Influence of on-road mobile monitoring design on ultrafine particle exposure models and cognitive health inferences

Abstract

On-road mobile monitoring is increasingly used to assess air pollution, but the implications of monitoring and analytic decisions on exposure prediction and health inferences remain unclear. This study evaluated the influence of on-road monitoring approaches in environmental epidemiology, specifically ultrafine particle (UFP) exposures and late-life cognitive function. We used data from a Seattle-based mobile monitoring campaign to develop a reference roadside UFP exposure model based on repeated measurements at 309 roadside locations and examine associations with cognitive function (Cognitive Abilities Screening Instrument—Item Response Theory [CASI-IRT]), in the Adult Changes in Thought cohort (N = 5283). To evaluate alternative designs, we subsampled on-road UFP measurements collected along 600 km of roadways, varying location visit frequencies, spatial balancing, and sampling times. New UFP models, some incorporating temporal and plume adjustments, were developed using universal kriging with partial least squares and used to estimate associations between UFP and CASI-IRT, after adjusting for age, year, sex, education, race, and socioeconomic status. Using the reference exposure model in the primary health model, the mean baseline CASI-IRT score increased by 0.007 (95% CI: –0.013, 0.027) per 1900 pt/cm³ increment in PNC. Associations were similar but relatively attenuated for all on-road sampling designs. Route-based sampling (which accounted for logistical field constraints and spatiotemporal correlation in the data) and very short (4- vs 12-visit) campaigns produced more variable health estimates. Applying temporal and plume adjustments had a minimal impact on the inferential results. In analyses where no association was observed between UFP and cognitive function, the on-road monitoring design produced similar but slightly attenuated point estimates. Secondary analyses with a reduced health model, which indicated a statistically significant but potentially confounded association, suggest that on-road design—particularly monitoring beyond weekday business hours—may have greater implications in other contexts. Mobile monitoring is increasingly used to develop air pollution exposure models, yet the influence of monitoring design on health inferences remains unclear. Using extensive ultrafine particle (UFP) data from a monitoring campaign and health measures from a long-standing cohort study, we assess how on-road campaigns can be designed for epidemiologic research. We evaluate the effects of visit number, spatial balance, time selection, temporal adjustment methods, and plume adjustments on exposure and health models, providing guidance for mobile monitoring design in air pollution health research.