GDP, trade, and government lending/borrowing correlated inconsistently (different for each model) across the models, with GDP and trade only identified as important variables in models lacking the geographical weighting matrix (Panel D in Figures 16; Appendix S4). Unemployment rate was never identified as important in any of the models.

    Total population correlated negatively with Y1 and Y2, but became insignificant for 1, 2, and Y0 (Panel E in Figures 16; Appendix S4). Population density, the proportion of the population over the age of 65, and the proportion of immigrants (people of international stock) consistently correlated positively and were important factors in 28.13%, 25%, and 31.25% of our models, though never in Y0 (Panel E in Figures 16; Appendix S4). Although rural population was identified as an important variable in all categories of models, the pattern of correlation was inconsistent, correlating negatively in Y1, Y2, and Y0 models and positively in 1 and 2 models (Panel E in Figures 16; Appendix S4).

    The number of airline passengers per year was nearly always important for Y1 and Y2 models, correlating positively in 100% and 75% of those models, respectively, though it never appeared in 1, 2, or Y0 models (Panel D in Figures 16; Appendix S4). Although the number of tourist arrivals was consistently positively correlated with 50% of Y1, Y2, and Y0 models, the direction of correlation was inconsistent in 1 and 2, despite being identified as a key determinant in each of those model categories (appearing in 75% and 100% of those models, respectively; Panel D in Figures 16; Appendix S4). The geographical weighting matrix was always present in models in which tourist arrivals correlated negatively, while the absence of this matrix in the models always coincided with tourist arrivals correlating positively.

    Ecological footprint was identified as an important variable in 25% of Y1, Y2, and Y0 models, in which it consistently correlated positively (Panel C in Figures 16; Appendix S4). Air pollution never appeared as an important determinant.

    The proportion of individuals with access to internet was identified as an important determinant in 28.13% of all our models; in 1, 2, and Y0 models, it correlated negatively, whereas in Y2 models it correlated positively (Panel F in Figures 16; Appendix S4). Government effectiveness and economic freedom score correlated negatively with Y0 and were identified as important in 75% of those models, but no other models (Panel F in Figures 16; Appendix S4). Personal freedom also correlated negatively with Y0 in 50% of those models, though it correlated positively in one Y1 model (Panel F in Figures 16; Appendix S4). Human freedom correlated positively in 50% of Y0 models, but negatively in 25% of 1 and 2 models (Panel F in Figures 16; Appendix S4). The average number of people per household appeared in 25% of 1 models, in which it correlated positively, but was not important in any of the other models (Panel F in Figures 16; Appendix S4). Education level, rule of law, and control of corruption never appeared among the top-10 variables in any of our models.

    The percentage of the population identifying as Christian appeared in 40.63% of models and was a consistently positive correlate in all model categories (Panel G in Figures 16; Appendix S4). In contrast, the percentage of the population identifying as Buddhist, appeared in 50% of Y1, Y2, and Y0 models and was a consistently negative correlate in those models (Panel G in Figures 16; Appendix S4). The remaining religious categories appeared rarely among the most important variables in our models, but when present, most correlated negatively (Panel G in Figures 16; Appendix S4).

    We divide our discussion into sections based on the category of socio-economic data. The relationships discussed here are all multivariate correlations, namely correlations occurring in models that contain all socio-economic data.

    In countries in which the population has greater access to essential healthcare services and in which the government invests more capital into healthcare, the results surprisingly showed an increase in the number of cases, number of deaths, and number of deaths/cases of COVID-19. In contrast, countries that have greater numbers of nurses and midwives and hospital beds per capita and in which diabetes is more prevalent have smaller numbers of cases, deaths, and deaths/cases. Interestingly, the number of physicians correlated positively with the number of cases and number of deaths, but negatively with the number of deaths/cases. Although seemingly contradictory, taken together, these data may provide indications that government healthcare spending needs to be allocated appropriately in order to effectively combat diseases like COVID-19. This explanation is still not satisfactory, as the possible inefficiencies in the healthcare system do not explain why countries with less capital investment in healthcare would not be struggling with the same inefficiencies. A further reason for this surprising result could be that the population of developed countries is more mobile, and some government epidemic prevention measures are not effective, resulting in higher number of cases and deaths. To narrow down this range of possible interpretations, we would need (among the others) more detailed data about resource allocation to hospitals (e.g. whether or not they experienced shortages of critical equipment). In addition to meeting basic space requirements in the form of hospital beds, access to medical personnel, like nurses and midwives, who interact for greater periods of time directly with patients, may facilitate the treatment of and recovery from both chronic conditions like diabetes and acute conditions like COVID-19. This may be a particularly important consideration for developing countries [6], which may have less effective medical infrastructure in place. The analysis seems also to caution against clustering doctors into large, centralized healthcare facilities when suitable care can be provided at home or in less dense facilities.

    Employment to population ratio and income distribution, as measured by the GINI index, were identified as important variables in 43.75% of models, and were consistently negatively correlated with number of cases, number of deaths, and number of deaths/cases of COVID-19, both with and without standardization by population. These data suggest that countries in which greater proportions of the population are employed, and where there is less economic disparity within the population, can be expected to feel the effects of COVID-19 less strongly. Indeed, these results appear to be corroborated by the current condition in the United States, a country in which concern about widening economic stratification and discrepancy is frequently discussed and the number of cases and deaths of COVID-19 are continuing to rise rapidly at the time of writing this article (First Trimester of 2021).

    Although the roles of GDP and trade were less prominent (appearing in only 12.5% and 6.25% of models, respectively) and were inconsistently correlated with COVID-19 variables, it is important to note that these variables only appeared in models in which there was no weighting by geographical distance. This is interesting given that [5] found that GDP was a strong positive predictor of both COVID-19 cases and deaths. Our results may indicate an influence of geographical clustering of countries with similar economic strength/health, and could potentially be used as an indication of regions in which countries can be expected to exhibit similarities in vulnerability to diseases like COVID-19.

    Variables relating to demographic structure consistently played a bigger role in models of number of cases and number of deaths than in number of deaths/cases. Indeed, only a single demographic variable (rural population) appeared in only one out of eight of our models in which the number of deaths/cases was the outcome variable. Thus, although the percentage of the population aged 65+, population density, and the percentage of the population of international origin (immigrants) all positively correlated with the number of cases and number of deaths, they do not appear to have an influence in models of the death rate when standardized by the number of cases of COVID-19. Total population correlated negatively with number of COVID-19 cases and deaths (Y1 and Y2), but it was no longer identified as an important variable when these outcome variables were standardized by population (1 and 2) or in models using deaths/cases (Y0). This may suggest a sub-exponential growth of the number of infections [30].

    The degree of mobility, as indicated by the number of airline passengers, correlated positively with number of cases, number of deaths, and number of deaths/cases, though this effect seemed to disappear in models that were standardized by total population size. Although short-term travel restrictions play an important role in reducing the impact of COVID-19, the fact that the number of airline passengers does not appear in models with outcome variables scaled by population size (1 and 2) suggests that increased mobility does not increase COVID-19 cases or deaths to a level disproportionate with the total population of the country. Further, our other variable that directly measures mobility (number of tourist arrivals per year) showed an interesting reversal in correlation structure when weighting by geography was added to our models: tourist arrivals correlated positively in models without the geographical weighting structure, but switched to correlating negatively in all models weighted with geographical data. Interestingly, ecological footprint, which in part could be heavily influenced by domestic mobility (e.g. car travel), appeared in a few models and was consistently positively correlated with cases, deaths, and deaths/cases. Thus, it is clear that the relationship between international travel and COVID-19 is a complicated one, and blanket policies restricting travel may not accurately reflect the impact that such mobility may have on the impact of COVID-19 in national populations.

    In general, religion, or lack thereof, plays a very minor role in our models, the exceptions being the percentage of the population identifying as Christian, which consistently correlates positively with all our outcome variables, and the percentage of the population identifying as Buddhist, which consistently correlates negatively with all our outcome variables. This is interesting given the speculation early in the pandemic that religious gatherings could be sources of superspreading events [31].

    Societal characteristics consistently increase in importance in models in which the number of deaths are standardized by the number of cases compared to models with number of cases or number of deaths as the outcome variables. Further, the majority of the societal variables that correlated strongly in our models had a net negative correlation. In particular, countries in which a greater proportion of their population have greater economic freedom (freedom to voluntarily acquire and dispose of their property [19]), and to a lesser extent personal freedom (freedom of movement, assembly, religion etc. in addition to safety and security [19]), and in which the government is more effective, tend to have lower numbers of deaths/cases. Further, countries in which internet usage is more widespread tend to see a drop in the number of cases, number of deaths, and number of deaths/cases. In contrast, education level, control of corruption, and rule of law never appeared among the top-10 variables in any of our models. Together, these results may suggest that countries with governments capable and effective at enacting policies for the protection of their citizens, who in their turn have the resources to keep themselves informed and freedom to act in their own best interests, may fare better against COVID-19. Note that the implied association between internet access and a more informed public is somehow speculative, possibly non-obvious in the days of misinformation on social media, and certainly requires further attention on its own. It might be more plausible that countries with high levels of internet access are simply in a better position to have a larger number of people work from home, which may not be an option otherwise.

    Our study is static and photographs the situation at 2nd May 2020. Furthermore, looking at data collected at a single point in time does not take in consideration the fact that the pandemic did not begin to spread in every location simultaneously. For example, European countries began to be affected much earlier than the US. Thus, the strong association between investment into healthcare and number of cases may simply be a consequence of the fact that Europe, a region where healthcare investment is consistently high, was one of the earliest regions to be affected. Also, regions that are affected in the early stages of a pandemic may be hardest-hit even with high capital investment simply because, at that point in time, the medical community still lacks the data and expertise necessary to effectively treat patients. Finally, it may also be that countries with more developed health infrastructure are able to test and diagnose more patients and that COVID-19 deaths will then be more likely to be correctly attributed to the disease.

    All of these limitations suggest to follow up this analysis with the analysis of the time evolution of the disease, its relationship with socio-economic covariates, and to account for the possibility of time-lag between different countries.

    On another note, it has been shown that underreporting can influence the severity of the pandemic [2], [28], [29], [32]. A future work will incorporate estimates of underreporting in our models.

    In this paper, we have studied the relationship between socio-economic determinants and the reported number of cases, deaths, and the ratio of deaths/cases in each country during the first months of the COVID-19 pandemic, by means of machine learning methods. We analyzed a total of 32 interpretable models and built two importance indices (AII and SII) for the covariates. Our statistical models included linear regression with independent outcomes and geographically weighted outcomes, and variable selection methods such as LASSO. We analyzed the raw data and MICE-imputed datasets.

    Our analysis suggests that governments might need to allocate healthcare resources heterogeneously, with a possible benefit in decentralizing healthcare. This could be a problem for developing countries, where the means are limited. As of May 2nd, 2020, countries with more economic equity among their citizens seemed less hit by COVID-19, possibly indicating the importance of having a minimal baseline assistance across the whole population of a country. The analysis of the demographic structure mildly indicated that the disease grows sub-exponentially in the first months of the diffusion. The reduced degree of mobility across countries, for example the degree to which tourism is constrained, had a positive effect in reducing the number of cases, deaths, and death rate per cases. However, there is an indication that a smart and alternating policy could lead to further containment of the disease. Furthermore, our analysis highlighted the benefit of informing the population for government measures to be more effective.

    Together, our results seem to indicate that blanket policies are sub-optimal and government measures related to healthcare and immigration have the potential to both help and damage the population, as, if not appropriately taken, they can lead to an increase or reduced decrease of COVID-19 cases, deaths, and deaths/cases rate.



    Conflict of Interest



    Clemens Drenowatz is an editorial board member for AIMS Public Health and was not involved in the editorial review or the decision to publish this article. All authors declare that there are no competing interests.

    [1] World Health OrganizationHealthy grow and development (2023). Available from: https://www.who.int/teams/maternal-newborn-child-adolescent-health-and-ageing/child-health/healthy-growth-and-development
    [2] Carreras M, Puig G, Sanchez-Perez I, et al. (2020) Morbidity and self-perception of health, two different approaches to health status. Gac Sanit 34: 601-607. https://doi.org/10.1016/j.gaceta.2019.04.005
    [3] Werneck AO, Silva DR, Agostinete RR, et al. (2018) Relationship of parental and adolescents' screen time to self-rated health: a structural equation modeling. Health Educ Behav 45: 764-771. https://doi.org/10.1177/1090198118757825
    [4] Zhang T, Lu G, Wu XY (2020) Associations between physical activity, sedentary behavior and self-rated health among the general population of children and adolescents: A systematic review and meta-analysis. BMC Public health 20: 1343. https://doi.org/10.1186/s12889-020-09447-1
    [5] Pascual SC, Casajús Mallén JA, González Gross M (2020) Adherence factors related to exercise prescriptions in healthcare settings: A review of the scientific literature. Res Q Exerc Sport 93: 16-25. https://doi.org/10.1080/02701367.2020.1788699
    [6] Burke S, Utley A, Belchamber C, McDowall L (2020) Physical activity in hospice care: A social ecological perspective to inform policy and practice. Res Q Exerc Sport 91: 500-513. https://doi.org/10.1080/02701367.2019.1687808
    [7] Da Silva AO, Diniz PRB, Santons M, et al. (2019) Health self-perception and its association with physical activity and nutritional status in adolescents. J Pediatr 95: 458-465. https://doi.org/10.1016/j.jped.2018.05.007
    [8] Moral Garcia JE, Agraso Lopez AD, Ramos Morcillo AJ, et al. (2020) The influence of physical activity, diet, weight status and substance abuse on students' self-perceived health. Int J Environ Res Public Health 17: 1387. https://doi.org/10.3390/ijerph17041387
    [9] Rittsteiger L, Hinz T, Oriwol D, et al. (2022) Changes of self-rated health status, overweight and physical activity during childhood and adolescence-the ratchet effect of high parental socioeconomic status. Front Sports Act Living 4: 781394. https://doi.org/10.3389/fspor.2022.781394
    [10] Brodersen K, Hammami N, Katapally TR (2023) Is excessive smartphone use associated with weight status and self-rated health among youth? A smart platform study. BMC Public Health 23: 234. https://do.org/10.1186/s12889-023-15037-8
    [11] Mintjens S, Menting MD, Daams JG, et al. (2018) Cardiorespiratory fitness in childhood and adolescence affects future cardiovascular risk factors: A systematic review of longitudinal studies. Sports Med 48: 2577-2605. https://doi.org/10.1007/s40279-018-0974-5
    [12] Tomkinson GR, Carver KD, Atkinson F, et al. (2018) European normative values for physical fitness in children and adolescents aged 9–17 years: Results from 2 779 165 Eurofit performances representing 30 countries. Br J Sports Med 52: 1445-14563. https://doi.org/10.1136/bjsports-2017-098253
    [13] García-Hermoso A, Ramírez-Campillo R, Izquierdo M (2019) Is muscular fitness associated with future health benefits in children and adolescents? A systematic review and meta-analysis of longitudinal studies. Sports Med 49: 1079-1094. https://doi.org/10.1007/s40279-019-01098-6
    [14] Ortega FB, Silventoinen K, Tynelius P, et al. (2012) Muscular strength in male adolescents and premature death: Cohort study of one million participants. BMJ 20: e7279. https://doi.org/10.1136/bmj.e7279
    [15] Högström G, Ohlsson H, Crump C, et al. (2019) Aerobic fitness in late adolescence and the risk of cancer and cancer-associated mortality in adulthood: A prospective nationwide study of 1.2 million Swedish men. Cancer Epidemiol 59: 58-63. https://doi.org/10.1016/j.canep.2019.01.012
    [16] Krause L, Mauz E (2018) Subjective, physical and mental health of children and adolescents in Thuringia: Representative results ealth Thuringia state module in KiGGS wave. Bundesgesundheitsblatt Gesundheitsforschung Gesundheitsschutz 61: 845-856. https://doi.org/10.1007/s00103-018-2753-8
    [17] Elissa K, Bratt EL, Axelsson ÅB, et al. (2020) Self-perceived health status and sense of coherence in children with type 1 diabetes in the west bank, Palestine. J Transcult Nurs 31: 153-161. https://doi.org/10.1177/1043659619854509
    [18] Husu P, Vaha-Ypya H, Vasankari T (2016) Objectively measured sedentary behavior and physical activity of Finnish 7- to 14-year-old children- association with perceived health status: A cross-sectional study. BMC Public Health 16: 338. https://doi.org/10.1186/s12889-016-3006-0
    [19] Padilla-Moledo C, Fernández-Santos JDR, Izquierdo-Gómez R, et al. (2020) Physical fitness and self-rated health in children and adolescents: Cross-sectional and longitudinal study. Int J Environ Res Public Health 17: 2413. https://doi.org/10.3390/ijerph17072413
    [20] Hanssen-Doose A, Kunina-Habenicht O, Oriwol D, et al. (2021) Predictive value of physical fitness on self-rated health: A longitudinal study. Scand J Me Sci Sports 31: 56-64. https://doi.org/10.1111/sms.13841
    [21] Kromeyer-Hauschild K, Wabitsch M, Kunze D, et al. (2001) Perzentile für den Body-Mass-Index für das Kindes- und Jugendalter unter Heranziehung verschiedener deutscher Stichproben. Monatsschrift Kinderheilkunde 149: 807-818. https://doi.org/10.1007/s001120170107
    [22] Bös K, Schlenker L, Büsch D, et al. (2009) Deutscher Motorik Test 6–18 (DMT 6–18). German motor performance test (DMT 6–18). Berlin, GER: Czwalina 6-18.
    [23] Utesch T, Dreiskamper D, Strauss B, et al. (2018) The development of the physical fitness construct across childhood. Scand J Med Sci Sports 28: 212-219. https://doi.org/10.1111/sms.12889
    [24] Ruedl G, Ewald P, Niedermeier M, et al. (2019) Long-term effect of migration background on the development of physical fitness among primary school children. Scand J Med Sci Sports 29: 124-131. https://doi.org/10.1111/sms.13316
    [25] UNESCO Institute for StatisticsInternational Standard Classification of Education (ISCED) (2012). Available from: http://uis.unesco.org/sites/default/files/documents/international-standard-classification-of-education-isced-2011-en.pdf
    [26] Ruedl G, Niedermeier M, Wimmer L, et al. (2021) Impact of parental education and physical activity on the long-term development of the physical fitness of primary school children: An observational study. Int J Environ Res Public Health 18: 8736. https://doi.org/10.3390/ijerph18168736
    [27] Ruedl G, Niedermeier M, Posch M, et al. (2022) Association of modifiable factors with the development of physical fitness of Austrian primary school children: A 4-year longitudinal study. J Sports Sciences 40: 920-927. https://doi.org/10.1080/02640414.2022.2038874
    [28] Burchatz A, Oriwol D, Kolb S, et al. (2021) Comparison of self-reported device-based, measured physical activity among children in Germany. BMC Public Health 21: 1081. https://doi.org/10.1186/s12889-021-11114-y
    [29] Cohen J (1988) Statistical Power Analysis for the Behavioral Sciences. New York, USA: Lawrence Erlbaum Associates 20-27.
    [30] Field A (2018) Discovering Statistics Using IBM SPSS Statistics. Thousand Oaks, CA: SAGE Publications Ltd 95-134.
    [31] Cocca A, Espino Verdugo F, Rodenas Cuenca L, et al. (2020) Effect of a game-based physical education program on physical fitness and mental health in elementary school children. Int J Environ Res Public Health 17: 4883. https://doi.org/10.3390/ijerph17134883
    [32] Cocca A, Carbajal Baca J, Hernandez Cruz G, et al. (2020) Does a multiple-sport intervention based on the TgfU model for physical education increase physical fitness in primary school children?. Int J Environ Res Public Health 17: 5532. https://doi.org/10.3390/ijerph17155532
    [33] Peralta M, Henriques-Neto D, Gouveia ER, et al. (2020) Promoting health-related cardiorespiratory fitness in physical education: A systematic review. PloS One 15: 0237019. https://doi.org/10.1371/journal.pone.0237019
    [34] Kantomaa MT, Tammelin T, Ebeling H, et al. (2015) High levels of physical activity and cardiorespiratory fitness are associated with good self-rated health in adolescents. J Phys Act Health 12: 266-272. https://doi.org/10.1123/jpah.2013-0062
    [35] Eriksen L, Curtis T, Gronbaek M, et al. (2013) The association between physical activity, cardiorespiratory fitness and self-rated health. Prev Med 57: 900-902. https://doi.org/10.1016/j.ypmed.2013.09.024
    [36] Herman KM, Hopman WM, Sabiston CM (2015) Physical activity, screen time and self-rated health and mental health in Canadian adolescents. Prev Med 73: 112-116. https://doi.org/10.1016/j.ypmed.2015.01.030
    [37] Breidablik HJ, Meland E, Lydersen S (2009) Self-rated health during adolescence: Stability and predictors of change (Young-HUNT study, Norway). Eur J Public Health 19: 73-78. https://doi.org/10.1093/eurpub/ckn111
    [38] Krause L, Lampert T (2015) Relation between overweight/obesity and self-rated health among adolescents in Germany. Do socio-economic status and type of school have an impact on that relation?. Int J Environ Res Public Health 12: 2262-2276. https://doi.org/10.3390/ijerph120202262
    [39] Mbogori T, Arthur TM (2019) Perception of body weight status is associated with the health and food intake behaviors of adolescents in the United States. Am J Lifestyle Med 15: 347-355. https://doi.org/10.1177/1559827619834507
    [40] Meland E, Breidablik HJ, Thuen F, et al. (2021) How body concerns, body mass, self-rated health and self-esteem are mutually impacted in early adolescence: A longitudinal cohort study. BMC Public Health 21: 496. https://doi.org/10.1186/s12889-021-10553-x
    [41] Matin N, Kelishadi R, Heshmat R, et al. (2017) Joint association of screen time and physical activity on self-rated health and life satisfaction in children and adolescents: The CASPIAN-IV study. Int Health 9: 58-68. https://doi.org/10.1093/inthealth/ihw044
    [42] Greier K, Drenowatz C, Ruedl G, et al. (2019) Association between daily TV time and fitness in 6- to 14-year old Austrian youth. Transl Pediatr 8: 371-377. https://doi.org/10.21037/tp.2019.03.03
    [43] Vermeiren AP, Willeboordse M, Oosterhoff M, et al. (2018) Socioeconomic multi-domain health inequalities in Dutch primary school children. Eur J Public Health 28: 610-616. https://doi.org/10.1093/eurpub/cky055
    [44] Okamoto S (2021) Parental socioeconomic status and adolescent health in Japan. Sci Rep 11: 12089. https://doi.org/10.1038/s41598-021-91715-0
    [45] Kerr J, Marshall S, Godbole S, et al. (2012) The relationship between outdoor activity and health in older adults using GPS. Int J Environ Res Public Health 9: 4615-4625. https://doi.org/10.3390/ijerph9124615
    [46] Coventry PA, Brown JE, Pervin J, et al. (2021) Nature-based outdoor activities for mental and physical health: Systematic review and meta-analysis. SSM Popul Health 16: 100934. https://doi.org/10.1016/j.ssmph.2021.100934
    [47] Sando OJ (2019) The outdoor environment and children's health: A multilevel approach. Int J Play 8: 39-52. https://doi.org/10.1080/21594937.2019.1580336
    [48] Pasanen TP, Tyrväinen L, Korpela KM (2014) The relationship between perceived health and physical activity indoors, outdoors in built environments, and outdoors in nature. Appl Psychol Health Well Being 6: 324-346. https://doi.org/10.1111/aphw.12031
    [49] Stodden DF, Goodway JD, Langendorfer SJ, et al. (2008) A developmental perspective on the role of motor skill competence in physical activity: An emergent relationship. Quest 60: 290-306. https://doi.org/10.1080/00336297.2008.10483582
    [50] (2018) World Health OrganizationAustria physical activity factsheet 2018. Geneva: WHO. Available from: https://www.euro.who.int/__data/assets/pdf_file/0009/382338/austria-eng.pdf.
    [51] Pouline T, Ludwig J, Hiemisch A, et al. (2019) Media use of mothers, media use of children, and parent–child interaction are related to behavioral difficulties and strengths of children. Int J Environ Res Public Health 16: 4651. https://doi.org/10.3390/ijerph16234651
  • This article has been cited by:

    1. Mohammad Javad Bordbari, Fuzhan Nasiri, Networked Microgrids: A Review on Configuration, Operation, and Control Strategies, 2024, 17, 1996-1073, 715, 10.3390/en17030715
    2. Mutiu Shola Bakare, Abubakar Abdulkarim, Aliyu Nuhu Shuaibu, Mundu Mustafa Muhamad, Energy management controllers: strategies, coordination, and applications, 2024, 7, 2520-8942, 10.1186/s42162-024-00357-9
    3. Kaka Sanaullah, Mingchao Xia, Arif Hussain, Kashif Hussain, Improving islanded distribution system stability with adaptive decision-making framework, 2024, 2194-5756, 10.1515/ijeeps-2024-0061
  • Reader Comments
  • © 2024 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(2570) PDF downloads(393) Cited by(0)

Figures and Tables

Figures(1)  /  Tables(4)

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog