A Multidimensional Analysis Framework for the Study of Contaminated Sites: Construction of Composite Indexes and their Association with Mortality

Abstract

INTRODUCTION
Contaminated Sites of National interest (SINs) represent highly complex settings in which environmental, demographic, socioeconomic, and productive factors interact, influencing the health status of resident populations. In these areas, consistently with the One Health paradigm, health results from the interac-tion among determinants belonging to different and closely interconnected domains. Given this complexity, studying the health status of populations living in SINs requires multidimen-sional analytical approaches capable of integrating heteroge-neous information and capturing the systemic nature of these territories. Consequently, there is a need to adopt statistical methodologies able to simultaneously represent the multiple dimensions characterizing these contexts, overcoming the lim-itations of analyses based on single indicators, which describe only isolated aspects of the phenomenon under study.

OBJECTIVES
This study proposes a methodological framework for the multidimensional characterization of SINs, based on the con-struction of composite indexes capable of integrating both the spatial component of the data and the multiple dimensions characterizing these contexts. The aim is to identify the main latent dimensions of the territory and assess their association with all-cause and cause-specific mortality, providing useful tools for epidemiological surveillance and public health de-cision-making.

METHODS
The approach is applied to the Sulcis-Iglesiente-Gus-pinese and Porto Torres SINs in Sardinia, using municipalities as the territorial unit of analysis. A set of indicators derived from institutional data sources describes demographic char-acteristics, environmental pressure, level of industrialization, socioeconomic vulnerability, and land use. Statistical techincs are employed to reduce the dimensionality of the dataset and construct composite indicators representing the main latent ter-ritorial dimensions. These synthetic indexes are subsequently incorporated into regression models to investigate their asso-ciation with all-cause mortality and specific causes of death. The analyses are further complemented by spatial statistical techniques aimed at evaluating the geographical distribution of territorial components and the role of spatial dependence in interpreting relationships between territory and health.

RESULTS
The proposed approach makes it possible to reduce and summarize a large number of variables into a limited set of composite indicators capable of representing the main struc-tural characteristics of the territories, while reducing problems related to information redundancy and multicollinearity. The integration of regression models and spatial analysis tech-niques provides deeper insight into the role of different terri-torial dimensions in the distribution of mortality and highlights potential geographical patterns of epidemiological interest.

CONCLUSIONS
The proposed framework integrates dimensionality reduc-tion methods, spatial analysis, and environmental epidemiol-ogy into a unified approach for studying territorial determi-nants of health. The methodology is flexible and replicable in other contexts characterized by high environmental complex-ity, contributing to the development of tools that support ep-idemiological surveillance and the planning of public health interventions according to the One Health paradigm.