The Joint Impact of Temperature and Fine Particles on Mortality in the Metropolitan Area of Florence (Italy): A Causal Inference Analysis Based on G-Computation

Abstract

Introduction

Air pollution and extreme temperatures are among the leading environmental determinants of premature mortality. Although the short-term associations between particulate matter with an aerodynamic diameter smaller than 10 μm (PM₁₀), temperature, and mortality have been extensively documented, the nature of their joint effects remains largely unexplored.

Aims

Conceptualizing both temperature and air pollution as modifiable treatments within a potential outcome (PO) approach to casual inference, we aimed to: (1) estimate the joint causal effect of particulate matter ≤10 µg/m³ (PM₁₀) and temperature on natural mortality in the metropolitan area of Florence (2008–2019) within the potential outcomes framework, and (2) quantify the attributable fraction among the exposed under a joint counterfactual reduction scenario.

Methods

Data

Daily time series data on air pollution and mortality were collected from 2008 to 2019 for an area including eight municipalities in Tuscany: Bagno a Ripoli, Calenzano, Campi Bisenzio, Florence, Lastra a Signa, Scandicci, Sesto Fiorentino, and Signa (597740 inhabitants in 2019). Mortality data, sourced from mortality registers, encompassed all natural causes, excluding the external ones. Air quality data were obtained from the Tuscany Regional Environmental Protection Agency (ARPA), while meteorological data including temperature and relative humidity were provided by the Tuscany Regional Hydrological Service (SIR) and the Laboratory of Monitoring and Environmental Modelling for Sustainable Development (LaMMA Consortium).

Statistical analysis

Daily time-series data on mortality, PM₁₀, and temperature were analyzed using g-computation under a bivariate treatment framework. POs were defined for joint exposure levels of PM₁₀ and temperature under assumptions of no unmeasured confounding, consistency, and positivity formulated for the two-dimensional exposure. A bi-dimensional average dose-response function (aDRF) was estimated to characterize how the joint variation in PM₁₀ and temperature influences mortality. Attributable deaths (AD) and attributable fractions among the exposed (AF) were computed by comparing observed exposure levels with a counterfactual scenario defined by PM₁₀ = 15 µg/m³ (WHO 2021 annual guideline) and temperature = 25°C.

Results

The estimated dose-response surface showed evidence of interaction between PM₁₀ and temperature, with amplified mortality risks at higher temperatures. Under the joint reduction scenario, 919 deaths were attributable to the combined effect of elevated pollution and temperature, corresponding to an AF of 9.45%. The AF was substantially larger under concurrent high-temperature and high-PM₁₀ conditions compared with the reduction of one exposure only (AF = 3.66% for PM₁₀ alone and 5.53% for temperature alone), supporting the presence of synergistic effects.

Conclusions

This study provides causal evidence of a synergistic short-term effect of PM₁₀ and temperature on mortality. The estimated attributable fractions highlight the public health relevance of joint exposure reduction strategies, particularly in the context of climate change and increasing heat events.