Comparing different spatio-temporal modeling methods in dengue fever data analysis in Colombia during 2012-2015

Predicting spatio-temporal dynamics of dengue using INLA (integrated nested laplace approximation) in Yogyakarta, Indonesia

INTRODUCTION

Dengue is a mosquito-borne disease caused by the dengue virus, primarily transmitted by Aedes aegypti and Aedes albopictus. Its incidence fluctuates due to spatial and temporal factors, necessitating robust modeling approaches for prediction and risk mapping.

OBJECTIVES

This study aims to develop a spatio-temporal Bayesian model for predicting dengue incidence, integrating climatic, sociodemographic, and environmental factors to improve outbreak forecasting.

METHODS

An ecological study was conducted in the Special Region of Yogyakarta, Indonesia (January 2017-December 2022) using monthly panel data from 78 sub-districts. Secondary data sources included dengue surveillance (Health Office), meteorological data (NASA POWER), sociodemographic data (BPS-Statistics Indonesia), and land use data (Sentinel-2, ESRI). Predictors included rainfall, temperature, humidity, wind speed, atmospheric pressure, population density, and land use patterns. Data analysis was performed using R-INLA, with model performance assessed using Deviance Information Criterion (DIC), Watanabe-Akaike Information Criterion (WAIC), marginal log-likelihood, Mean Absolute Error (MAE), and Root Mean Squared Error (RMSE).

RESULTS

The INLA-based Bayesian model effectively captured spatial and temporal dengue dynamics. Key predictors included rainfall lag 1 and 2 (mean = 0.001), temperature (mean = 0.151, CI: 0.090-0.210), humidity (mean = 0.056, CI: 0.040-0.073), built area (mean = 0.001), and water area (mean = 0.008, CI: 0.005-0.011). Spatial clustering (BYM model, precision = 2163.53) indicated that dengue cases were concentrated in specific areas. The RW2 model (precision = 49.11) confirmed seasonal trends, highlighting climate's role in disease transmission. Model evaluation metrics (DIC = 15017.88, WAIC = 15294.54, log-likelihood = -7845.857) demonstrated good predictive performance. Furthermore, the model's accuracy was assessed using MAE and RMSE values, where MAE = 1.77 indicates an average prediction error of 1-2 cases, while RMSE = 2.97 suggests the presence of occasional larger discrepancies. The RMSE's higher value relative to MAE highlights instances where prediction errors were more significant, as RMSE is more sensitive to large deviations.

CONCLUSIONS

The INLA-based spatio-temporal model is an effective tool for dengue prediction, offering valuable insights for early warning systems and targeted vector control strategies, thereby improving disease prevention and response efforts.

A Bayesian Multivariate Model With Temporal Dependence on Random Partition of Areal Data for Mosquito-Borne Diseases

More than half of the world's population is exposed to mosquito-borne diseases, leading to millions of cases and hundreds of thousands of deaths every year. Analyzing this type of data is complex and poses several interesting challenges, mainly due to the usually vast geographic area involved, the peculiar temporal behavior, and the potential correlation between infections. Motivation for this work stems from the analysis of tropical disease data, namely, the number of cases of dengue and chikungunya, for the 145 microregions in Southeast Brazil from 2018 to 2022. As a contribution to the literature on multivariate disease data, we develop a flexible Bayesian multivariate spatio-temporal model where temporal dependence is defined for areal clusters. The model features a prior distribution for the random partition of areal data that incorporates neighboring information. It also incorporates an autoregressive structure and terms related to seasonal patterns into temporal components that are disease- and cluster-specific. Furthermore, it considers a multivariate directed acyclic graph autoregressive structure to accommodate spatial and inter-disease dependence. We explore the properties of the model through simulation studies and show results that prove our proposal compares well to competing alternatives. Finally, we apply the model to the motivating dataset with a twofold goal: finding clusters of areas with similar temporal trends for some of the diseases and exploring the existence of correlation between two diseases transmitted by the same mosquito.

A systematic review of the data, methods and environmental covariates used to map Aedes-borne arbovirus transmission risk

BACKGROUND

Aedes (Stegomyia)-borne diseases are an expanding global threat, but gaps in surveillance make comprehensive and comparable risk assessments challenging. Geostatistical models combine data from multiple locations and use links with environmental and socioeconomic factors to make predictive risk maps. Here we systematically review past approaches to map risk for different Aedes-borne arboviruses from local to global scales, identifying differences and similarities in the data types, covariates, and modelling approaches used.

METHODS

We searched on-line databases for predictive risk mapping studies for dengue, Zika, chikungunya, and yellow fever with no geographical or date restrictions. We included studies that needed to parameterise or fit their model to real-world epidemiological data and make predictions to new spatial locations of some measure of population-level risk of viral transmission (e.g. incidence, occurrence, suitability, etc.).

RESULTS

We found a growing number of arbovirus risk mapping studies across all endemic regions and arboviral diseases, with a total of 176 papers published 2002-2022 with the largest increases shortly following major epidemics. Three dominant use cases emerged: (i) global maps to identify limits of transmission, estimate burden and assess impacts of future global change, (ii) regional models used to predict the spread of major epidemics between countries and (iii) national and sub-national models that use local datasets to better understand transmission dynamics to improve outbreak detection and response. Temperature and rainfall were the most popular choice of covariates (included in 50% and 40% of studies respectively) but variables such as human mobility are increasingly being included. Surprisingly, few studies (22%, 31/144) robustly tested combinations of covariates from different domains (e.g. climatic, sociodemographic, ecological, etc.) and only 49% of studies assessed predictive performance via out-of-sample validation procedures.

CONCLUSIONS

Here we show that approaches to map risk for different arboviruses have diversified in response to changing use cases, epidemiology and data availability. We identify key differences in mapping approaches between different arboviral diseases, discuss future research needs and outline specific recommendations for future arbovirus mapping.

Spatial Analysis of Dengue Clusters at Department, Municipality and Local Scales in the Southwest of Colombia, 2014-2019

Dengue is an arbovirus transmitted by mosquitoes of the genus Aedes and is one of the 15 main public health problems in the world, including Colombia. Where limited financial resources create a problem for management, there is a need for the department to prioritize target areas for public health implementation. This study focuses on a spatio-temporal analysis to determine the targeted area to manage the public health problems related to dengue cases. To this end, three phases at three different scales were carried out. First, for the departmental scale, four risk clusters were identified in Cauca (RR ≥ 1.49) using the Poisson model, and three clusters were identified through Getis-Ord Gi* hotspots analysis; among them, Patía municipality presented significantly high incidence rates in the time window (2014-2018). Second, on the municipality scale, altitude and minimum temperature were observed to be more relevant than precipitation; considering posterior means, no spatial autocorrelation for the Markov Chain Monte Carlo was found (Moran test ˂ 1.0), and convergence was reached for b₁-b₁₀₅ with 20,000 iterations. Finally, on the local scale, a clustered pattern was observed for dengue cases distribution (nearest neighbour index, NNI = 0.202819) and the accumulated number of pupae (G = 0.70007). Two neighbourhoods showed higher concentrations of both epidemiological and entomological hotspots. In conclusion, the municipality of Patía is in an operational scenario of a high transmission of dengue.

Spatiotemporal dynamics of dengue in Colombia in relation to the combined effects of local climate and ENSO

Dengue virus (DENV) is an endemic disease in the hot and humid low-lands of Colombia. We characterize the association of monthly series of dengue cases with indices of El Niño/Southern Oscillation (ENSO) at the tropical Pacific and local climatic variables in Colombia during the period 2007-2017 at different temporal and spatial scales. For estimation purposes, we use lagged cross-correlations (Pearson test), cross-wavelet analysis (wavelet cross spectrum, and wavelet coherence), as well as a novel nonlinear causality method, PCMCI, that allows identifying common causal drivers and links among high dimensional simultaneous and time-lagged variables. Our results evidence the strong association of DENV cases in Colombia with ENSO indices and with local temperature and rainfall. El Niño (La Niña) phenomenon is related to an increase (decrease) of dengue cases nationally and in most regions and departments, with maximum correlations occurring at shorter time lags in the Pacific and Andes regions, closer to the Pacific Ocean. This association is mainly explained by the ENSO-driven increase in temperature and decrease in rainfall, especially in the Andes and Pacific regions. The influence of ENSO is not stationary, given the reduction of DENV cases since 2005, and that local climate variables vary in space and time, which prevents to extrapolate results from one region to another. The association between DENV and ENSO varies at national and regional scales when data are disaggregated by seasons, being stronger in DJF and weaker in SON. Overall, the Pacific and Andes regions control the relationship between dengue dynamics and ENSO at national scale. Cross-wavelet analysis indicates that the ENSO-DENV relation in Colombia exhibits a strong coherence in the 12 to 16-months frequency band, which implies the frequency locking between the annual cycle and the interannual (ENSO) timescales. Results of nonlinear causality metrics reveal the complex concomitant effects of ENSO and local climate variables, while offering new insights to develop early warning systems for DENV in Colombia.

Fine scale analysis of malaria incidence in under-5: hierarchical Bayesian spatio-temporal modelling of routinely collected malaria data between 2012-2018 in Cameroon

The current study aims to provide a fine-scale spatiotemporal estimate of malaria incidence among Cameroonian under-5, and to determine its associated environmental factors, to set up preventive interventions that are adapted to each health district of Cameroon. Routine data on symptomatic malaria in children under-5 collected in health facilities, between 2012 and 2018 were used. The trend of malaria cases was assessed by the Mann–Kendall (M–K) test. A time series decomposition was applied to malaria incidence to extract the seasonal component. Malaria risk was estimated by the standardised incidence ratio (SIR) and smoothed by a hierarchical Bayesian spatiotemporal model. In total, 4,052,216 cases of malaria were diagnosed between 2012 and 2018. There was a gradual increase per year, from 369,178 in 2012 to 652,661 in 2018. After adjusting the data for completeness, the national incidence ranged from 489‰ in 2012 to 603‰ in 2018, with an upward trend (M–K test p-value < 0.001). At the regional level, an upward trend was observed in Adamaoua, Centre without Yaoundé, East, and South regions. There was a positive spatial autocorrelation of the number of malaria incident-cases per district per year as suggested by the Moran’s I test (statistic range between 0.11 and 0.53). The crude SIR showed a heterogeneous malaria risk with values ranging from 0.00 to 8.90, meaning that some health districts have a risk 8.9 times higher than the national annual level. The incidence and risk of malaria among under-5 in Cameroon are heterogeneous and vary significantly across health districts and seasons. It is crucial to adapt malaria prevention measures to the specificities of each health district, in order to reduce its burden in health districts where the trend is upward.

Spatial connectivity in mosquito-borne disease models: a systematic review of methods and assumptions

Spatial connectivity plays an important role in mosquito-borne disease transmission. Connectivity can arise for many reasons, including shared environments, vector ecology and human movement. This systematic review synthesizes the spatial methods used to model mosquito-borne diseases, their spatial connectivity assumptions and the data used to inform spatial model components. We identified 248 papers eligible for inclusion. Most used statistical models (84.2%), although mechanistic are increasingly used. We identified 17 spatial models which used one of four methods (spatial covariates, local regression, random effects/fields and movement matrices). Over 80% of studies assumed that connectivity was distance-based despite this approach ignoring distant connections and potentially oversimplifying the process of transmission. Studies were more likely to assume connectivity was driven by human movement if the disease was transmitted by an Aedes mosquito. Connectivity arising from human movement was more commonly assumed in studies using a mechanistic model, likely influenced by a lack of statistical models able to account for these connections. Although models have been increasing in complexity, it is important to select the most appropriate, parsimonious model available based on the research question, disease transmission process, the spatial scale and availability of data, and the way spatial connectivity is assumed to occur.

Exploring for Municipality-Level Socioeconomic Variables Related to Zika Virus Incidence in Colombia

Colombia experienced an outbreak of Zika virus infection during September 2015 until July 2016. This study aimed to identify the socioeconomic factors that at the municipality level correlate with this outbreak and therefore could have influenced its incidence. An analysis of publicly available, municipality-aggregated data related to eight potential explanatory socioeconomic variables was conducted. These variables are school dropout, low energy strata, social security system, savings capacity, tax, resources, investment, and debt. The response variable of interest in this study is the number of reported cases of Zika virus infection per people (projected) per square kilometer. Binomial regression models were performed. Results show that the best predictor variables of Zika virus occurrence, assuming an expected inverse relationship with socioeconomic status, are "school", "energy", and "savings". Contrary to expectations, proxies of socioeconomic status such as "investment", "tax", and "resources" were associated with an increase in the occurrence of Zika virus infection, while no association was detected for "social security" and "debt". Energy stratification, school dropout rate, and the percentage of the municipality's income that is saved conformed to the hypothesized inverse relationship between socioeconomic standing and Zika occurrence. As such, this study suggests these factors should be considered in Zika risk modeling.