Model building and assessment of the impact of covariates for disease prevalence mapping in low-resource settings: to explain and to predict

Abstract

This paper provides statistical guidance on the development and application of model-based geostatistical methods for disease prevalence mapping. We illustrate the different stages of the analysis, from exploratory analysis to spatial prediction of prevalence, through a case study on malaria mapping in Tanzania. Throughout the paper, we distinguish between predictive modelling, whose main focus is on maximizing the predictive accuracy of the model, and explanatory modelling, where greater emphasis is placed on understanding the relationships between the health outcome and risk factors. We demonstrate that these two paradigms can result in different modelling choices. We also propose a simple approach for detecting over-fitting based on inspection of the correlation matrix of the estimators of the regression coefficients. To enhance the interpretability of geostatistical models, we introduce the concept of domain effects in order to assist variable selection and model validation. The statistical ideas and principles illustrated here in the specific context of disease prevalence mapping are more widely applicable to any regression model for the analysis of epidemiological outcomes but are particularly relevant to geostatistical models, for which the separation between fixed and random effects can be ambiguous.

Spatial distribution and risk factors for human cysticercosis in Colombia

“…Despite the lack of data concerning some geographical areas in Colombia, the authors still consider the study outcomes as valuable and indicative of the situation of cysticercosis in the country. In addition, the information provided in the current study could be further used to build models that can spatially predict the disease seroprevalence in non-sampled areas [ 17 ], offering a cost-effective tool for decision-makers in places where direct sampling did not take place.…”

“…Geostatistical approaches will play an important role in identifying areas of high transmission, particularly if they can be parameterized to identify likely areas of high transmission using Geographical Information System (GIS) data that have comprehensive global coverage. Although our study focused on the identification of risk factors associated with exposure to T. solium and residual degrees of spatial clustering, similar geostatistical and machine learning approaches can be used that focus on predicting the spatial distribution of disease using GIS data [ 17 ]. Such approaches, conducted at national and global scales, will be crucial in assisting progress towards the WHO’s 2030 goals [ 22 , 58 ].…”

“…Before performing the geospatial analysis, an initial exploratory analysis was undertaken (using R version 4.0.5 [ 17 ]). Given the clustered nature of the data, a hierarchical univariate mixed-effects logistic regression model was fitted to test the association between each explanatory variable (covariate) and human seropositivity to T. solium cysticerci, with each model including two independent random effects terms to capture correlation at the municipality and neighbourhood/village (depending on urban or rural location) levels.…”

“…Following fitting of the multivariable mixed-effects model, a variogram analysis was performed to assess the presence of residual spatial correlation [ 17 ]. Since the geographical coordinates were available only for the municipalities and not for the neighbourhoods/villages, the empirical variogram was computed only on , the estimated random effects at the municipality level.…”

Public Health Policy Pillars for the Sustainable Elimination of Zoonotic Schistosomiasis

“…Many infection and population characteristics which influence disease transmission, such as present prevalence, age-intensity distributions, community-and age-contact rates and transmission rates, vary spatially. Therefore, studies mapping the spatial distribution in endemic areas need to be performed (27,28,29,30,31,32). Furthermore, some biological aspects of infection and hosts cannot be, or are difficult to be, measured.…”