Ambient air pollution and stroke risk among different accumulation patterns of comorbidities: Casual inference study based on marginal ‘between-within’ model

Abstract

Previous studies showed that ambient air pollution is a risk factor for stroke. But its effects on patients with different comorbidity accumulation patterns and the causal relationship remain unclear. Hospitalization data were collected from the Beijing Municipal Health Commission Information Center. Pollutant data, including particulate matter (PM_2.5, PM₁₀, PM₁, PM_1-2.5), sulfur dioxide (SO₂), nitrogen dioxide (NO₂), carbon monoxide (CO), and ozone (O₃) from 2014 to 2017, were obtained from the widely-used ChinaHighAirPollutants dataset. An individual-level case-crossover design was employed to identify case days and control days. The exposure-response association was estimated by fitting conditional logistic regression models. The newly proposed marginal ‘between-within’ models were used to estimate counterfactual hospitalization probability. In a total of 237 487 stroke patients, 211 741 diagnosed with ischemic stroke (IS) and 25 641 with hemorrhagic stroke (HS). Except for SO₂, the seven pollutants were associated with higher risk of overall stroke and IS. Particulate matter showed larger estimated effects in patients with hypertension and diabetes but without hyperlipidemia. NO₂ and CO showed larger estimated effects in patients with hypertension but without diabetes. O₃ showed larger estimated effects in patients with hypertension, diabetes, and hyperlipidemia. PM₁, PM₁₀, NO₂, and CO had statistically significant and persistent causal relationships with stroke risk during the lag periods. This study highlighted the need for targeted interventions of air pollution. Prioritizing control measures for PM₁, PM₁₀, NO₂, and CO is particularly crucial in stroke prevention efforts.