Objective: There are well-known biological pathways in which air pollution can contribute to Parkinson’s disease (PD) pathology. We examined the effect of exposure to common air pollutants: particulate matter ≤2.5 µm (PM_2.5), nitrogen dioxide (NO₂), and ozone (O₃), on PD hospitalization in the same year among Medicare Fee-For-Service beneficiaries.

Material and Methods: Among the 49+ million beneficiaries in our study, 1.5% experienced a hospitalization with PD, defined as an initial hospitalization during the study period with a PD-related diagnosis code (2001-2016). Annual air pollution concentrations from an ensemble of machine learning algorithms were linked to residential ZIP codes of Medicare beneficiaries. For each exposure, four different causal inference models were fitted: 1) traditional outcome 2) marginal structural (MSM), and 3) doubly robust (DR) Poisson regression models, adjusted for same-year sociodemographic and meteorological confounders and long-term trends.

Results: All approaches yielded consistent results, with the estimates from the MSM being the most similar to those from the DR models. Higher same-year concentrations of all air pollutants were observed to increase rates of hospitalizations with PD; the DR models yielded incidence rate ratios (IRRs) of 1.084 (95% CI: 1.070, 1.098), 1.066 (95% CI: 1.048, 1.084), and 1.034 (95% CI: 1.022, 1.046) for an interquartile range increase in PM_2.5 (3.717 µg/m³), NO₂ (13.838 ppb), and O₃ (10.091 ppb), respectively. 

Conclusion: In this nationwide retrospective study, we observed exposure to ambient air pollution to affect the rate of hospitalizations with PD. Confidence in our findings is bolstered by our use of several modeling methods. Future work on other clinical PD endpoints is needed.