Our research covers a wide range of topics, including: methodologies, global health modelling, climate change and health, air pollution, and spatio-temporal modelling.

Focus areas: 

Statistical Methodologies

Distributed lag linear and non-linear models

Distributed lag models (DLMs) represent an elegant methodology for describing lagged association in time series data. Originally developed in econometrics, they are now frequently used in epidemiological analysis. Pioneering work by Ben Armstrong extended them to distributed lag non-linear models (DLNMs) for non-linear temperature-mortality relationships.

We proposed a unified statistical framework for the DLM/DLNM class, based on the definition of a cross-basis, a bi-dimensional space of functions that describes the association simultaneously along the space of predictor and lag. Later, we generalised the methodology beyond time series data, allowing applications in various epidemiological fields. Finally, we extended the framework to penalised DLNM implemented through generalized additive models (GAM). The DLM/DLNM methodology is implemented in the R package dlnm.

 

An extended meta-analytical framework

Standard methods for meta-analysis are limited to pooling associations represented by a single effect size estimated from a set of independent studies. However, this setting can be too restrictive for modern meta-analytical applications.

 We first contributed to developing multivariate meta-analytical methods for pooling multiparameter estimates representing complex associations. We then developed a general framework for meta-analysis based on linear mixed-effects models that includes, as special cases, multivariate, network, multilevel, dose-response, and longitudinal meta-analysis and meta-regression. This meta-analytical development has been applied in extended two-stage designs for environmental epidemiology. The methodology has been implemented in the R packages mvmeta and mixmeta.

 

Study Designs

The case time series design

Modern linkage methods and data technologies provide a way to reconstruct detailed longitudinal profiles of health outcomes and predictors. This rich data setting, however, poses important methodological and computational problems that traditional epidemiological methods are not well suited to address.

We have developed the case time series (CTS) design, a novel methodology that combines the longitudinal structure typical of aggregated time series with the individual-level self-matched methods. The modelling framework is highly adaptable to various outcome and exposure definitions, and it is based on efficient methods that make it suitable for the analysis of highly informative longitudinal data resources.

The main article introduced the CTS design and illustrated applications in case studies using environmental and clinical data. A following tutorial article adapted the CTS methodology for the analysis of small-area data.

 

Small-area analysis of environmental risks

The increased availability of data on health outcomes and risk factors collected at fine geographical resolution makes possible conducting small-area epidemiological studies. However, this setting poses important methodological and computational issues, related to modelling complexities and data linkage. 

The EHM-Lab has developed cutting-edge study designs for the analysis of small-area data. These methods allow the use of finely disaggregated health data linked with high-resolution environmental exposure measurements through GIS techniques. The framework offers the opportunity to study local variations in risk and the role of area-level characteristics in modifying the vulnerability to environmental stressors. We provided a methodological description of the design in a tutorial article, and an application to study small-area temperature-related risks.

 

Extensions of two-stage designs

The two-stage design has become a standard tool in environmental epidemiology to model multi-location data. We have recently proposed multiple design extensions of the classical two-stage design structure, all implemented within a unified analytical framework based on linear mixed-effects models.

 

 

The extended two-stage methodology, described in a recent tutorial article, permits the analysis of associations characterised by combinations of multivariate outcomes, hierarchical geographical structures, repeated measures, and/or longitudinal settings. We have applied it in various epidemiological analyses, including for modelling complex associations with temperature, investigating air pollution effects clustered at multiple geographical levels, assessing differential risks by age and geographical areas, quantifying excess mortality during the COVID-19 outbreak, and estimating the role of air conditioning in attenuating heat-related mortality.

 

Interrupted time series design

Interrupted time series (ITS) analysis is a valuable study design for evaluating the public health interventions. Its quasi-experimental nature allows quantifying effects of policies or events using a pre-post comparison while controlling for temporal trends.

We first illustrated the application of the ITS method for epidemiological analysis in a tutorial article that discussed design features and assumptions. Specific methodological contributions focused instead on model selection and the use of controls. The EHM-Lab has contributed to several applications of the ITS design, for instance for assessing the association between smoking bans and cardiovascular risk, the effect of the financial crisis on suicides, the impact of media coverage on the use of statins, the relationship between self-defence laws on firearm-related homicides, the effect of taxation on the sales of sugar-sweetened beverages, the impact of healthcare reforms on hospital care, and the excess mortality during the COVID-19 outbreak.

 

The MCC Study

We lead the Multi-Country Multi-City (MCC) Collaborative Research Network, an international collaboration of research teams producing epidemiological evidence on associations between environmental stressors, climate, and health.

Climate and health

Further details coming soon

Air pollution

Further details coming soon

Spatio-temporal modelling

Further details coming soon