Background: Reliable estimation of the short-term effects of PM2.5 on mortality requires flexible modeling of the exposure–response function and a clear causal interpretation. Within a formal causal framework, we extended a recently proposed Generalized Propensity Score (GPS) matching approach to estimate the average dose–response function and to quantify attributable deaths under counterfactual reduction scenarios aligned with World Health Organization (WHO) guidelines and European Union (EU) standards. The method was applied to the metropolitan area surrounding Florence, Italy, for the period 2010–2019. Methods: The exposure range was partitioned into quantile-based intervals, and potential outcomes at the median exposure of each interval were imputed using one-to-one matching, pairing each day with the observed day that minimized a bivariate Manhattan distance based on GPS and exposure. The average dose-response function (aDRF) was estimated by interpolating the imputed outcomes with a monotone increasing spline. Attributable deaths were computed from the attributable fraction derived from the estimated aDRF. Bootstrap resampling provided 90\% confidence intervals for all estimates. Results: The aDRF showed positive associations between PM2.5 and natural and cardiovascular mortality, with a steeper slope at lower concentrations for natural mortality. The curve for respiratory mortality was flatter and imprecise. The maximum impact was observed when reducing daily exposures above 5 mug/m^3 to that threshold (WHO annual limit), preventing 2,876 natural deaths (90% CI: 599–19,695), of which 1,527 (90% CI: 426–9,869) were cardiovascular. Reducing daily exposures to exactly 10 mug/m^3 (EU annual limit) would have prevented 2,425 natural deaths (90% CI: 432–4,525). Respecting the WHO daily guideline of 15 mug/m^3 would have prevented 1,300 natural deaths (90% CI: 92–2,199) when exceedances were fixed to the threshold, and 1,583 deaths (90% CI: 212–3,319) when counterfactual exposures were sampled from the observed distribution below the threshold. Conclusions: Exposure to PM2.5 substantially contributed to natural and cardiovascular mortality in the study area. Despite limitations mainly related to the high uncertainty of the estimates, the proposed method shows strong potential for broader application in environmental epidemiology, enabling transparent estimation of health impacts under realistic counterfactual scenarios.

Estimating the short-term PM2.5-mortality relationship and attributable deaths under alternative scenarios using a novel approach based on GPS matching: application to the metropolitan area of Florence, Italy (2010–2019) / Giulio Biscardi, C.M.. - In: ENVIRONMENTAL HEALTH. - ISSN 1476-069X. - STAMPA. - (In corso di stampa), pp. 1-22.

Estimating the short-term PM2.5-mortality relationship and attributable deaths under alternative scenarios using a novel approach based on GPS matching: application to the metropolitan area of Florence, Italy (2010–2019)

Giulio Biscardi
;
Chiara Marzi;Alessandra Mattei;Michela Baccini
In corso di stampa

Abstract

Background: Reliable estimation of the short-term effects of PM2.5 on mortality requires flexible modeling of the exposure–response function and a clear causal interpretation. Within a formal causal framework, we extended a recently proposed Generalized Propensity Score (GPS) matching approach to estimate the average dose–response function and to quantify attributable deaths under counterfactual reduction scenarios aligned with World Health Organization (WHO) guidelines and European Union (EU) standards. The method was applied to the metropolitan area surrounding Florence, Italy, for the period 2010–2019. Methods: The exposure range was partitioned into quantile-based intervals, and potential outcomes at the median exposure of each interval were imputed using one-to-one matching, pairing each day with the observed day that minimized a bivariate Manhattan distance based on GPS and exposure. The average dose-response function (aDRF) was estimated by interpolating the imputed outcomes with a monotone increasing spline. Attributable deaths were computed from the attributable fraction derived from the estimated aDRF. Bootstrap resampling provided 90\% confidence intervals for all estimates. Results: The aDRF showed positive associations between PM2.5 and natural and cardiovascular mortality, with a steeper slope at lower concentrations for natural mortality. The curve for respiratory mortality was flatter and imprecise. The maximum impact was observed when reducing daily exposures above 5 mug/m^3 to that threshold (WHO annual limit), preventing 2,876 natural deaths (90% CI: 599–19,695), of which 1,527 (90% CI: 426–9,869) were cardiovascular. Reducing daily exposures to exactly 10 mug/m^3 (EU annual limit) would have prevented 2,425 natural deaths (90% CI: 432–4,525). Respecting the WHO daily guideline of 15 mug/m^3 would have prevented 1,300 natural deaths (90% CI: 92–2,199) when exceedances were fixed to the threshold, and 1,583 deaths (90% CI: 212–3,319) when counterfactual exposures were sampled from the observed distribution below the threshold. Conclusions: Exposure to PM2.5 substantially contributed to natural and cardiovascular mortality in the study area. Despite limitations mainly related to the high uncertainty of the estimates, the proposed method shows strong potential for broader application in environmental epidemiology, enabling transparent estimation of health impacts under realistic counterfactual scenarios.
In corso di stampa
1
22
Giulio Biscardi, Chiara Marzi, Daniela Nuvolone, Alessandra Mattei, Michela Baccini
File in questo prodotto:
File Dimensione Formato  
paper_Florence_ENV_HEALTH.pdf

Accesso chiuso

Descrizione: Articolo
Tipologia: Versione finale referata (Postprint, Accepted manuscript)
Licenza: Tutti i diritti riservati
Dimensione 1.54 MB
Formato Adobe PDF
1.54 MB Adobe PDF   Richiedi una copia all'autore

I documenti in FLORE sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificatore per citare o creare un link a questa risorsa: https://hdl.handle.net/2158/1500760
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact