Disease Mapping Models Research Articles

Some findings on zero‐inflated and hurdle poisson models for disease mapping

Zero excess in the study of geographically referenced mortality data sets has been the focus of considerable attention in the literature, with zero-inflation being the most common procedure to handle this lack of fit. Although hurdle models have also been used in disease mapping studies, their use is more rare. We show in this paper that models using particular treatments of zero excesses are often required for achieving appropriate fits in regular mortality studies since, otherwise, geographical units with low expected counts are oversmoothed. However, as also shown, an indiscriminate treatment of zero excess may be unnecessary and has a problematic implementation. In this regard, we find that naive zero-inflation and hurdle models, without an explicit modeling of the probabilities of zeroes, do not fix zero excesses problems well enough and are clearly unsatisfactory. Results sharply suggest the need for an explicit modeling of the probabilities that should vary across areal units. Unfortunately, these more flexible modeling strategies can easily lead to improper posterior distributions as we prove in several theoretical results. Those procedures have been repeatedly used in the disease mapping literature, and one should bear these issues in mind in order to propose valid models. We finally propose several valid modeling alternatives according to the results mentioned that are suitable for fitting zero excesses. We show that those proposals fix zero excesses problems and correct the mentioned oversmoothing of risks in low populated units depicting geographic patterns more suited to the data.

Poisson-generalized gamma empirical Bayes model for disease mapping

In spatial disease mapping, the use of Bayesian models of estimation technique is becoming popular for smoothing relative risks estimates for disease mapping. The most common Bayesian conjugate model for disease mapping is the Poisson-Gamma Model (PG). To explore further the activity of smoothing of relative risk estimates for Bayesian disease mapping, this study focused on the use of generalized gamma distribution as conjugate priors with respect to Poisson likelihood. Two new empirical Bayesian (EB) models are developed; these include Poisson-Generalized Gamma model (PGG) and modified Poisson-Generalized Gamma model (MPGG). The model simulation results indicated that PGG and MPGG models are more likely to handle dispersion in zero-deflated data, contaminated data and zero-inflated data for small and large sample data. Hence, the new EB models are highly competitive to improve the efficiency of relative risk estimation for disease mapping.Keywords: Disease Mapping, Empirical Bayes, Generalized Gamma, Dispersion, Poisson

Smooth individual level covariates adjustment in disease mapping.

Spatial models for disease mapping should ideally account for covariates measured both at individual and area levels. The newly available "indiCAR" model fits the popular conditional autoregresssive (CAR) model by accommodating both individual and group level covariates while adjusting for spatial correlation in the disease rates. This algorithm has been shown to be effective but assumes log-linear associations between individual level covariates and outcome. In many studies, the relationship between individual level covariates and the outcome may be non-log-linear, and methods to track such nonlinearity between individual level covariate and outcome in spatial regression modeling are not well developed. In this paper, we propose a new algorithm, smooth-indiCAR, to fit an extension to the popular conditional autoregresssive model that can accommodate both linear and nonlinear individual level covariate effects while adjusting for group level covariates and spatial correlation in the disease rates. In this formulation, the effect of a continuous individual level covariate is accommodated via penalized splines. We describe a two-step estimation procedure to obtain reliable estimates of individual and group level covariate effects where both individual and group level covariate effects are estimated separately. This distributed computing framework enhances its application in the Big Data domain with a large number of individual/group level covariates. We evaluate the performance of smooth-indiCAR through simulation. Our results indicate that the smooth-indiCAR method provides reliable estimates of all regression and random effect parameters. We illustrate our proposed methodology with an analysis of data on neutropenia admissions in New South Wales (NSW), Australia.

Lung Cancer Mortality in Tuscany from 1971 to 2010 and Its Connections with Silicosis: A Space-Cohort Analysis Based on Shared Models.

Lung cancer mortality in Tuscany (Italy) for males, from 1971 and 2010, is investigated. A hierarchical Bayesian model for space-time disease mapping is introduced. Such a model belongs to the class of shared random effect models and exploits the birth-cohort as the relevant time dimension. It allows for highlighting common and specific patterns of risk for each birth-cohort. The results show that different birth-cohorts exhibit quite different spatial patterns, even if the socioeconomic status is taken into account. In fact, there were different occupational exposures before and after the Second World War. The birth-cohort 1930–35 exhibits high relative risks related to particular areas. This fact could be connected with occupational exposure to risk factors for silicosis, perhaps a prognostic status for lung cancer.

A More Robust Random Effects Model for Disease Mapping

Disease mapping studies have found wide applications within geographical epidemiology and public health and are typically analysed within a Bayesian hierarchical model formulation. The most popular disease mapping model is the Besag-York-Molli´e model. A distinguishing feature of this model is the use of two sets of random effects: one spatially structured to model spatial autocorrelation and the other spatially unstructured to describe residual unstructured heterogeneity. Very often the spatially unstructured random effect is assumed to be normally distributed. Under practical situations, this normality assumption is found to be over restrictive. In this study, we investigate a more robust spatially unstructured random effect distribution by considering the Inverse Gaussian (IG) distribution in the disease mapping problem. The distribution has the normal distribution as special case. The inferences under this model are carried out within a bayesian hierarchical model formulation implemented in WinBUGS. The IG distribution is introduced in WinBUGS using zero tricks. The usefulness of the proposed model is investigated with a simulation study and applied in real data; mapping HIV in Kenya. In this work we showed that the IG distribution can produce better results when the normality assumption is violated due to the skewness of the data. For the case of data in which the random effects are truly normal, the IG distribution adjusts to a normal distribution as dictated by the data itself. On the other hand, the spatially structured random effect is normally modelled using the intrinsic conditional autoregressive (iCAR) prior. This prior is improper and has the undesirable large scale property of leading to a negative pairwise correlation for regions located further apart. In addition, the BYM model presents some identifiability problems of the spatial and non-spatial effects. In this work, we consider Leroux CAR (named lCAR hereafter) prior, a less widely used prior in disease mapping, as the prior distribution for the spatially structured random effects.

Spatio-Temporal Variation of HIV Infection in Kenya

Disease mapping is the study of the distribution of disease relative risks or rates in space and time, and normally uses generalized linear mixed models (GLMMs) which includes fixed effects and spatial, temporal, and spatio-temporal random effects. Model fitting and statistical inference are commonly accomplished through the empirical Bayes (EB) and fully Bayes (FB) approaches. The EB approach usually relies on the penalized quasi-likelihood (PQL), while the FB approach, which has increasingly become more popular in the recent past, usually uses Markov chain Monte Carlo (McMC) techniques. However, there are many challenges in conventional use of posterior sampling via McMC for inference. This includes the need to evaluate convergence of posterior samples, which often requires extensive simulation and can be very time consuming. Spatio-temporal models used in disease mapping are often very complex and McMC methods may lead to large Monte Carlo errors if the dimension of the data at hand is large. To address these challenges, a new strategy based on integrated nested Laplace approximations (INLA) has recently been recently developed as a promising alternative to the McMC. This technique is now becoming more popular in disease mapping because of its ability to fit fairly complex space-time models much more quickly than the McMC. In this paper, we show how to fit different spatio-temporal models for disease mapping with INLA using the Leroux CAR prior for the spatial component, and we compare it with McMC using Kenya HIV incidence data during the period 2013-2016.

Estimating relative risk for dengue disease in Peninsular Malaysia using INLA

Study in spatio-temporal disease mapping models give a great worth in epidemiology, in describing the pattern of disease incidence across geographical space and time. This paper studies generalized linear mixed models (GLMM) for the analysis of spatial and temporal variability of dengue disease rates. For spatio-temporal study, the models accommodate spatially correlated random effects as well as temporal effects together with the space time interaction. The space time interaction is used to capture any additional effects that are not explained by the main factors of space and time. However, as study including time dimension is quite complex for disease mapping, the temporal effects that only relate to structured and unstructured time pattern are considered in these models as initial screening in studying disease pattern and time trend. The models are fitted within a hierarchical Bayesian framework using Integrated Nested Laplace Approximation (INLA) methodology. For this study, there are three main objectives. First, to choose the best model that represent the disease phenomenon. Second, to estimate the relative risk of disease based on the model selected and lastly, to visualize the risk spatial pattern and temporal trend using graphical representation. The models are applied to monthly dengue fever data in Peninsular Malaysia reported to Ministry of Health Malaysia for year 2015 by district level.

Relative risk estimation of dengue disease at small spatial scale

BackgroundDengue is a high incidence arboviral disease in tropical countries around the world. Colombia is an endemic country due to the favourable environmental conditions for vector survival and spread. Dengue surveillance in Colombia is based in passive notification of cases, supporting monitoring, prediction, risk factor identification and intervention measures. Even though the surveillance network works adequately, disease mapping techniques currently developed and employed for many health problems are not widely applied. We select the Colombian city of Bucaramanga to apply Bayesian areal disease mapping models, testing the challenges and difficulties of the approach.MethodsWe estimated the relative risk of dengue disease by census section (a geographical unit composed approximately by 1–20 city blocks) for the period January 2008 to December 2015. We included the covariates normalized difference vegetation index (NDVI) and land surface temperature (LST), obtained by satellite images. We fitted Bayesian areal models at the complete period and annual aggregation time scales for 2008–2015, with fixed and space-varying coefficients for the covariates, using Markov Chain Monte Carlo simulations. In addition, we used Cohen’s Kappa agreement measures to compare the risk from year to year, and from every year to the complete period aggregation.ResultsWe found the NDVI providing more information than LST for estimating relative risk of dengue, although their effects were small. NDVI was directly associated to high relative risk of dengue. Risk maps of dengue were produced from the estimates obtained by the modeling process. The year to year risk agreement by census section was sligth to fair.ConclusionThe study provides an example of implementation of relative risk estimation using Bayesian models for disease mapping at small spatial scale with covariates. We relate satellite data to dengue disease, using an areal data approach, which is not commonly found in the literature. The main difficulty of the study was to find quality data for generating expected values as input for the models. We remark the importance of creating population registry at small spatial scale, which is not only relevant for the risk estimation of dengue but also important to the surveillance of all notifiable diseases.

In spatio-temporal disease mapping models, identifiability constraints affect PQL and INLA results

Disease mapping studies the distribution of relative risks or rates in space and time, and typically relies on generalized linear mixed models (GLMMs) including fixed effects and spatial, temporal, and spatio-temporal random effects. These GLMMs are typically not identifiable and constraints are required to achieve sensible results. However, automatic specification of constraints can sometimes lead to misleading results. In particular, the penalized quasi-likelihood fitting technique automatically centers the random effects even when this is not necessary. In the Bayesian approach, the recently-introduced integrated nested Laplace approximations computing technique can also produce wrong results if constraints are not well-specified. In this paper the spatial, temporal, and spatio-temporal interaction random effects are reparameterized using the spectral decompositions of their precision matrices to establish the appropriate identifiability constraints. Breast cancer mortality data from Spain is used to illustrate the ideas.

Estimating cross-validatory predictive p-values with integrated importance sampling for disease mapping models.

An important statistical task in disease mapping problems is to identify divergent regions with unusually high or low risk of disease. Leave-one-out cross-validatory (LOOCV) model assessment is the gold standard for estimating predictive p-values that can flag such divergent regions. However, actual LOOCV is time-consuming because one needs to rerun a Markov chain Monte Carlo analysis for each posterior distribution in which an observation is held out as a test case. This paper introduces a new method, called integrated importance sampling (iIS), for estimating LOOCV predictive p-values with only Markov chain samples drawn from the posterior based on a full data set. The key step in iIS is that we integrate away the latent variables associated the test observation with respect to their conditional distribution without reference to the actual observation. By following the general theory for importance sampling, the formula used by iIS can be proved to be equivalent to the LOOCV predictive p-value. We compare iIS and other three existing methods in the literature with two disease mapping datasets. Our empirical results show that the predictive p-values estimated with iIS are almost identical to the predictive p-values estimated with actual LOOCV and outperform those given by the existing three methods, namely, the posterior predictive checking, the ordinary importance sampling, and the ghosting method by Marshall and Spiegelhalter (2003). Copyright © 2017 John Wiley & Sons, Ltd.

Zero-inflated spatio-temporal models for disease mapping.

In this paper, our aim is to analyze geographical and temporal variability of disease incidence when spatio-temporal count data have excess zeros. To that end, we consider random effects in zero-inflated Poisson models to investigate geographical and temporal patterns of disease incidence. Spatio-temporal models that employ conditionally autoregressive smoothing across the spatial dimension and B-spline smoothing over the temporal dimension are proposed. The analysis of these complex models is computationally difficult from the frequentist perspective. On the other hand, the advent of the Markov chain Monte Carlo algorithm has made the Bayesian analysis of complex models computationally convenient. Recently developed data cloning method provides a frequentist approach to mixed models that is also computationally convenient. We propose to use data cloning, which yields to maximum likelihood estimation, to conduct frequentist analysis of zero-inflated spatio-temporal modeling of disease incidence. One of the advantages of the data cloning approach is that the prediction and corresponding standard errors (or prediction intervals) of smoothing disease incidence over space and time is easily obtained. We illustrate our approach using a real dataset of monthly children asthma visits to hospital in the province of Manitoba, Canada, during the period April 2006 to March 2010. Performance of our approach is also evaluated through a simulation study.

Disease mapping of zero-excessive mesothelioma data in Flanders

PurposeTo investigate the distribution of mesothelioma in Flanders using Bayesian disease mapping models that account for both an excess of zeros and overdispersion. MethodsThe numbers of newly diagnosed mesothelioma cases within all Flemish municipalities between 1999 and 2008 were obtained from the Belgian Cancer Registry. To deal with overdispersion, zero inflation, and geographical association, the hurdle combined model was proposed, which has three components: a Bernoulli zero-inflation mixture component to account for excess zeros, a gamma random effect to adjust for overdispersion, and a normal conditional autoregressive random effect to attribute spatial association. This model was compared with other existing methods in literature. ResultsThe results indicate that hurdle models with a random effects term accounting for extra variance in the Bernoulli zero-inflation component fit the data better than hurdle models that do not take overdispersion in the occurrence of zeros into account. Furthermore, traditional models that do not take into account excessive zeros but contain at least one random effects term that models extra variance in the counts have better fits compared to their hurdle counterparts. In other words, the extra variability, due to an excess of zeros, can be accommodated by spatially structured and/or unstructured random effects in a Poisson model such that the hurdle mixture model is not necessary. ConclusionsModels taking into account zero inflation do not always provide better fits to data with excessive zeros than less complex models. In this study, a simple conditional autoregressive model identified a cluster in mesothelioma cases near a former asbestos processing plant (Kapelle-op-den-Bos). This observation is likely linked with historical local asbestos exposures. Future research will clarify this.

The spatio-temporal relationship between alcohol outlets and violence before and after privatization: A natural experiment, Seattle, Wa 2010–2013

Alcohol-related violence is a well-documented public health concern, where various individual and community-level factors contribute to this relationship. The purpose of this study is to examine the impact of a significant policy change at the local level, which privatized liquor sales and distribution. Specifically, we explored the relationship between alcohol and violence in Seattle, WA, 2010–2013, via hierarchical spatio-temporal disease mapping models. To measure and map this complex spatio-temporal relationship at the census block group level (n=567), we examined a variety of models using integrated nested Laplace approximations and used the deviance information criterion to gauge model complexity and fit. For each additional off-premises and on-premises alcohol outlet in a given census block group, we found a significant increase of 8% and 5% for aggravated assaults and 6% and 5% for non-aggravated assaults, respectively. Lastly, our maps showed variation in the estimated relative risks across the city of Seattle.

An intuitive Bayesian spatial model for disease mapping that accounts for scaling.

In recent years, disease mapping studies have become a routine application within geographical epidemiology and are typically analysed within a Bayesian hierarchical model formulation. A variety of model formulations for the latent level have been proposed but all come with inherent issues. In the classical BYM (Besag, York and Mollié) model, the spatially structured component cannot be seen independently from the unstructured component. This makes prior definitions for the hyperparameters of the two random effects challenging. There are alternative model formulations that address this confounding; however, the issue on how to choose interpretable hyperpriors is still unsolved. Here, we discuss a recently proposed parameterisation of the BYM model that leads to improved parameter control as the hyperparameters can be seen independently from each other. Furthermore, the need for a scaled spatial component is addressed, which facilitates assignment of interpretable hyperpriors and make these transferable between spatial applications with different graph structures. The hyperparameters themselves are used to define flexible extensions of simple base models. Consequently, penalised complexity priors for these parameters can be derived based on the information-theoretic distance from the flexible model to the base model, giving priors with clear interpretation. We provide implementation details for the new model formulation which preserve sparsity properties, and we investigate systematically the model performance and compare it to existing parameterisations. Through a simulation study, we show that the new model performs well, both showing good learning abilities and good shrinkage behaviour. In terms of model choice criteria, the proposed model performs at least equally well as existing parameterisations, but only the new formulation offers parameters that are interpretable and hyperpriors that have a clear meaning.

Individual level covariate adjusted conditional autoregressive (indiCAR) model for disease mapping.

BackgroundMapping disease rates over a region provides a visual illustration of underlying geographical variation of the disease and can be useful to generate new hypotheses on the disease aetiology. However, methods to fit the popular and widely used conditional autoregressive (CAR) models for disease mapping are not feasible in many applications due to memory constraints, particularly when the sample size is large. We propose a new algorithm to fit a CAR model that can accommodate both individual and group level covariates while adjusting for spatial correlation in the disease rates, termed indiCAR. Our method scales well and works in very large datasets where other methods fail.ResultsWe evaluate the performance of the indiCAR method through simulation studies. Our simulation results indicate that the indiCAR provides reliable estimates of all the regression and random effect parameters. We also apply indiCAR to the analysis of data on neutropenia admissions in New South Wales (NSW), Australia. Our analyses reveal that lower rates of neutropenia admissions are significantly associated with individual level predictors including higher age, male gender, residence in an outer regional area and a group level predictor of social disadvantage, the socio-economic index for areas. A large value for the spatial dependence parameter is estimated after adjusting for individual and area level covariates. This suggests the presence of important variation in the management of cancer patients across NSW.ConclusionsIncorporating individual covariate data in disease mapping studies improves the estimation of fixed and random effect parameters by utilizing information from multiple sources. Health registries routinely collect individual and area level information and thus could benefit by using indiCAR for mapping disease rates. Moreover, the natural applicability of indiCAR in a distributed computing framework enhances its application in the Big Data domain with a large number of individual/group level covariates. CI NSW Study Reference Number: 2012/07/410. Dated: July 2012.

On the specification of prior distributions for variance components in disease mapping models

In this paper, we consider the problem of specifying priors for the variance components in the Bayesian analysis of the Besag-York-Mollie model, a model that is popular among epidemiologists for disease mapping. The model encompasses two sets of random effects: one spatially structured to model spatial autocorrelation and the other spatially unstructured to describe residual heterogeneity. In this model, prior specification for variance components is an important problem because these priors maintain their influence on the posterior distributions of relative risks when mapping rare diseases. We propose using generalised inverse Gaussian priors, a broad class of distributions that includes many distributions commonly used as priors in this context, such as inverse gammas. We discuss the prior parameter choice with the aim of balancing the prior weight of the two sets of random effects on total variation and controlling the amount of shrinkage. The suggested prior specification strategy is compared to popular alternatives using a simulation exercise and applications to real data sets.

Evidence for Community Transmission of Community-Associated but Not Health-Care-Associated Methicillin-Resistant Staphylococcus Aureus Strains Linked to Social and Material Deprivation: Spatial Analysis of Cross-sectional Data.

BackgroundIdentifying and tackling the social determinants of infectious diseases has become a public health priority following the recognition that individuals with lower socioeconomic status are disproportionately affected by infectious diseases. In many parts of the world, epidemiologically and genotypically defined community-associated (CA) methicillin-resistant Staphylococcus aureus (MRSA) strains have emerged to become frequent causes of hospital infection. The aim of this study was to use spatial models with adjustment for area-level hospital attendance to determine the transmission niche of genotypically defined CA- and health-care-associated (HA)-MRSA strains across a diverse region of South East London and to explore a potential link between MRSA carriage and markers of social and material deprivation.Methods and FindingsThis study involved spatial analysis of cross-sectional data linked with all MRSA isolates identified by three National Health Service (NHS) microbiology laboratories between 1 November 2011 and 29 February 2012. The cohort of hospital-based NHS microbiology diagnostic services serves 867,254 usual residents in the Lambeth, Southwark, and Lewisham boroughs in South East London, United Kingdom (UK). Isolates were classified as HA- or CA-MRSA based on whole genome sequencing. All MRSA cases identified over 4 mo within the three-borough catchment area (n = 471) were mapped to small geographies and linked to area-level aggregated socioeconomic and demographic data. Disease mapping and ecological regression models were used to infer the most likely transmission niches for each MRSA genetic classification and to describe the spatial epidemiology of MRSA in relation to social determinants. Specifically, we aimed to identify demographic and socioeconomic population traits that explain cross-area extra variation in HA- and CA-MRSA relative risks following adjustment for hospital attendance data. We explored the potential for associations with the English Indices of Deprivation 2010 (including the Index of Multiple Deprivation and several deprivation domains and subdomains) and the 2011 England and Wales census demographic and socioeconomic indicators (including numbers of households by deprivation dimension) and indicators of population health. Both CA-and HA-MRSA were associated with household deprivation (CA-MRSA relative risk [RR]: 1.72 [1.03–2.94]; HA-MRSA RR: 1.57 [1.06–2.33]), which was correlated with hospital attendance (Pearson correlation coefficient [PCC] = 0.76). HA-MRSA was also associated with poor health (RR: 1.10 [1.01–1.19]) and residence in communal care homes (RR: 1.24 [1.12–1.37]), whereas CA-MRSA was linked with household overcrowding (RR: 1.58 [1.04–2.41]) and wider barriers, which represent a combined score for household overcrowding, low income, and homelessness (RR: 1.76 [1.16–2.70]). CA-MRSA was also associated with recent immigration to the UK (RR: 1.77 [1.19–2.66]). For the area-level variation in RR for CA-MRSA, 28.67% was attributable to the spatial arrangement of target geographies, compared with only 0.09% for HA-MRSA. An advantage to our study is that it provided a representative sample of usual residents receiving care in the catchment areas. A limitation is that relationships apparent in aggregated data analyses cannot be assumed to operate at the individual level.ConclusionsThere was no evidence of community transmission of HA-MRSA strains, implying that HA-MRSA cases identified in the community originate from the hospital reservoir and are maintained by frequent attendance at health care facilities. In contrast, there was a high risk of CA-MRSA in deprived areas linked with overcrowding, homelessness, low income, and recent immigration to the UK, which was not explainable by health care exposure. Furthermore, areas adjacent to these deprived areas were themselves at greater risk of CA-MRSA, indicating community transmission of CA-MRSA. This ongoing community transmission could lead to CA-MRSA becoming the dominant strain types carried by patients admitted to hospital, particularly if successful hospital-based MRSA infection control programmes are maintained. These results suggest that community infection control programmes targeting transmission of CA-MRSA will be required to control MRSA in both the community and hospital. These epidemiological changes will also have implications for effectiveness of risk-factor-based hospital admission MRSA screening programmes.

A unifying modeling framework for highly multivariate disease mapping.

Multivariate disease mapping refers to the joint mapping of multiple diseases from regionally aggregated data and continues to be the subject of considerable attention for biostatisticians and spatial epidemiologists. The key issue is to map multiple diseases accounting for any correlations among themselves. Recently, Martinez-Beneito (2013) provided a unifying framework for multivariate disease mapping. While attractive in that it colligates a variety of existing statistical models for mapping multiple diseases, this and other existing approaches are computationally burdensome and preclude the multivariate analysis of moderate to large numbers of diseases. Here, we propose an alternative reformulation that accrues substantial computational benefits enabling the joint mapping of tens of diseases. Furthermore, the approach subsumes almost all existing classes of multivariate disease mapping models and offers substantial insight into the properties of statistical disease mapping models.

STANOVA: a smoothed-ANOVA-based model for spatio-temporal disease mapping

Spatio-temporal disease mapping can be viewed as a multivariate disease mapping problem with a given order of the geographic patterns to be studied. As a consequence, some of the techniques in multivariate literature could also be used to build spatio-temporal models. In this paper we propose using the smoothed ANOVA multivariate model for spatio-temporal problems. Under our approach the time trend for each geographic unit is modeled parametrically, projecting it on a preset orthogonal basis of functions (the contrasts in the smoothed ANOVA nomenclature), while the coefficients of these projections are considered to be spatially dependent random effects. Despite the parametric temporal nature of our proposal, we show with both simulated and real datasets that it may be as flexible as other spatio-temporal smoothing models proposed in the literature and may model spatio-temporal data with several sources of variability.

On fitting spatio-temporal disease mapping models using approximate Bayesian inference

Spatio-temporal disease mapping comprises a wide range of models used to describe the distribution of a disease in space and its evolution in time. These models have been commonly formulated within a hierarchical Bayesian framework with two main approaches: an empirical Bayes (EB) and a fully Bayes (FB) approach. The EB approach provides point estimates of the parameters relying on the well-known penalized quasi-likelihood (PQL) technique. The FB approach provides the posterior distribution of the target parameters. These marginal distributions are not usually available in closed form and common estimation procedures are based on Markov chain Monte Carlo (MCMC) methods. However, the spatio-temporal models used in disease mapping are often very complex and MCMC methods may lead to large Monte Carlo errors and a huge computation time if the dimension of the data at hand is large. To circumvent these potential inconveniences, a new technique called integrated nested Laplace approximations (INLA), based on nested Laplace approximations, has been proposed for Bayesian inference in latent Gaussian models. In this paper, we show how to fit different spatio-temporal models for disease mapping with INLA using the Leroux CAR prior for the spatial component, and we compare it with PQL via a simulation study. The spatio-temporal distribution of male brain cancer mortality in Spain during the period 1986-2010 is also analysed.