Hierarchical spatial modeling of uncertainty in air pollution and birth weight study

The impact of GPS-derived activity spaces on personal PM2.5 exposures in the MADRES cohort

BACKGROUND

In-utero exposure to particulate matter with aerodynamic diameter less than 2.5 μm (PM_2.5) is associated with low birth weight and health risks later in life. Pregnant women are mobile and locations they spend time in contribute to their personal PM_2.5 exposures. Therefore, it is important to understand how mobility and exposures encountered within activity spaces contribute to personal PM_2.5 exposures during pregnancy.

METHODS

We collected 48-h integrated personal PM_2.5 samples and continuous geolocation (GPS) data for 213 predominantly Hispanic/Latina pregnant women in their 3^rd trimester in Los Angeles, CA. We also collected questionnaires and modeled outdoor air pollution and meteorology in their residential neighborhood. We calculated three GPS-derived activity space measures of exposure to road networks, greenness (NDVI), parks, traffic volume, walkability, and outdoor PM_2.5 and temperature. We used bivariate analyses to screen variables (GPS-extracted exposures in activity spaces, individual characteristics, and residential neighborhood exposures) based on their relationship with personal, 48-h integrated PM_2.5 concentrations. We then built a generalized linear model to explain the variability in personal PM_2.5 exposure and identify key contributing factors.

RESULTS

Indoor PM_2.5 sources, parity, and home ventilation were significantly associated with personal exposure. Activity-space based exposure to roads was associated with significantly higher personal PM_2.5 exposure, while greenness was associated with lower personal PM_2.5 exposure (β = -3.09 μg/m³ per SD increase in NDVI, p-value = 0.018). The contribution of outdoor PM_2.5 to personal exposure was positive but relatively lower (β = 2.05 μg/m³ per SD increase, p-value = 0.016) than exposures in activity spaces and the indoor environment. The final model explained 34% of the variability in personal PM_2.5 concentrations.

CONCLUSIONS

Our findings highlight the importance of activity spaces and the indoor environment on personal PM_2.5 exposures of pregnant women living in Los Angeles, CA. This work also showcases the multiple, complex factors that contribute to total personal PM_2.5 exposure.

In-utero personal exposure to PM2.5 impacted by indoor and outdoor sources and birthweight in the MADRES cohort

BACKGROUND

In-utero exposure to outdoor particulate matter with aerodynamic diameter less than 2.5 μm (PM_2.5) is linked with low birthweight. However, previous results are mixed, likely due to measurement error introduced by estimating personal exposure from ambient data. This study investigated the effect of total personal PM_2.5 exposure on birthweight and whether it differed when it was more heavily impacted by sources of indoor vs outdoor origin in the MADRES cohort study.

METHODS

Personal PM_2.5 exposure was measured in 205 pregnant women in the 3rd trimester using 48 h integrated, filter-based sampling. Linear regression was used to test the association between personal PM_2.5 exposure and birthweight, adjusting for key covariates. Interactions of PM_2.5 with variables representing indoor sources of PM_2.5, home ventilation, or time spent indoors tested whether the effect of total PM_2.5 on birthweight varied when it was more impacted by sources of indoor vs outdoor origin.

RESULTS

In a sample of largely Hispanic (81%) pregnant women, total personal PM_2.5 was not significantly associated with birthweight (β = 38.6 per 1SD increase in PM_2.5; 95% CI:-21.1, 98.2). This association however, differed by home type (single family home: 156.9 (26.9, 287.0), 2-4 attached units:-16.6 (-111.9, 78.7), 5+ units:-62.6 (-184.9, 59.6), missing: 145.4 (-4.1, 294.9), interaction p = 0.028) and by household air conditioner use (none of the time: -27.6 (-101.5, 46.3) vs. some of the time: 139.9 (42.9, 237.0), interaction p = 0.008) Additionally, the effect of personal PM_2.5 on birthweight varied by time spent indoors (none or little of the time: - 45.1 (-208.3, 118.1) vs. most or all of the time: 57.1 (-7.3, 121.6), interaction p = 0.255).

CONCLUSIONS

While no significant association between total personal PM_2.5 exposure and birthweight was found, there was evidence that multi-unit housing (vs. single-family homes), candle and/or incense smoke, and greater outdoor source contributions to personal PM_2.5 were more strongly associated with lower birthweight.

Multiscale measurement error models for aggregated small area health data

Spatial data are often aggregated from a finer (smaller) to a coarser (larger) geographical level. The process of data aggregation induces a scaling effect which smoothes the variation in the data. To address the scaling problem, multiscale models that link the convolution models at different scale levels via the shared random effect have been proposed. One of the main goals in aggregated health data is to investigate the relationship between predictors and an outcome at different geographical levels. In this paper, we extend multiscale models to examine whether a predictor effect at a finer level hold true at a coarser level. To adjust for predictor uncertainty due to aggregation, we applied measurement error models in the framework of multiscale approach. To assess the benefit of using multiscale measurement error models, we compare the performance of multiscale models with and without measurement error in both real and simulated data. We found that ignoring the measurement error in multiscale models underestimates the regression coefficient, while it overestimates the variance of the spatially structured random effect. On the other hand, accounting for the measurement error in multiscale models provides a better model fit and unbiased parameter estimates.