Policy Research Working Paper 10328 Estimating Spatially Disaggregated Probability of Severe Covid-19 and the Impact of Handwashing Interventions The Case of Zimbabwe George Joseph Sveta Milusheva Hugh Sturrock Tonderai Mapoko Sophie Charlotte Amy Ayling Yi Rong Hoo Water Global Practice February 2023 Policy Research Working Paper 10328 Abstract The severity of COVID-19 disease varies substantially risk at the district level. Results show that the current risk between individuals, with some infections being asymp- of severe disease is heterogeneous across the country, due tomatic while others are fatal. Several risk factors have been to differences in individual characteristics and household identified that affect the progression of SARS-CoV-2 to conditions. This study demonstrates how household level severe COVID-19. They include age, smoking and pres- improved access to handwashing could lead to reductions ence of underlying comorbidities such as respiratory illness, in the risk of severe COVID-19 of up to 16% from the HIV, anemia and obesity. Given that respiratory illness is estimated current levels across all districts. Taken alongside one such comorbidity and is affected by hand hygiene, it the likely impact on transmission of SARS-CoV-2 itself, as is plausible that improving access to handwashing could well as countless other pathogens, this result adds further lower the risk of severe COVID-19 among a population. support for the expansion of access to handwashing across In this paper, the authors estimate the potential impact of the country. It also highlights the spatial differences in risk improved access to handwashing on the risk of respiratory of severe COVID-19, and thus the opportunity for better illness and its knock-on impact on the risk of developing planning to focus limited resources in high risk areas in severe COVID-19 disease across Zimbabwe. They use a geo- order to potentially reduce the number of severe cases. spatial model that allows us to estimate differential clinical This paper is a product of the Water Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at gjoseph@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Estimating spatially disaggregated probability of severe COVID-19 and the impact of handwashing interventions: The case of Zimbabwe George Joseph1, Sveta Milusheva, Hugh Sturrock, Tonderai Mapoko, Sophie Charlotte Amy Ayling, Yi Rong Hoo JEL: I12;I18;O55 WASH; Handwashing; Spatial Modelling: Public Health, Coronavirus (COVID 19) 1 George Joseph (corresponding author), Water Global Practice, World Bank, Washington DC 20433, email: gjoseph@worldbank.org. Sveta Milusheva, Development Impact Evaluation, World Bank , Washington DC 20433, smilusheva@worldbank.org, Hugh Sturrock, Co-Founder and Chief Science Officer at Locational.UK, hugh@locational.io, Tonderai Mapako, Chair, Biomedical Research and Training Institute, Institutional Review Board, Harare, Zimbabwe, mapakot2008@gmail.com, Sophie Charlotte Amy Ayling, Center for Advanced Spatial Analysis, University College, London UK, sayling@worldbank.org,and Yi Rong Hoo, World Bank, Washington DC 20433, yhoo@worldbank.org. We would like to acknowledge the support of the Global Water Security and Sanitation Partnership (GWSP) of the World Bank and the ieConnect for Impact Program funded with UK aid from the UK government in completing this work. Also this analysis formed part of a broader collaboration of the World Bank with the Zimbabwe National Modeling Consortium. In particular the Biomedical Research and Training Institute - Institutional Review Board (BRTI-IRB) and the COVID-19 Modeling and Spatial analysis sub-group (modeling consortium); Mr. Tendayi Kureya, Executive Secretary of the Medical Research Council of Zimbabwe (MRC-Z); Dr Shungu Munyati, Director General of BRTI; Professor in Demography and Behavioral Science Simon Gregson and Dr. Mike Pickles, Research Fellow of Imperial College London (ICL). From the World Bank we would like to thank Chenjerai Sismayi from the Health, Nutrition and Population Global Practice of the World Bank as well as the Country Management Unit of the World Bank in Zimbabwe. The findings, interpretations and conclusions expressed in this paper do not necessarily reflect the views of the World Bank, the Executive Directors of the World Bank or the governments whom they represent. The World Bank does not guarantee the accuracy of the data included in this work. 2 Estimating spatially disaggregated probability of severe COVID-19 and the impact of handwashing interventions: The case of Zimbabwe Introduction The Coronavirus pandemic (SARS-CoV-2) has affected 220 countries around the world, with over 5.4 million recorded deaths as of December 2021 (JHU, 2021). In Zimbabwe, there have been just under 5,000 recorded deaths following multiple waves. (JHU, 2022). The severity of COVID-19 disease varies substantially between individuals, with some infections being asymptomatic while others are fatal. Research studies have identified a number of risk factors for severe disease which include older age, smoking and underlying comorbidities including hypertension, diabetes, cardiovascular disease and chronic respiratory disease (Elmore et al., 2020; Q. Ma et al., 2020; Ou et al., 2020; Wu and McGoogan, 2020; Hu and Wang, 2021) . Handwashing is widely understood to modify risk of respiratory illness as hands have been known to harbor viral respiratory pathogens (Gwaltney, 1978; Ansari et al., 1991; Weber and Stilianakis, 2008). This happens when hands come into contact with contaminated fomites (environmental surfaces) or through a person-to-person contact with a disease carrier who sheds the respiratory pathogens through their nose or mouth (Hendley, Wenzel and Gwaltney, 1973; Brady, Evans and Cuartas, 1990). Transmission occurs following contact between hands and nasal mucosa or mouth of the new host. There are at least two channels through which handwashing can theoretically have an impact on reducing the number of severe COVID-19 cases in the population. The more direct pathway is by reducing transmission of the SARS-CoV-2 virus itself (Figure 1 (1)). Health authorities have established that the main transmission routes for SARS-CoV-2 are via either respiratory droplets, inhaled directly by a susceptible individual or through bodily contact, or via an infected surface or fomite (such as door handles or other frequently touched surfaces) (van Doremalen et al., 2020; WHO, 2020). For this reason, handwashing will likely have an impact on transmission of SARS-CoV-2 virus. Hand-washing has been encouraged during the pandemic to reduce the likelihood of spreading the infection from one individual to another (WHO, 2020). There have been more recent efforts to quantify (Talic, 2021) and determine the extent to which both hand-washing facility access (Zvobgo and Do, 2020) and handwashing behaviour (Meda et al., 2020, 2020; Yang, 2020; Hirose et al., 2021) could impact the transmission of SARS- CoV-2. However, epidemiological modeling that sought to incorporate handwashing scenarios in predicting disease outcomes in the population have had to make assumptions to quantify the impact of handwashing on transmission reduction (Winskill et al., 2020). The second, more indirect pathway for handwashing in reducing prevalence of severe COVID-19 is via its impact on reducing other respiratory infections. Reducing transmission and incidence of other respiratory illnesses, would reduce prevalence of a risk factor for developing severe COVID- 19 (Udwadia et al., 2020, Ma et al, Gao et al., 2021; Hu & Wang, 2021; C. Ma et al., 2020; Ou et al., 2020). A large body of empirical studies in household and community settings have demonstrated the effectiveness of hand washing in reducing other respiratory illnesses around the 3 world, be that through campaigns or through the presence of the facility itself. Washing hands with soap (both plain and antibacterial soap) is a proven mechanism to eliminate bacteria and respiratory viruses (Luby et al., 2001; Gibson et al., 2002; Burton et al., 2011). A systematic review of eight studies from a pool of 410 articles found that hand washing lowered the risk of respiratory infection, with risk reductions ranging from 6% to 44%, but noted that a greater number of rigorous studies are urgently needed (Rabie and Curtis, 2006). Another meta-analysis suggested that regular hand hygiene provided a significant protective effect (OR = 0.62;95% CI 0.52–0.73; I2= 0%) against 2009 pandemic influenza infection (Saunders-Hastings et al., 2017). A randomized controlled trial among squatter settlements in Karachi, Pakistan, found that children younger than 5 years old living in households that were assigned to antibacterial soap had roughly a 50 percent lower incidence of pneumonia than controls (Luby et al., 2005). Similarly, another randomized control trial across 60 elementary schools in Egypt found that a hand hygiene intervention campaign significantly reduced the incidence of laboratory-confirmed influenza by 50 percent and subsequently absenteeism among the students caused by the illnesses (Talaat et al., 2011). In Bangladesh, a study over 60,000 low income households, found that across the sample, respiratory illness was lower among people who had soap and water present in the hand washing station than among those who did not (Najnin et al., 2019). A two-stage cluster RCT with 60 villages in Kenya also showed that children in schools where hand washing and drinking water stations were made available were consistently less likely to have an acute respiratory infection over the following 2-10 months than in schools without (2% versus 3%, Estimated Difference in Medians -2%, 90% CI-3% to -1% in rounds 1-17) (Hoekstra et al., 2012). Taken together, it is therefore plausible that handwashing could impact the number of hospitalizations and deaths from COVID via two pathways: 1) by reducing transmission and incidence of COVID-19 and 2) by reducing the incidence of underlying respiratory illness in a population, thereby reducing susceptibility to severe COVID-19. If a host who becomes infected with SARS-CoV-2 already has a respiratory infection the chances of that developing to severe COVID-19 disease are increased. What we focus on in the analysis presented in this paper is therefore the second, more indirect pathway described above, to impact on prevalence of severe COVID-19 cases. This study contributes to the literature in three main ways. Firstly, it contributes to the growing body of spatially disaggregated approaches to COVID-19, but focusing on clinical risk data. Spatially disaggregated models thus far have made use of mobility data to examine transmission dynamics (Mu, Yeh and Zhang, 2021; Zachreson et al., 2021) and examined the impact of government policies(Guaitoli and Pancrazi, 2021), through lockdowns, or compared population characteristics in their response to government measures(Macharia, Joseph and Okiro, 2020; Gauvin et al., 2021). However, to the best of our knowledge, this study is one of the few, particularly in the sub-Saharan African context, that have used available spatially disaggregated data on clinical risk factors to map underlying risk of severe COVID-19 in a population. Understanding spatial variation in risk using spatial modeling has proved useful for a number of diseases including malaria (Weiss et al., 2019), schistosomiasis (Sturrock et al., 2013) , and HIV ((Dwyer-Lindgren et al., 2019). In turn, clinical risk for severe COVID-19 is likely to differ across a country, so using a spatial approach has potential to be strategically useful for targeting a health policy response. Its application to severe COVID-19 risk factors is therefore applied here. 4 Secondly, though it is widely accepted that hand hygiene is important for reducing the spread of respiratory illnesses, and that it is important for reducing transmission of SARS-CoV-2, there have been limited efforts to map out the pathways for how handwashing can impact upon a reduction in severe COVID-19 cases. This is despite the use of hand hygiene scenarios that reduce transmission being incorporated into epidemiological models (Winskill et al., 2020). In this study we seek to build on existing literature that sought to quantify the impact of handwashing on reducing the transmission of respiratory pathogens, and examine how that could have knock-on implications for the extent of severe COVID-19 prevalence. Thirdly, this paper relies on available data to map underlying risk factors of severe COVID-19 disease and potential impact of handwashing interventions in a way that can be used for guiding policy and planning interventions. This approach is particularly useful in data scarce environments, common to developing countries. Data and Methods Ethics Statement This study was approved by the Medical Research Council of Zimbabwe. The approval number is MRCZ/E/302 and it was assessed as exempt. All data obtained and herein reported were anonymized prior to analysis. Study Area In this paper we focus on Zimbabwe as a case study for the methodology being presented. Though there is an emerging collection of COVID-19 related studies within the African continent (Guleid et al., 2021), the majority of studies thus far have been concentrated in Asia, North America, and Europe, while Africa focused studies are in the minority (Wang and Tian, 2021). With lower rates of vaccination on average than in other parts of the world, vulnerability to severe disease remains an important concern in countries like Zimbabwe. Spatial analysis methods of this nature can be taken as a useful model for similar contexts where vaccine prioritization may be important in public health planning. It is also a data scarce environment when compared to some of the countries studied more intensively during the pandemic, and thus can provide a useful example of how similar contexts can be studied under such circumstances. Data The main data sets used in this analysis are: the Demographic and Health Survey (DHS) from 2015 in Zimbabwe; population density, night-time light intensity for 2015 and distance to nearest Open Street Map (OSM) road obtained via WorldPop (www.worldpop.org); and poverty Small Area Estimations (SMEs) from the World Bank at ward level (Masaki et al., 2020). The 2015 DHS was used to inform the creation of a hand-washing risk index drawing from data on access to hand-washing facilities with soap and water at the household level. Respiratory illness data was taken from the same DHS survey but only available for children under five years of age. Studies have indicated that children can often be the source of respiratory illness transmission to 5 adults within the same household (Longini et al., 1982; MacIntyre et al., 2012), and thus we have used this indicator as a proxy for respiratory illness prevalence in adults at the household level. Anemia, overweight prevalence and HIV prevalence were also taken from this source. There are 400 clusters in the DHS distributed across the ten provinces (60 districts) of the country. Though the data is originally collected to be representative at the province level, the application of spatial modeling to such data allow for higher resolution predictions of outcomes (Diggle et al., 2013; DHS Spatial Interpolation Working Group, 2014) such as malaria prevalence or vaccine coverage e.g. (Takahashi et al., 2017; Utazi et al., 2020, 2020)2. Spatial modeling approaches leveraged are described in more detail in the methods section. The odds ratios for risk of severe disease from different clinical risk factors were taken from a meta-analysis and systematic review study of incidence and clinical characteristics and prognostic factors of patients with COVID-19 across 53,000 patients (C. Ma et al., 2020) and the age specific probabilities of hospitalization are taken from Estimates of the severity of coronavirus disease 2019: a model-based analysis (Verity et al., 2020). Methods The methodology outlined here includes three parts. First we calculate the prevalence of risk factors and comorbidities for severe COVID-19 at the district level. This includes the prevalence of respiratory illness modeled as a function of handwashing risk. Secondly, we use these prevalence estimates to calculate the probability of developing severe COVID-19 disease. Thirdly, we modify the level of handwashing risk according to different counterfactual scenarios in order to see the impact it has on reducing prevalence of respiratory illness and severe COVID-19 risk across the country. 1. Predicting prevalence of risk factors and comorbidities 1.1 Handwashing risk Using the three variables related to handwashing in the DHS data, we assign each household a hand washing ‘risk’ score. For households with a hand washing facility with water and soap, this was considered equivalent to a hand washing risk of 0. With water but no soap, there was a risk of 0.5, while with no water or no facility (even if soap was present) the risk was set to 1. The mean household level observation was then calculated for each cluster. Cluster level georeferenced handwashing risk scores were used to predict handwashing risk across the whole country and generate a continuous map using a beta Generalized Additive Model (GAM) which includes a bivariate smooth on latitude and longitude. Population density, night-time light intensity for 2015 and distance to nearest Open Street Map (OSM) road obtained via WorldPop 2 Though DHS clusters are intentionally displaced by 0-2km in urban areas and 0-5 km in rural clusters, with a further, randomly-selected 1% (every 100th) of rural clusters displaced by up to 10km (Burgert, 2013), this is taken into account in the interpolation process. The application of interpolation techniques specifically to DHS data is described in the referenced publication (DHS Spatial Interpolation Working Group, 2014) 6 (www.worldpop.org) were all resampled to 1km resolution and included as covariates with thin- plate splines used to model non-linear effects. These covariates were chosen due to their plausible association with poverty and access to handwashing facilities. While other covariates could be used, here the goal of the spatial modeling was to produce predictions of handwashing risk as opposed to examining factors associated with handwashing risk in detail. We also included an additional penalty term in the model to allow covariates that were not related to the outcome to essentially be removed from the model (Marra and Wood, 2011). This model was used to predict handwashing risk at 1km resolution. District level handwashing risk was then estimated by calculating the population weighted mean for all cells within a district. 1.2 Respiratory illness To generate continuous maps of respiratory illness, we calculated the prevalence of respiratory illness symptoms at any time in the two weeks preceding the survey at each DHS cluster, available for individuals aged under 5 years of age (Croft, Marshall and Allen, 2020). The prevalence of respiratory illness symptoms was then modeled using a binomial GAM with handwashing risk, distance to road, population density and poverty as covariates in addition to a spatial bivariate smooth. The ward level poverty estimates were rasterized to be on the same 1km resolution grid as all other covariates and included as an additional covariate. This model was then used to predict prevalence of respiratory illness at 1km resolution. District level prevalence of respiratory illness was then estimated by calculating the population weighted mean for all cells within a district. 1.3 Comorbidities and factors for the probability of severe COVID-19 In addition to respiratory illness, we generated estimates of the prevalence of several additional comorbidities and probability factors for severe COVID-19 at the district level. Individual probability factors that are included at this stage are based on existing literature and presently are HIV prevalence (without Viral Load Suppression), anemia prevalence, and obesity prevalence, in addition to respiratory illness. For each of these, prevalence was calculated at the cluster level and the same GAM model was fit to each outcome to generate predicted values at the 1km cell level. Once these gridded surfaces are created at a 1km resolution, estimates are reached at the district level by weighting, using a raster file of population count at the pixel level. We then calculate the probability of having any of these comorbidities, defined as 1 minus the probability of having none of these comorbidities: = 1 − (1 − (pri ⋅ pHIV ⋅ pa ⋅ po)) where pco = proportion with a comorbidity, pri = proportion with respiratory illness, pHIV = proportion with HIV (without viral load suppression), pa = proportion with anaemia and po = proportion with obesity. . We apply the same geospatial methods to calculate the proportion of the population that is a smoker based on smoking prevalence within DHS clusters. We additionally calculate the proportion of the population that is over age 50 from WorldPop data. 7 2. Predicting probability of severe COVID-19 To calculate the probability of severe COVID-19, we apply the odds ratios estimated in Ma et al to the proportion with the corresponding probability factor/comorbidity at the district level. () = (2.6) + (1.7) + (2.6)50 + intercept where psc = probability of severe COVID-19, pco = proportion with a comorbidity, psm = proportion that smoke, p50 = proportion over 50. In order to apply this equation, we need the intercept value specific to Zimbabwe. This was set so that the mean district level probability of severe COVID was equal to that expected on the basis of the age-breakdown of the population. To estimate this expected probability of severe COVID, we applied the Imperial College London (ICL) clinical study age-specific estimates of probability of severe disease to the age breakdown data from the Zimbabwe census from 2012. This generated a mean probability of severe disease across the population in Zimbabwe of 2.9%. After testing a range of different intercepts, an intercept of -4.2 was chosen as this produced a mean district level probability of severe disease of 2.9%. 3. Generating counterfactual scenarios We produce two counterfactual scenarios for probability of severe COVID-19. These are based on predictions of prevalence of respiratory illness under different levels of handwashing risk, which affect the probability of having a comorbidity and in turn the final probability score. The first scenario uses a handwashing risk of 0.25 throughout the country, equivalent to the lowest level of handwashing risk (i.e. the highest level of handwashing access) currently observed at the district level. This is a realistic scenario where it is assumed that all the districts can improve their handwashing access levels at least to the level of the best performing district. The second scenario, which shows an ideal case, uses a handwashing risk of 0, equivalent to assuming that every household has access to a hand washing facility, soap and water. Results Handwashing risk estimates Figure 1 shows the cluster level handwashing risk and the district level estimates. There is considerable variation across Zimbabwe, with the lowest cluster level handwashing risk of 0.036 and highest of 0.923. At the district level, hand washing risk varied from 0.25 in Bulawayo to 0.69 in Mberengwa. 8 Figure 2. DHS Cluster level handwashing risk across Zimbabwe in 2015 (left) and estimated district level handwashing risk (right). Respiratory illness probability estimates Figure 2 shows the relationships between respiratory illness and each of the covariates explored, including the spatial effect. As expected, handwashing risk showed a general positive relationship with respiratory illness as did population density. There appeared to be a negative, but non-linear, relationship with distance to road. Poverty did not show an association with respiratory illness. In addition to the covariate effects, there appeared to be residual spatial variation in respiratory illness across the country not explained by those covariates included in the model. Figure 3. The estimated effects of the covariates included in the respiratory illness model. 9 Predicted prevalence of respiratory illness varied across the country, with Makonde, Gokwe North and Kariba districts all having a predicted prevalence of > 0.54 and Insiza, Beitbridge and Matobo having a prevalence of < 0.26 (Figure 4A). Figure 4. Predicted prevalence of respiratory illness at the district level in Zimbabwe Estimates of probability of severe COVID-19 Estimates of the probability of severe disease among symptomatic cases are displayed in Figure 5. Under current conditions, probability of severe disease among symptomatic cases at the district level is predicted to vary between 0.026 – 0.033, with Makonde, Mutasa and Marondera districts predicted to have the highest levels of probability (Figure 5). 10 Figure 5. Predicted prevalence of respiratory illness at the district level in Zimbabwe 11 3. Counterfactual Scenarios Under the first counterfactual scenario of reducing handwashing risk to the lowest observed levels at district level, mean prevalence of respiratory illness is predicted to drop from 0.39 to 0.32 (Figure B). Under the second scenario of perfect handwashing access, it drops to 0.15. (Figure C). Figure 6. Predicted prevalence of respiratory illness at the district level in Zimbabwe - current, and under improved handwashing scenarios This decrease in the prevalence of respiratory illness leads to a drop in the probability of severe disease (Figure 7B and C). This drop is marginal if handwashing risk is set to 0.25 throughout every district (Figure 7B), with the mean district level probability of severe disease dropping from 0.029 to 0.028. This drop in risk is more substantial in a scenario of perfect access to handwashing facilities (i.e. a handwashing risk of 0), with the mean district level probability of severe disease dropping by around 16% to 0.025 (Figure 7C). Figure 7. Predicted probability of severe COVID-19 among symptomatic cases at the district level in Zimbabwe under different hand washing risk scenarios. 12 Discussion The severity of COVID-19 disease caused by SARS-CoV-2 is highly variable with some individuals and populations being more vulnerable to complications than others. Using information on known factors for probability of severe disease, and mapping it to available data on such factors from the DHS, here we estimate and map this variation in probability of severe disease across Zimbabwe. Results show that the probability of severe disease varies substantially with some districts having a probability of severe disease nearly 30% higher than others. Furthermore, we show that there is a relationship between handwashing and respiratory illness, a potential risk factor for severe disease, and that improving access to handwashing across the country could reduce the probability of severe disease. From a policy perspective, this study highlights a potential added benefit of improvements to WASH facilities in addition to the already known numerous other benefits improved WASH brings. Hand washing is also an important intervention for gastro-intestinal and worm infections (Ejemot et al., 2015; Bieri et al., 2013; Aiello et al., 2008; Ejemot et al., 2008; Rabie & Curtis, 2006), in addition to the relationships highlighted here for reducing transmission of respiratory infections. Taken together, this study adds another perspective on why handwashing access, particularly in developing countries, can continue to be considered as an important tool in the fight against COVID-19. Other studies have also shown it to be highly cost-effective in prevention of the spread of the virus (Bagepally et al., 2021), an important consideration in light of the considerable economic and societal costs of the pandemic, and associated restrictions (World Bank, 2020; Vagliasindi, 2021). Indeed, investments in WASH are likely to lead to substantial economic returns in the absence of COVID-19. The pandemic should make this investment all the more urgent. While the link between handwashing and probability of severe COVID-19 is plausible, there are a number of important limitations and assumptions made as part of this study that deserve discussion. Firstly, we assumed that the prevalence of respiratory illness symptoms in children is representative of the prevalence of illness in the wider population. In reality, rates of symptoms and disease may vary between age-groups. Furthermore, many types of respiratory illness may not be severe enough to constitute being a risk factor for severe COVID-19. Second, our results rely on the strong relationship between handwashing and respiratory illness risk based on previous studies, the causality of which has been indicated in the literature reviewed in the introduction (Hendley, Wenzel and Gwaltney, 1973; Ansari et al., 1991; Luby et al., 2005). However, this model does not take into account additional factors affecting respiratory illness risk e.g. indoor pollution (Smith et al., 2000; Mocumbi et al., 2019). Further to this point, out of necessity, these analyses are conducted at district level using odds ratios obtained from analysis at the individual level which is not representative of the full population. While we don’t expect this to impact the main conclusions of the study, it raises the possibility of the atomistic fallacy. Third, the comorbidities and risk factors for severe disease included in this study are not exhaustive and are limited to those for which data are available. Other factors that may pose a risk for severe disease, include TB, malaria or hypertension (or yet to be discovered others). Further to this point, we only consider prevalence of any comorbidity as a risk factor. In reality, having more than one 13 comorbidity is likely to increase the probability of severe disease. To incorporate this into this study would require estimates of odds ratios for each risk factor and their interactions as well as an understanding of the correlation between risk factors as these are likely to not be independent. For example, obesity has been linked to hypertension (Aneja, 2004). What the model seeks to do is provide a framework into which more data on risk factors can be added as and when it becomes available. Fourth, we make the assumption that the impact of improvements in access to handwashing on prevalence of respiratory illness occurs over a timescale relevant to the pandemic. While previous research from Kenya suggests that the impact can be observed 2-9 months after intervention (Hoekstra et al., 2012), if this timescale is longer in Zimbabwe, or if the improvements themselves take a long time to be implemented, we may not see the desired effects. One potential way to maximize impact would be to prioritize districts for improved handwashing according to their current handwashing risk. Intervention impacts could be maximized by targeting districts or communities with highest handwashing risk first. Outputs from ongoing spatially discrete agent- based modeling, informed by human movement data, could also be used to identify those districts most vulnerable to transmission. Finally, this paper has only focused on underlying risk factors in the population and has not considered risk of transmission due to e.g. population mobility. The outputs of this work will likely therefore be used as part of a larger agent based modeling exercise to identify those districts with highest probability of severe disease when transmission is also taken into account, in a separate, forthcoming paper. In conclusion, this study shows the spatial disparity in severe COVID-19 risk across Zimbabwe. Secondly, it shows that there is a relationship between access to handwashing facilities and respiratory illness, a known risk factor for severe COVID-19. In addition to any direct effects on transmission of SARS-CoV-2, improving access to handwashing may therefore reduce the probability of severe disease among the population which could in turn reduce the number of individuals requiring hospitalization, critical care or dying from the disease. References Aneja, A. (2004) ‘Hypertension and Obesity’, Recent Progress in Hormone Research, 59(1), pp. 169–205. doi:10.1210/rp.59.1.169. Ansari, S.A. et al. (1991) ‘Potential role of hands in the spread of respiratory viral infections: studies with human parainfluenza virus 3 and rhinovirus 14.’, Journal of Clinical Microbiology, 29(10), pp. 2115–2119. doi:10.1128/JCM.29.10.2115-2119.1991. Bagepally, B.S. et al. (2021) ‘Cost-effectiveness of surgical mask, N-95 respirator, hand-hygiene and surgical mask with hand hygiene in the prevention of COVID-19: Cost effectiveness analysis from Indian context’, Clinical Epidemiology and Global Health, 10, p. 100702. doi:10.1016/j.cegh.2021.100702. Brady, M.T., Evans, J. and Cuartas, J. (1990) ‘Survival and disinfection of parainfluenza viruses 14 on environmental surfaces’, American Journal of Infection Control, 18(1), pp. 18–23. doi:10.1016/0196-6553(90)90206-8. Burgert, C. (2013) ‘Geographic Displacement Procedure and Georeferenced Data Release Policy for the Demographic and Health Surveys’. Available at: https://dhsprogram.com/pubs/pdf/SAR7/SAR7.pdf. Burton, M. et al. (2011) ‘The Effect of Handwashing with Water or Soap on Bacterial Contamination of Hands’, International Journal of Environmental Research and Public Health, 8(1), pp. 97–104. doi:10.3390/ijerph8010097. Croft, T., Marshall, A. and Allen, C. (2020) Guide to DHS Statistics. DHS program. Available at: https://dhsprogram.com/data/Guide-to-DHS-Statistics/Guide_to_DHS_Statistics_DHS-7.htm. DHS Spatial Interpolation Working Group (2014) ‘SPATIAL INTERPOLATION WITH DEMOGRAPHIC AND HEALTH SURVEY DATA: KEY CONSIDERATIONS’. DHS Spatial Analysis Reports No. 9. Available at: https://dhsprogram.com/pubs/pdf/SAR9/SAR9.pdf. Diggle, P.J. et al. (2013) ‘Spatial and Spatio-Temporal Log-Gaussian Cox Processes: Extending the Geostatistical Paradigm’, Statistical Science, 28(4), pp. 542–563. doi:10.1214/13-STS441. van Doremalen, N. et al. (2020) ‘Aerosol and Surface Stability of SARS-CoV-2 as Compared with SARS-CoV-1’, New England Journal of Medicine, 382(16), pp. 1564–1567. doi:10.1056/NEJMc2004973. Dwyer-Lindgren, L. et al. (2019) ‘Mapping HIV prevalence in sub-Saharan Africa between 2000 and 2017’, Nature, 570(7760), pp. 189–193. doi:10.1038/s41586-019-1200-9. Elmore, R. et al. (2020) ‘Risk and Protective Factors in the COVID-19 Pandemic: A Rapid Evidence Map’, Frontiers in Public Health, 8, p. 582205. doi:10.3389/fpubh.2020.582205. Gao, Y. et al. (2021) ‘Risk factors for severe and critically ill COVID‐19 patients: A review’, Allergy, 76(2), pp. 428–455. doi:10.1111/all.14657. Gauvin, L. et al. (2021) ‘Socio-economic determinants of mobility responses during the first wave of COVID-19 in Italy: from provinces to neighbourhoods’, Journal of The Royal Society Interface, 18(181), p. 20210092. doi:10.1098/rsif.2021.0092. Gibson, L.L. et al. (2002) ‘Quantitative assessment of risk reduction from hand washing with antibacterial soaps: QUANTITATIVE MICROBIAL RISK ASSESSMENT’, Journal of Applied Microbiology, 92, pp. 136S-143S. doi:10.1046/j.1365-2672.92.5s1.17.x. Guaitoli, G. and Pancrazi, R. (2021) ‘Covid-19: Regional policies and local infection risk: Evidence from Italy with a modelling study’, The Lancet Regional Health - Europe, 8, p. 100169. doi:10.1016/j.lanepe.2021.100169. Guleid, F.H. et al. (2021) ‘A bibliometric analysis of COVID-19 research in Africa’, BMJ Global Health, 6(5), p. e005690. doi:10.1136/bmjgh-2021-005690. 15 Gwaltney, J.M. (1978) ‘Hand-to-Hand Transmission of Rhinovirus Colds’, Annals of Internal Medicine, 88(4), p. 463. doi:10.7326/0003-4819-88-4-463. Hendley, J.O., Wenzel, R.P. and Gwaltney, J.M. (1973) ‘Transmission of Rhinovirus Colds by Self-Inoculation’, New England Journal of Medicine, 288(26), pp. 1361–1364. doi:10.1056/NEJM197306282882601. Hirose, R. et al. (2021) ‘Survival of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS- CoV-2) and Influenza Virus on Human Skin: Importance of Hand Hygiene in Coronavirus Disease 2019 (COVID-19)’, Clinical Infectious Diseases, 73(11), pp. e4329–e4335. doi:10.1093/cid/ciaa1517. Hoekstra, R.M. et al. (2012) ‘Impact of a Hygiene Curriculum and the Installation of Simple Handwashing and Drinking Water Stations in Rural Kenyan Primary Schools on Student Health and Hygiene Practices’, The American Journal of Tropical Medicine and Hygiene, 87(4), pp. 594–601. doi:10.4269/ajtmh.2012.11-0494. Hu, J. and Wang, Y. (2021) ‘The Clinical Characteristics and Risk Factors of Severe COVID- 19’, Gerontology, 67(3), pp. 255–266. doi:10.1159/000513400. Longini, I.M. et al. (1982) ‘ESTIMATING HOUSEHOLD AND COMMUNITY TRANSMISSION PARAMETERS FOR INFLUENZA’, American Journal of Epidemiology, 115(5), pp. 736–751. doi:10.1093/oxfordjournals.aje.a113356. Luby, S.P. et al. (2001) ‘Microbiologic effectiveness of hand washing with soap in an urban squatter settlement, Karachi, Pakistan’, Epidemiology and Infection, 127(2), pp. 237–244. doi:10.1017/S0950268801005829. Luby, S.P. et al. (2005) ‘Effect of handwashing on child health: a randomised controlled trial’, The Lancet, 366(9481), pp. 225–233. doi:10.1016/S0140-6736(05)66912-7. Ma, C. et al. (2020) Incidence, clinical characteristics and prognostic factor of patients with COVID-19: a systematic review and meta-analysis. preprint. Infectious Diseases (except HIV/AIDS). doi:10.1101/2020.03.17.20037572. Ma, Q. et al. (2020) ‘Potential utilities of mask‐wearing and instant hand hygiene for fighting SARS‐CoV‐2’, Journal of Medical Virology, 92(9), pp. 1567–1571. doi:10.1002/jmv.25805. Macharia, P.M., Joseph, N.K. and Okiro, E.A. (2020) ‘A vulnerability index for COVID-19: spatial analysis at the subnational level in Kenya’, BMJ Global Health, 5(8), p. e003014. doi:10.1136/bmjgh-2020-003014. MacIntyre, C.R. et al. (2012) ‘Respiratory viruses transmission from children to adults within a household’, Vaccine, 30(19), pp. 3009–3014. doi:10.1016/j.vaccine.2011.11.047. Marra, G. and Wood, S.N. (2011) ‘Practical variable selection for generalized additive models’, 16 Computational Statistics & Data Analysis, 55(7), pp. 2372–2387. doi:10.1016/j.csda.2011.02.004. Masaki, T. et al. (2020) Small Area Estimation of Non-Monetary Poverty with Geospatial Data. World Bank, Washington, DC. doi:10.1596/1813-9450-9383. Meda, M. et al. (2020) ‘Unintended consequences of long-sleeved gowns in a critical care setting during the COVID-19 pandemic’, Journal of Hospital Infection, 106(3), pp. 605–609. doi:10.1016/j.jhin.2020.07.036. Mocumbi, A.O. et al. (2019) ‘Cardiovascular Effects of Indoor Air Pollution from Solid Fuel: Relevance to Sub-Saharan Africa’, Current Environmental Health Reports, 6(3), pp. 116–126. doi:10.1007/s40572-019-00234-8. Mu, X., Yeh, A.G.-O. and Zhang, X. (2021) ‘The interplay of spatial spread of COVID-19 and human mobility in the urban system of China during the Chinese New Year’, Environment and Planning B: Urban Analytics and City Science, 48(7), pp. 1955–1971. doi:10.1177/2399808320954211. Najnin, N. et al. (2019) ‘Impact of a Large-Scale Handwashing Intervention on Reported Respiratory Illness: Findings from a Cluster-Randomized Controlled Trial’, The American Journal of Tropical Medicine and Hygiene, 100(3), pp. 742–749. doi:10.4269/ajtmh.18-0644. Ou, M. et al. (2020) ‘Risk factors of severe cases with COVID-19: a meta-analysis’, Epidemiology and Infection, 148, p. e175. doi:10.1017/S095026882000179X. Rabie, T. and Curtis, V. (2006) ‘Handwashing and risk of respiratory infections: a quantitative systematic review’, Tropical Medicine and International Health, 11(3), pp. 258–267. doi:10.1111/j.1365-3156.2006.01568.x. Saunders-Hastings, P. et al. (2017) ‘Effectiveness of personal protective measures in reducing pandemic influenza transmission: A systematic review and meta-analysis’, Epidemics, 20, pp. 1– 20. doi:10.1016/j.epidem.2017.04.003. Smith, K.R. et al. (2000) ‘Indoor air pollution in developing countries and acute lower respiratory infections in children’, Thorax, 55(6), p. 518. doi:10.1136/thorax.55.6.518. Sturrock, H.J.W. et al. (2013) ‘The Use of Bivariate Spatial Modeling of Questionnaire and Parasitology Data to Predict the Distribution of Schistosoma haematobium in Coastal Kenya’, PLoS Neglected Tropical Diseases. Edited by X.-N. Zhou, 7(1), p. e2016. doi:10.1371/journal.pntd.0002016. Takahashi, S. et al. (2017) ‘The geography of measles vaccination in the African Great Lakes region’, Nature Communications, 8(1), p. 15585. doi:10.1038/ncomms15585. Talaat, M. et al. (2011) ‘Effects of Hand Hygiene Campaigns on Incidence of Laboratory- 17 confirmed Influenza and Absenteeism in Schoolchildren, Cairo, Egypt’, Emerging Infectious Diseases, 17(4), pp. 619–625. doi:10.3201/eid1704.101353. Talic, S. (2021) ‘Effectiveness of public health measures in reducing the incidence of covid-19, SARS-CoV-2 transmission, and covid-19 mortality: systematic review and meta-analysis’, BMJ, p. n2997. doi:10.1136/bmj.n2997. Udwadia, Z.F. et al. (2020) ‘COVID-19 -Tuberculosis interactions: When dark forces collide’, Indian Journal of Tuberculosis, 67(4), pp. S155–S162. doi:10.1016/j.ijtb.2020.07.003. Utazi, C.E. et al. (2020) ‘Geospatial variation in measles vaccine coverage through routine and campaign strategies in Nigeria: Analysis of recent household surveys’, Vaccine, 38(14), pp. 3062–3071. doi:10.1016/j.vaccine.2020.02.070. Vagliasindi, M. (2021) Measuring the Economic Impact of COVID-19 with Real-Time Electricity Indicators. The World Bank (Policy Research Working Papers). doi:10.1596/1813-9450-9806. Verity, R. et al. (2020) ‘Estimates of the severity of coronavirus disease 2019: a model-based analysis’, The Lancet Infectious Diseases, 20(6), pp. 669–677. doi:10.1016/S1473- 3099(20)30243-7. Wang, P. and Tian, D. (2021) ‘Bibliometric analysis of global scientific research on COVID-19’, Journal of Biosafety and Biosecurity, 3(1), pp. 4–9. doi:10.1016/j.jobb.2020.12.002. Weber, T.P. and Stilianakis, N.I. (2008) ‘Inactivation of influenza A viruses in the environment and modes of transmission: A critical review’, Journal of Infection, 57(5), pp. 361–373. doi:10.1016/j.jinf.2008.08.013. Weiss, D.J. et al. (2019) ‘Mapping the global prevalence, incidence, and mortality of Plasmodium falciparum, 2000–17: a spatial and temporal modelling study’, The Lancet, 394(10195), pp. 322–331. doi:10.1016/S0140-6736(19)31097-9. WHO (2020) Transmission of SARS-CoV-2: implications for infection prevention precautions. Available at: https://www.who.int/news-room/commentaries/detail/transmission-of-sars-cov-2- implications-for-infection-prevention-precautions (Accessed: 10 August 2021). Winskill, P. et al. (2020) Report 22: Equity in response to the COVID-19 pandemic: an assessment of the direct and indirect impacts on disadvantaged and vulnerable populations in low- and lower middle-income countries. Imperial College London. doi:10.25561/78965. World Bank (2020) The Global Economic Outlook During the COVID-19 Pandemic: A Changed World, World Bank. Available at: https://www.worldbank.org/en/news/feature/2020/06/08/the- global-economic-outlook-during-the-covid-19-pandemic-a-changed-world (Accessed: 16 February 2021). Wu, Z. and McGoogan, J.M. (2020) ‘Characteristics of and Important Lessons From the 18 Coronavirus Disease 2019 (COVID-19) Outbreak in China: Summary of a Report of 72 314 Cases From the Chinese Center for Disease Control and Prevention’, JAMA, 323(13), p. 1239. doi:10.1001/jama.2020.2648. Yang, C. (2020) ‘Does hand hygiene reduce SARS-CoV-2 transmission?’, Graefe’s Archive for Clinical and Experimental Ophthalmology, 258(5), pp. 1133–1134. doi:10.1007/s00417-020- 04652-5. Zachreson, C. et al. (2021) ‘Risk mapping for COVID-19 outbreaks in Australia using mobility data’, Journal of The Royal Society Interface, 18(174), p. 20200657. doi:10.1098/rsif.2020.0657. Zvobgo, L. and Do, P. (2020) ‘COVID-19 and the call for “Safe Hands”: Challenges facing the under-resourced municipalities that lack potable water access - A case study of Chitungwiza municipality, Zimbabwe’, Water Research X, 9, p. 100074. doi:10.1016/j.wroa.2020.100074. 19 Appendix Table A1. Assumed hospitalization rates for symptomatic infections by age category. Estimates are based on Verity (Verity et al., 2020) at and multiplied by 1.67 to reflect rates among symptomatic infections assuming 60% of infections are symptomatic. Age category Hospitalization rate (%) <20 0.3 20-24 0.8 25-29 1.7 30-34 2.7 35-39 3.8 40-44 4.8 45-49 6.5 50-54 9.7 55-59 12.0 60-64 17.0 65-69 19.5 70-74 24.3 75-79 29.5 80+ 30.0