key: cord-0752974-93tvzab8 authors: Ekumah, Bernard; Armah, Frederick Ato; Yawson, David Oscar; Quansah, Reginald; Nyieku, Florence Esi; Owusu, Samuel Asiedu; Odoi, Justice; Afitiri, Abdul-Rahaman title: Disparate on-site access to water, sanitation, and food storage heighten the risk of COVID-19 spread in Sub-Saharan Africa date: 2020-07-19 journal: Environ Res DOI: 10.1016/j.envres.2020.109936 sha: 25d8796cbcce43bd456e75baff735fc7439ac4ae doc_id: 752974 cord_uid: 93tvzab8 COVID-19 is an active pandemic that likely poses an existential threat to humanity. Frequent handwashing, social distancing, and partial or total lockdowns are among the suite of measures prescribed by the World Health Organization (WHO) and being implemented across the world to contain the pandemic. However, existing inequalities in access to certain basic necessities of life (water, sanitation facility, and food storage) create layered vulnerabilities to COVID-19 and can render the preventive measures ineffective or simply counterproductive. We hypothesized that individuals in households without any of the named basic necessities of life are more likely to violate the preventive (especially lockdown) measures and thereby increase the risk of infection or aid the spread of COVID-19. Based on nationally-representative data for 25 sub-Saharan African (SSA) countries, multivariate statistical and geospatial analyses were used to investigate whether, and to what extent, household family structure is associated with in-house access to basic needs which, in turn, could reflect on a higher risk of COVID-19 infection. The results indicate that approximately 46% of the sampled households in these countries (except South Africa) did not have in-house access to any of the three basic needs and about 8% had access to all the three basic needs. Five countries had less than 2% of their households with in-house access to all three basic needs. Ten countries had over 50% of their households with no in-house access to all the three basic needs. There is a social gradient in in-house access between the rich and the poor, urban and rural richest, male- and female-headed households, among others. We conclude that SSA governments would need to infuse innovative gender- and age-sensitive support services (such as water supply, portable sanitation) to augment the preventive measures prescribed by the WHO. Short-, medium- and long-term interventions within and across countries should necessarily address the upstream, midstream and downstream determinants of in-house access and the full spectrum of layers of inequalities including individual, interpersonal, institutional, and population levels. Since December 2019 the world has been afflicted by the coronavirus disease , and its causative virus severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). According to the World Health Organisation (WHO, 2020) , most people infected with the COVID-19 virus will experience mild to moderate respiratory illness and recover without requiring special treatment. However, older people, and those with underlying medical problems such as cardiovascular disease, diabetes, chronic respiratory disease, and cancer are more likely to develop serious illness (Emami et al., 2020; Jordan et al., 2020) . Notwithstanding this, the scale of infections and mortality make COVID-19 an existential threat to humankind. It has taken a heavy toll on all facets of the global economy and human life, including commerce and trade, health, livelihoods of populations, education, gender (Wenham et al., 2020) , mobility (more than 100 countries have imposed travel restrictions and others in partial/total lockdown), lifestyle choices, environment (lower atmosphere nitrogen dioxide levels have dropped sharply in France, Spain, Italy and China) and interpersonal relationships, among others. It is therefore unsurprising that a plethora of measures have been proffered to curb the spread and adverse outcomes of the disease on economies and human life. Obviously, policy responses have been uneven, often delayed, and there have been missteps. From a systems perspective, dealing with COVID-19 is a complex issue or wicked problem. It requires multi-sector, multi-disciplinary, interdisciplinary, trans-disciplinary and multistakeholder responses. This is reflected, in part, by the range of measures being proposed or implemented by different countries worldwide. The measures typically involve specific actions at various levels-individual, interpersonal, household, community, national, international, and global. Prominent actions that have been suggested so far include social/physical distancing, personal hygiene protocols, quarantine, isolation, environmental cleaning and ventilation, and partial/total lockdown. Most of these actions are community-based measures taken by planners, administrators, and employers to protect groups, employees and the population. While these measures are appropriate and laudable, they are often incognizant of existing intra-country and inter-country inequalities that could amplify vulnerabilities and potentially undermine wellintentioned interventions. Multidimensional inequality is a defining characteristic of all societies (Chakravarty & Lugo, 2019) . A key aspect of these inequalities relates to access to water and sanitation facility. According to the WHO, in 2017, 71% of the global population (5.3 billion people) used a safely managed drinking-water service -that is, one located on premises, available when needed, and free from contamination. Approximately 90% of the global population (6.8 billion people) used at least a basic service; that is an improved drinking-water source within a round trip of 30 minutes to collect water. About 785 million people lack even a basic drinking-water service, including 144 million people who are dependent on surface water (WHO/UNICEF, 2017). In least developed countries, 22% of health care facilities have no water service, 21% no sanitation service, and 22% no waste management service (WHO, 2017) . In 2015, 844 million people had no access to safe drinking water, and 2.3 billion people did not have ready access to basic sanitation services (WHO/UNICEF, 2017) . These issues are most severe in sub-Saharan Africa and central and southern Asia. In sub-Saharan Africa, people in urban areas are twice as likely as people in rural areas to have clean, safe water. Another way that we visualise the urban-rural divide is in sanitation. While rural areas often have less access to sanitation facilities, in sub-Saharan Africa the situation is very poor (Armah et al., 2018) . Only 24% of the rural population and 44% of the urban population have access to sanitation facilities (WHO/UNICEF, 2017). Access to safe water, sanitation and hygienic conditions play an essential role in protecting human health during all infectious disease outbreaks, including the current COVID-19 outbreak (UNICEF, 2020) . When we consider the fact that personal hygiene features prominently in most of the community-based measures to curb the spread of COVID-19 and is inextricably linked to access to improved water and sanitation, it is pertinent to underscore why sub-Saharan Africa may be greatly impacted by the disease even though the region has so far recorded the lowest number of COVID-19 cases. Pandemics such as COVID-19 require data-driven decisions. Hitherto research on the pandemic has focused almost exclusively on the clinical, virology and epidemiological dimensions (see Lipsitch et al., 2020) . This is understandable as COVID-19 is relatively new and active. Apart from the emerging insight that old persons and those with certain pre-existing health conditions are high risk group, there are gaps in our understanding of the human dimensions and behavioural responses of populations and groups to the implementation of community-based measures. Currently, most SSA countries are implementing partial or full lockdowns with varied enforcement mechanisms such as enactment of movement restrictions acts and deployment of low enforcement agencies to ensure compliance. However, it is not clear how existing inequalities in access to water, sanitation and food storage at the household level will attenuate, moderate or enhance the spread of infections in the population. We hypothesize that individuals in households with constrained access to water, sanitation, and food are more likely to violate lockdown, social distancing and personal hygiene measures and thereby stand a higher risk of COVID-19 infection. We use data on 25 countries in sub-Saharan Africa to investigate how vulnerability to COVID-19 is layered and heightened and how this might compromise the intended policy outcome of curbing the spread of the disease. In the context of this paper, vulnerability is defined as the degree to which a population, individual or group is unable to anticipate, cope with, resist and recover from the impacts of disease (including pandemics as . This study holds tremendous implications for theory (conceptualization of complexity and feedback relationships between policy intervention and human behaviour during pandemics), practice (intervention design) and policy (partial/total lockdowns). In particular, this study underscores the need for water, sanitation and hygiene services and products to be made available for confined households or areas with large vulnerable groups, exposed collective sites and public spaces. Demographic and Health Surveys (DHS) data for 25 SSA countries were used in this study. DHS provide data for comparative quantitative assessment of several important indicators in the areas of health, nutrition, population, and household energy across developing countries. Demographic and Health Surveys are nationally representative population-based surveys with large sample sizes. The determination of optimal sample size of DHS surveys is usually a trade-off between available budget and the desired survey precision (Aliaga & Ren, 2006) . A two-stage probabilistic samples drawn from an existing sample frame, generally the most recent census frame. In the first stage, primary sampling units (PSUs) are selected from a frame list with probability proportional to a size measure and in the second stage, a fixed number of households are selected from a list of households obtained from the selected PSUs. Probability sampling is provides unbiased estimation and enables evaluation of sampling errors (ICF International, 2012) . DHS data are open source and can be accessed on DHS website (www.dhsprogram.com). The inclusion criteria for a country in this study were as follows: (i) the country should be The dependent variable considered in this study was in-house access to basic needs. In this study, basic needs include water, sanitation facility and food storage. Emphasis was placed on the location of these three basic needs. When these basic needs are found in the house or on the compound, the propensity of members of a household flouting lockdown regulations and the risk of getting infected by COVID-19 could be low. In-house in the study context connotes inside a house or on the immediate compound of a house. It is worthy to note that the importance of these three basic needs with regards to adhering to lockdown regulations are not the same. For instance, the likelihood of a household without in-house sanitation facility violating the lockdown regulations is not the same as that of a household without refrigerator to store food. Other means of preservation can be used to store food for some few days but one cannot cope with living in a house without a sanitation facility for a day or two under total lockdown situation. In this regard, we gave the highest priority to sanitation facility followed by water (since some households have access to water sachets) and refrigerator was given the least priority. In-house access to sanitation facility was generated from the "type of toilet facility" variable in DHS datasets. In-house access to water was derived from the "location of source of water". In-house access to refrigerator was derived from "has refrigerator". The three variables were combined to generate the outcome variable called "in-house access to basic needs" with eight (8) categories, ranked and recoded in order of importance. In-house access to all the three basic needs was recoded "0", representing least propensity of flouting lockdown regulations and in-house access to none of the basic needs was recorded "7", which represents the group with the highest propensity of violating lockdown regulations. The dependent variable "in-house access to basic needs" was treated as an ordinal variable. Eight groups make up the dependent variable; "0"Access to all, "1" Sanitation facility and water only "2" Sanitation facility and refrigerator only, "3" Sanitation facility only "4" Water and refrigerator, "5" Water only "6" Refrigerator only, "7" No access. Both theoretically and practically, access to basic necessities within a household is a function of empowerment, which in turn depends on relationship structure. The explanatory variable was derived from the "relationship structure" variable and "sex of household member" in the DHS datasets. The two variables were combined to generate the key explanatory variable "household family structure". The groups were: "1" up to a single adult household, "2" two opposite sex adults household, "3" two female adults household, "4" two male adults household, "5" more than two related adults household and "6" more than two unrelated adults household. We controlled for compositional and contextual factors in assessing the relationship between household family structure and in-house access to basic needs. Compositional factor is related to the socio-demographic characteristics of an individual or group (Collins et al., 2017; Pol and Thomas, 2000) . It is categorized into biosocial and sociocultural factors. Biosocial factors are underlying biological or physical characteristics of individuals that are present at birth and not amenable to change (Pol and Thomas, 2000) . Socio-cultural factors refer to lifestyles, economy, beliefs, values and customs (see Armah et al., 2018) . The biosocial factors considered in this study included "sex of household member" (male or female), and "household age distribution". Household age distribution was derived from age and number of household members variables. The age variable was recoded into three age groups; young adult (18-40), middle-age adult (41-64) and old-age adult (65 and above). Household size which is a continuous variable in DHS datasets was converted to a categorical variable by grouping the "Number of household members"; small (less equal to 2), medium (3-5) and large (6 and above). The two recoded variables were then combined to produce household age distribution with 9 groups-"1" young adult in a small household, "2"young adult in a medium-size household, "3" young adult in a large-size household, "4"middle-aged adult in a small household, "5"middle-aged adult in a medium-size household, "6" middle-aged adult in a large-size household, "7" old-aged adult in a small household, "8" old-aged adult in a medium-size household and "9" old-aged adult in a large-size household. The socio-cultural variables included in the study were the wealth index (poorest, poorer, middle, richer and richest quintile) and highest educational level attained (no school/preschool, primary, secondary and higher). Factors that define the broader neighbourhood characteristics or location-specific opportunities in a region are referred to as contextual factors (Collins et al., 2017; Ross and Mirowsky, 2008) . The contextual factors included in the study were type of place of residence "urbanicity" (urban/rural) and country. Twenty four countries were included in the multivariate analysis. South Africa was dropped because preliminary analysis revealed that it is an outlier. The statistical analyses were carried out in STATA 13 MP (StataCorp, College Station, TX, USA). Descriptive analysis was used to examine the distribution of in-house access to basic needs in the 25 study countries. Univariate and multivariate statistical techniques were employed to assess associations between the in-house access to basic needs and the categorical explanatory variables. Spatial heterogeneity analysis of in-house access to basic needs was also carried out. With the exception of the descriptive, all other analyses (univariate, multivariate and spatial) were carried out on 24 countries excluding South Africa. South Africa was removed from the analyses because it had by far the highest rate of in-house access to all the basic needs (50%, others being below 30%), and also households were not geolocated. Environmental Systems Research Institute (ESRI) ArcGIS 10.3 was used to produce the maps. Pearson's Chi-square test was employed to determine the association between outcome variable (in-house access to basic needs) and the key predictor (household family structure) as well as the compositional and contextual factors. Cramer's V statistic was used to estimate the strength of the association between the variables. An association is considered strong when Cramer's V is equal or greater than 0.3 whereas an association is weak when Cramer's V is less than 0.3. Ordered logistic regression was employed to model the relationship between in-house access to basic needs and household family structure. Ordered logistic regression was used because, it takes into account the order or ranking in the dependent categorical variable. In-house access to basic needs was ranked with regards to the propensity to flout lockdown regulations due to lack of in-house access to basic needs. Covariates such as compositional (sex of household member, household age distribution, wealth index and highest educational level attained) and contextual (urbanicity and country) factors were adjusted for in the model. The effect size of the relationship was estimated using odds ratio (OR). An OR of 1 implies that the explanatory variable does not affect the odds of in-house access to basic needs; OR >1 means that the explanatory variable is related with higher odds of in-house access to basic needs; OR <1 means that the explanatory variable is related with lower odds of in-house access to basic needs. Robust estimates of variance was employed to correct any statistical outliers in the estimation of standard errors. Four models were run; household family structure (model 1), household family structure and biosocial factors (model 2) household family structure, biosocial factors and sociocultural factors (model 3), household family structure, biosocial, sociocultural and contextual factors (model 4). Recent datasets from DHS surveys have Global Positioning System (GPS) coordinates of the centroid point of enumeration clusters which enable spatial analysis and mapping. In order to protect the identity of respondents who participate in the surveys, the exact coordinates of the cluster locations are geomasked. Geomasking provides an approximate location of each households (Wilson et al., 2020) however, geolocations of households can still be used to assess spatial variations in DHS datasets (Kandala et al., 2011) . The spatial variations, with regards to in-house access to basic needs in the study countries, were mapped and assessed through categorical spatial interpolation in R using the Weighted K-Nearest Neighbor Classifier (kknn) (Hechenbichler & Schliep, 2004) . The DHS household data were joined to the cluster spatial point data using cluster number variable as the common field. A regular grid of the dissolve boundary of the study countries was created. The geometric point data which contain the in-house access to basic needs as a factor variable was converted to a regular data frame. The common group in the in-house access to basic needs were retained in the process to avoid "noise" in the map. The "kknn" classifier was then employed to interpolate the point data on the grid cells created from the dissolved boundary layer of the study countries. The "k" was set to 50 in order to examine the lower level disparities within countries. The "k" specifies the number of neighboring points that is used in classifying point dataset into different categories on the grid cells. The DHS datasets used in the study were collected through standardised procedures and questionnaires reviewed and approved by the International Classification of Functioning, Disability and Health (ICF) Institutional Review Board (IRB). Moreover, the survey protocols for countries are also subjected to the host country laws. Figure 2 shows that 7.5% of households in the 24 SSA countries (excluding South Africa) had in-house access to all the three basic needs and 46.4 had no in-house access to any of the basic needs. South Africa, which was not included in the analysis because it was considered an outlier, had 52.8% of its households having in-house access to all the basic needs and 5.8% had no inhouse access to any of the basic needs. Rural poorest group recorded the highest proportion (79.4%) of households without in-house access to all the three basic needs even though there was no substantial difference compared to that of urban poorest (76.3%). Urban richest group had the highest proportion (74.7%) of households with in-house access to all the basic needs. No access and sanitation only were dominant in poor categories irrespective of the type of residence whereas the dominance in the richer and richest categories depended on the type of residence. Africa with over 50% of its households having in-house access to all the basic needs, Zimbabwe was the only country which had in-house access to all basic needs above 20%. Ten countries had over 50% of their households with no in-house access to all the three basic needs. Table 1 . Percentage in-house access to water, sanitation facility and refrigerator in 25 SSA countries. The Pearson chi-square and Cramer's V statistics results of the association between in-house access to basic needs and household family structure and the compositional and contextual variables are provided in the Supplementary material (Table S1 ). The results rejected the null hypothesis that in-house access to basic needs was independent of household family structure. The results indicated a weak association (P<0.0001, Cramer's V = 0.076) between in-house access to basic needs and household family structure. Regarding the compositional and contextual factors, the results showed that there was a strong association between the outcome In the multivariate analyses, four models: household family structure (model 1), biosocial (model 2), sociocultural (model 3), and contextual (model 4), were built to assess their relationship with in-house access to basic needs. Table 2 shows the proportional odds ratios, robust standard errors, probability values and confidence intervals associated with the key predictor (household family structure) as well as the compositional and contextual factors. Model 1 indicates that households with two opposite sex adults were 28% more likely to belong to a group with worse outcomes in terms of in-house access to basic needs compared to households with up to a single adult. Households with two female adults and more than two unrelated adults were 35% and 65% less likely respectively to have worse outcomes compared to households with up to a single adult. In model 2 when biosocial factors (sex of household member and household age distribution) were controlled, the relationship between household family structure and in-house access to basic needs changed in magnitude. Households with two opposite sex adults became statistically insignificant indicating complete mediation of the relationship. Households with two male adults and more than two related adults which were not statistically significant in model 1 became significant in model 2. They were 21% and 35% respectively less likely to belong to a group with worse outcomes in terms of in-house access to basic needs compared to households with up to a single adult. The odds of households with more than two unrelated adults increased by 30% in model 2. Females were 7% less likely to belong to a group with worse outcomes in terms of in-house access to basic needs compared to males. Regarding household age distribution, all groups were more likely to belong to a group with higher level of worse outcomes in terms of inhouse access to basic needs compared to the reference group, young adults in small households. In the sociocultural model, the direction of the relationship between in-house access to basic needs and household family structure did not change compared to the model 2 but, there were changes in the magnitudes of some sub-groups. The proportional odds ratios for households with two female adults, two male adults, with more than two related adults and more than two unrelated adults decreased by 39%, 16%, 30% and 180%, respectively. Besides, the level of significance for households with two female adults reduced indicating partial mediation of the relationship. The direction of the relationship between sex of household member and in-house access to basic needs did not change but the magnitude increased by 17%. Regarding the effect of sociocultural factors on the relationship between in-house access to basic needs and household age distribution, three main things were observed. There was complete mediation of the relationship between middle-aged adults in medium-sized households and in-house access basic needs. Middle-aged adults, old adults in small households, old adults in medium-sized households and old adults in large households became 19%, 26%, 9%, and 18% less likely to belong to a group with higher order of worse outcomes in terms of in-house access to basic needs compared to young adults in small households. In addition, there was decrease in magnitude for young adults in medium-sized households (8%), young adults (46%) and middle-aged adults (49%) in large households when sociocultural factors were controlled in model 3. Individuals in poorer, middle, richer and richest wealth quintile were 55%, 73%, 88%, and 97% were less likely to belong to a group with worse outcomes in terms of in-house access to basic needs compared to poorest households. Individuals educated to the primary level (19%), secondary level (45%) and higher level (71%) were less likely to belong to a group with higher order of worse outcomes in terms of in-house access to basic needs compared to individuals with no education or educated to the preschool level. When contextual factors (urbanicity and country) were accounted for in model 4, suppression, partial and complete mediation were observed in the relationship between household family structure and the outcome variable. Households with two opposite sex adults group was not significant in the sociocultural model (model 3) but became marginally less likely to belong to a group with higher level of worse outcomes with regards to in-house access to basic needs compared to households with a single adult in model 4. There was complete mediation effect on households with two female adults which was statistically significant in model 3 became insignificant in model 4. There was also a partial mediation effect on household with two male adults. The proportional odds ratios for households with more than two related adults and households with more than two unrelated adults reduced by 10% and 19%, respectively. The direction of the relationship between sex of household member and in-house access to basic needs did not change but the magnitude reduced by 14%. The relationship between age distribution of household and the dependent variable was mediated, nonetheless, the direction of the association did not change. The level of significance for young adults in medium-sized households and young adults in large households reduced (partial mediation). Parameter estimates for middle-aged adults in large households became statistically insignificant indicating complete mediation by the contextual factors. With regards to wealth index, the direction of the relationship with in-house access to basic needs remained the same and there was no substantial change in magnitude of the proportional odds ratios. Contextual factors did not affect the direction of the effect of highest educational attainment but there were substantial changes in magnitude. The magnitude of the relationship for individuals educated to the primary, secondary and higher level decreased by 18%, 37% and 43%, respectively. Regarding urbanicity, households in urban areas were 47% less likely to have worse outcomes in terms of in-house access to basic needs compared to households in rural areas. Households in Angola, Ghana, Senegal, Tanzania and Zimbabwe were less likely to belong to a group with higher order of worse outcomes in terms of in-house access to basic needs compared to households in Rwanda. With the exception of Kenya and Zambia, which were not statistically significant, households in the remaining 15 countries were more likely to belong to a group with higher level of worse outcomes with regards to in-house access to basic needs compared to households in Rwanda.