Estimating the population at risk with soil transmitted helminthiasis and annual drug requirements for preventive chemotherapy in Ogun State, Nigeria

Soil transmitted helminth (STH) infections are among the most common human infections worldwide with over 1 billion people affected. Many estimates of STH infection are often based on school-aged children (SAC). This study produced predictive risk-maps of STH on a more finite scale, estimated the number of people infected, and the amount of drug required for preventive chemotherapy (PC) in Ogun state, Nigeria. Georeferenced STH infection data obtained from a cross-sectional survey at 33 locations between July 2016 and November 2018, together with remotely-sensed environmental and socio-economic data were analyzed using Bayesian geostatistical modelling. Stepwise variable selection procedure was employed to select a parsimonious set of predictors to predict risk and spatial distribution of STH infections. The number of persons (pre-school ages children, SAC and adults) infected with STH were estimated, with the amount of tablets needed for preventive chemotherapy. An overall prevalence of 17.2% (95% CI 14.9, 19.5) was recorded for any STH infection. Ascaris lumbricoides infections was the most predominant, with an overall prevalence of 13.6% (95% CI 11.5, 15.7), while Hookworm and Trichuris trichiura had overall prevalence of 4.6% (95% CI 3.3, 5.9) and 1.7% (95% CI 0.9, 2.4), respectively. The model-based prevalence predictions ranged from 5.0 to 23.8% for Ascaris lumbricoides, from 2.0 to 14.5% for hookworms, and from 0.1 to 5.7% for Trichuris trichiura across the implementation units. The predictive maps revealed a spatial pattern of high risk in the central, western and on the border of Republic of Benin. The model identified soil pH, soil moisture and elevation as the main predictors of infection for A. lumbricoides, Hookworms and T. trichiura respectively. About 50% (10/20) of the implementation units require biannual rounds of mass drug administration. Approximately, a total of 1.1 million persons were infected and require 7.8 million doses. However, a sub-total of 375,374 SAC were estimated to be infected, requiring 2.7 million doses. Our predictive risk maps and estimated PC needs provide useful information for the elimination of STH, either for resource acquisition or identifying priority areas for delivery of interventions in Ogun State, Nigeria.


Scientific Reports
| (2022) 12:2027 | https://doi.org/10.1038/s41598-022-06012-1 www.nature.com/scientificreports/ We therefore present findings from a geostatistical analysis of soil-transmitted helminth infection data that were obtained from a state-wide community-based survey in Ogun State, Nigeria. The aims of this study were (1) to map and predict the spatial distribution of soil-transmitted helminth infections at 2 km spatial scale using a Bayesian geostatistical approach; (2) identify the most important climatic, environmental and socioeconomic determinants of soil-transmitted helminth infections (3) calculate the number of persons infected and; (4) estimate the annual drug requirements for preventive chemotherapy according to guidelines put forward by the World Health Organization (WHO).

Methods
Study area, design and population. This study was conducted in Ogun State, Nigeria (Fig. 1). Details of the study area, design and population surveyed have been described elsewhere 19 . In brief, the study was carried out between July 2016 and November 2018 spanning across both wet and dry season. We designed a crosssectional survey, and employed a systematic grid sampling method in the selection of communities to ensure an unbiased representation across the state 19 . A total of 1499 children and adults, from 33 spatially selected communities participated in the study 19 . In each community, all households and their occupants were considered eligible for participation and invited to participate in the study. STH Infection data. The field and laboratory procedures have been previously described in 19 . In brief, stool container was distributed to consenting household members in advance. Participants' unique identifiers were marked on the containers and detailed instructions of how to collect a fresh morning stool sample were given. Stool samples were processed in a designated area provided by the community leader. Duplicate sediment slides were prepared from 1 g of each stool using SAF-Ether concentration method 19,29 . The slides were examined under a light microscope by experienced laboratory technicians 2 h post sample collection. Infection was defined as the presence of at least one helminth egg on one of the two slides. The parasites' eggs were counted for each species, and number of eggs per species and per stool examined was recorded for each participant 19 . Environmental and socio-economic predictors. Nine environmental variables (elevation, enhanced vegetation index (EVI), normalized difference vegetation index (NDVI), land surface temperature for day (LSTD), land surface temperature for night (LSTN), rainfall, population, soil pH, and soil moisture; three socioeconomic variables (night light emission (NLE), improved access to sanitation facilities and improved access to drinking water facilities) were used in the analysis. NLE was used as a proxy for urbanization and economic growth. These variables were chosen because they are either directly associated with prevalence of STH infections or they serve as proxy for other factors that are known to influence STH transmission 30 . All environmental  Variable selection procedures for the geostatistical model. To select the best set of covariates for the geostatistical model, we first examined the covariates for correlation using Pearson's rank correlation index. Pairs of covariates with high correlation values (Pearson correlation > 0.7) were identified and only one of the correlated variables was included in the modelling process. The one included was chosen by visualizing the association via a scatterplot. Then we used both forward and backward stepwise selection to select a parsimonious set of covariates required for the prediction of STH among all the candidate set of covariates. This is achieved by fitting a non-spatial generalized linear model relating the prevalence of each STH species with the covariates. The final set of covariates result in a model with the lowest Akaike information criterion (AIC) and a further inclusion of any of the covariates does not improve the performance of the model. Geostatistical modelling. The prevalence survey data available for this analysis are, at any geographical location x, the number of individuals tested, m and the number of people tested positive for each of the STH species, y k , where y 1 is the number of people that tested positive for Ascaris, y 2 for Hookworm, and y 3 for Trichuris. The sampling distribution of y k is binomial with number of trials m and probability of positive outcome P k (x) , the prevalence at x. The variation of P(x) was modelled using a combination of socio-economic and environmental covariate effects d(x); unexplained residual spatial variation, S(x). Therefore, we developed a binomial logistic geospatial model given as S(x) is modelled as a zero-mean discretely indexed Gaussian Markov Random Field (GMRF) defined on the triangulation of the domain of interest, such that the correlation between any two locations x i and x j is modelled using Matérn correlation function. S(x) serves two purposes in the model; (1) it helps to capture the geographical variation; and (2) it helps to predict prevalence at unobserved locations. The Matérn covariance function is given as where d = x i − x j , κ is a scaling parameter, K ν is the modified Bessel function of second kind and order ν > 0 and Ŵ(.) is the Gamma-function, σ 2 is the variance, and the spatial range ρ = √ 8/κ , the distance at which the spatial correlation is becomes negligible (< 0.1).
The model was fitted using the Integrated Nested Laplace Approximation (INLA) 36,37 and the Stochastic Partial Differential Equation (SPDE) 38 approaches. INLAallows us to perform a fast Bayesian inference. Because Cov S(x i ), S x j = σ 2 2 ν−1 Ŵ(ν) (κu) ν K ν (κu), www.nature.com/scientificreports/ no prior information was available, an independent vague zero-mean Gaussian prior distribution was assigned to the fixed and random effects parameters. Posterior distributions were obtained for all the parameters and were summarised to obtain the mean and 95% credible interval (CI). Prediction of the prevalence of each of the STH species were provided at 2 km spatial resolution throughout the study area. We then estimate the prevalence of any STH prevalence by assuming that each species is independent 38 . The predicted prevalence was represented as the posterior mean.

Model validation.
We validated our geostatistical model by assessing the predictive performance of the model using the 5-folds cross-validation. All survey data were randomly splitted into 5 groups. We hold-out each unique group then fit the model on the remaining groups and evaluate the predictive performance of the model on the hold-out group. The withheld data was matched with the predictions to summarize the performance of the model using the correlation, bias, root mean square error (RMSE) and the coverage probability.
Estimating the amount of the anthelmintic treatment required. We estimated the amount of anthelmintic treatment (albendazole or mebendazole) needed to treat the population annually at the local government areas in Ogun state. According to the WHO STH treatment decision tree 39 , prevalence of STH should be examined after 5-6 rounds of annual or biannual PC. Subsequent chemotherapy campaigns after this evaluation should continue according to a set of endemicity classes defined by the following prevalence thresholds; suspend PC if prevalence is < 2%; biennial PC if prevalence is between 2 and 10%; annual PC if prevalence is between 10 and 20%; biannual PC if prevalence is between 20 and 50%; triannual PC is prevalence is greater than 50% 39 . Hence, we computed the total number of anthelmintic drugs by classifying each pixel according to the treatment decision, thereby estimating the number of MDA rounds. Then multiply the number of MDA rounds by the total population of that pixel. Hence, we aggregate across the pixels the number of anthelmintic drugs over the local government areas. Also, we estimated the number of anthelmintic drugs required to treat school-aged children (SAC) by multiplying the number of MDA rounds per pixel by the population of SAC per pixel. Then we constructed the local government level estimate by aggregating across the pixels.

Estimating the number of people infected with STH.
To estimate the number of people infected with STH parasites, we multiplied the prevalence at each pixel by the total population at that pixel. Also, to estimate the number of SAC infected with STH parasite, we multiplied the prevalence at each pixel by the population of SAC at that pixel. Hence, to construct the local government level estimate, we aggregate the values across the pixels. The 95% confidence interval of the estimate was constructed by using the prevalence values at the 2.5% and 97.5% quantiles.
Ethical approval. Ethical clearance for this study (HPRS/381/183) was obtained from the Ethics review committee of Department of Planning, Research and Statistics, Ogun State Ministry of Health, Oke Imosan Abeokuta, Nigeria. Prior to data collection, visitations were made to the LGAs and the selected communities were the objectives and study procedures were explained and permissions for field survey were sought. Written informed consents were obtained from household heads and corresponding occupants of their households. Children below age sixteen, completed assent forms through their parents or guardians. All methods including recruitment of participants, collection of participant's data and samples, laboratory analysis and data management were performed in accordance with the 1964 Declarations of Helsinki.

Results
Data summaries. A total of 1027 infection data was included in this survey. The demographic characteristics of the study population have been described elsewhere 19 . However, Table 2 summarizes the soil-transmitted helminths species-specific prevalence among the examined participants. In short, an overall prevalence of 17.2% (95% confidence interval (CI) 14.9, 19.5) was recorded for any STH infection. Ascaris lumbricoides infections was the most predominant, with an overall prevalence of 13.6% (95% CI 11.5, 15.7), while Hookworm and Trichuris trichiura had overall prevalence of 4.6% (95% CI 3.3, 5.9) and 1.7% (95% CI 0.9, 2.4), respectively. The geographical distribution of the empirical prevalence for each soil-transmitted helminth species has been described elsewhere 19 .
Geostatistical variable selection, model parameter estimates and model validation. Following the variable selection, Soil pH, soil moisture and elevation were selected for Ascaris, Hookworms, and Trichuris infections respectively (  (Table 4). Figure 2 present the overall and species-specific predictive risk maps of soil-transmitted helminth infections. Predictive risk map for overall STH infection shows a high prevalence (> 20%) for LGAs in the central and western part of the state. Pockets of very high prevalence (> 40%) were also predicted for the LGAs around the boundary regions in the south-western part of the state. However, predicted prevalence were predominantly between 12 and 15% in LGA N  www.nature.com/scientificreports/ the eastern part of the state. For Ascaris lumbricoides, pockets of high prevalence (> 20%) was predicted in the central and western part of Ogun State, with hotspots in the LGAs located close the border regions in the southwestern part of the country. Moderate to high prevalence (5-20%) were also predicted in these regions. However, low predicted prevalence (5-10%) were observed in the eastern part of the country, with a sparse predicted prevalence within 10-12% around the border regions (Fig. 2). For hookworms, the predicted risk map shows that most regions have prevalence below 20%, except some pocket areas in Ipokia LGA, around the boundary lines. Predicted prevalence were predominantly between 5 and 10% in the central and western part of the state. However, predicted prevalence were lower (2-5%) in the eastern regions (Fig. 2). For Trichuris trichiura infections, most of the LGAs in the northern part of the state had predicted prevalence value between 0 and 1%. Similarly, in the southern region, predicted prevalence were predominantly between 5 and 10% (Fig. 2). The uncertainty of these estimates is presented in the standard error maps which can be found in Supplementary Fig. S1.  10 LGAs has over 50,000 infected persons each, requiring more than 300,000 albendazole or mebendazole tablets. Ado Odo-Ota LGA has the highest number of infected population (168,591 persons) and requires 1,204,423 albendazole or mebendazole tablets. The least number of infected population (9103 persons) were estimated for Ijebu Northeast, requiring 64,181 tablets. However, for school-aged population, a total of 375,374 were infected, requiring 2,685,618 drugs for preventive chemotherapy. Ten LGAs in the central and western part of the state had over 10,000 infected school-aged children, and requires over 100,000 albendazole or mebendazole tablets each, for mass administration campaigns. The highest number of infected school-aged children (56,556) were estimated for Ado-Odo Ota LGA, requiring a total of 404,123 albendazole or mebendazole tablets. The least number of infected school-aged population (3118) were estimated for Ijebu North-east, requiring 21,980 tablets.

Discussion
In this study, we utilized soil-transmitted helminth infection data from a state-wide cross-sectional survey to produce model-based estimates of infection risk, number of people infected, rounds of MDA and annual drug requirements for preventive chemotherapy. Empirical prevalence estimates for each of the three STH were below Table 4. Soil-transmitted helminth species-specific model-based predicted prevalence across the 20 LGAs in Ogun State, Nigeria. LGA

Ascaris lumbricoides
Hookworms www.nature.com/scientificreports/ 20%, with Ascaris lumbricoides been the most predominant species (13.6%), followed by hookworm (4.6%) and Trichuris trichiura (1.7%) 19 . However, based on our model predictions, prevalence ranged from 5.0 to 23.8% for Ascaris lumbricoides, from 2.0 to 14.5% for hookworms, and from 0.1 to 5.7% for Trichuris trichiura across the IUs. However, location-specific predictions shows that overall STH and Ascaris infections were as high as 53% and 34% respectively, and greatest around the border of Republic of Benin in the west. Also, heavy risk approaching thresholds level necessitating preventive chemotherapy were observed in the central and western region. The risk of hookworms, also exhibit a similar pattern, however the predicted prevalence was further reduced below PC threshold levels. Majority of the LGAs were at very low risk of Trichuris infection. The spatial patterns observed in this study is in-line with the findings of 24 for the three STH species, except for Ascaris lumbricoides where additional risk was reported in the eastern part of the state. This observation might be explained by the differences in composition of the study population (total population versus SAC only) and sampling point (communities versus school) in this present study 19,24 . Furthermore, our results indicate the influence of some environmental covariates on transmission of soil transmitted helminth infections. For example, soil pH was negatively associated with Ascaris, suggesting that as pH of the soil increases, the survivability of Ascaris egg reduces. The effect of elevated pH on inactivation of Ascaris eggs have been previously reported 40 . Similarly, soil moisture was negatively associated with the risk of hookworm infection. This finding corroborates the observation of 41 . Temperature and moisture are determining factors in the development of helminth eggs 42 , with rainfall playing a major role in the restoration of the latter 43 . However, there are presumptions that heavy rainfalls might wash out soil transmitted helminth eggs from the soil 23,42,44 . This might explain the negative relationship between soil moisture and hookworm infections. Also, elevation was negatively associated with Trichuris. This supports already established evidence that the risk of Trichuris trichiura is rare or absent as altitude increases 45,46 . Our findings are also in-line with previous reports from Bolivia 47 and Nigeria 24 .
STH infections thrives in areas lacking sanitation, potable water source, personal and domestic hygiene 4,7,9,19 , hence we expected infections to be associated with socio-economic predictors such as access to improved  (1) probable loss of variability as a result of household data aggregation for community analysis 22 , (2) insufficient coverage of water and sanitation resource facilities 48 , (3) Lack of standard and better water, sanitation and hygiene assessment tool leading to information bias 49 and latrine efficiency in containing excreta 49 . Prior efforts to model the treatment needs and number of SAC infected in Ogun State, were based on national survey data collected across 555 locations in Nigeria, with less than 20 location-specific data in Ogun State 24 . Indeed, these data may not reflect the actual situation of infected SAC and drug requirements for PC in the state 24 . Our study therefore presents, a robust estimate for the state, using more recent survey data collected across 1027 locations in the State. Based on our estimate, about 55% (11/20) of the IUs in the State requires biannual rounds of MDA, while 35% (7/20) and 10% (2/20) requires annual and biennial MDA rounds respectively. We therefore estimated a total of 1.1 million infected persons (comprising pre-school aged children, school-aged children and adults) and a total of 7.8 million albendazole or mebendazole tablets in Ogun State. More specifically, we estimate that 375,374 SAC were infected and a total of 2.7 million albendazole or mebendazole tablets will be required for PC. These estimates are twice as high as the number of tablets reported in 24 for the state.
This study has shown the predicted prevalence using a robust geostatistical approach, and as well the spatial pattern of disease spread. The empirical and predicted prevalence for Ascaris infections were above 20%, hence necessitating annual PC in most regions. However, there were significant reduction in the prevalence and spread of Hookworm and Trichuris infections. This observation reflects the yields of investment made by the WHO, Table 5. Estimates of number of people infected and annual drug requirements in Ogun State, Nigeria. *Total population comprises of preschool-aged children, school-aged children and adults; CrI : Credible interval.
LGAs Total population* School-aged population www.nature.com/scientificreports/ donor agencies, and various governmental and non-governmental health development agencies supporting PC in the country. Our predictive maps and estimated drug requirements are therefore important in planning, targeting and delivery of prioritized interventions. The maps can also be utilized for designing more robust spatial surveys to meet more specialized needs including evaluation of STH control programs or long-term surveillance. Furthermore, we believe our estimations on the number of pre-school aged children and adults infected are useful, in the phase of expanding PC to adult population, to sustain accrued gains in morbidity control and interruption of transmission 51 .

Conclusion
The work presented here contributes to the existing body of knowledge on model-based estimates of the geographical distribution of soil-transmitted helminth infection risk at more finite scale (i.e., scale smaller than the implementation units) in Ogun State in Nigeria. We used data generated across a community based cross-sectional study focusing on all sub-sets of a population (pre-school aged children, school-aged children and adults) to; (1) predict disease distributions, (2) identify associated environmental and socioeconomic risk factors, (3) estimate number of persons infected, and (4) estimate annual drug requirements. Our prediction maps provide useful information for identifying priority areas where interventions targeting soil transmitted helminthiasis are most urgently required. In addition, our estimations of drug needs are useful in the process of resource acquisition, planning and delivery of interventions.

Data availability
The datasets for environmental and socio-economic variables are publicly available in the remote sensing data repositories cited within the text. The primary STH infection datasets analyzed for this study have also been previously published and are available at https:// doi. org/ 10. 1371/ journ al. pone. 02334 23. s001.