Uncertainty Reduction and Environmental Justice in Air Pollution Epidemiology: The Importance of Minority Representation.

Ambient air pollution is an increasing threat to society, with rising numbers of adverse outcomes and exposure inequalities worldwide. Reducing uncertainty in health outcomes models and exposure disparity studies is therefore essential to develop policies effective in protecting the most affected places and populations. This study uses the concept of information entropy to study tradeoffs in mortality uncertainty reduction from increasing input data of air pollution versus health outcomes. We study a case scenario for short-term mortality from particulate matter (PM2.5) in North Carolina for 2001-2016, employing a case-crossover design with inputs from an individual-level mortality data set and high-resolution gridded data sets of PM2.5 and weather covariates. We find a significant association between mortality and PM2.5, and the information tradeoffs indicate that a 10% increase in mortality information reduces model uncertainty three times more than increased resolution of the air pollution model from 12 to 1 km. We also find that Non-Hispanic Black (NHB) residents tend to live in relatively more polluted census tracts, and that the mean PM2.5 for NHB cases in the mortality model is significantly higher than that of Non-Hispanic White cases. The distinct distribution of PM2.5 for NHB cases results in a relatively higher information value, and therefore faster uncertainty reduction, for new NHB cases introduced into the mortality model. This newfound influence of exposure disparities in the rate of uncertainty reduction highlights the importance of minority representation in environmental research as a quantitative advantage to produce more confident estimates of the true effects of environmental pollution.

An illustration of model agnostic explainability methods applied to environmental data.

Historically, two primary criticisms statisticians have of machine learning and deep neural models is their lack of uncertainty quantification and the inability to do inference (i.e., to explain what inputs are important). Explainable AI has developed in the last few years as a sub-discipline of computer science and machine learning to mitigate these concerns (as well as concerns of fairness and transparency in deep modeling). In this article, our focus is on explaining which inputs are important in models for predicting environmental data. In particular, we focus on three general methods for explainability that are model agnostic and thus applicable across a breadth of models without internal explainability: "feature shuffling", "interpretable local surrogates", and "occlusion analysis". We describe particular implementations of each of these and illustrate their use with a variety of models, all applied to the problem of long-lead forecasting monthly soil moisture in the North American corn belt given sea surface temperature anomalies in the Pacific Ocean.

Information entropy tradeoffs for efficient uncertainty reduction in estimates of air pollution mortality.

Implementing effective policy to protect human health from the adverse effects of air pollution, such as premature mortality, requires reducing the uncertainty in health outcomes models. Here we present a novel method to reduce mortality uncertainty by increasing the amount of input data of air pollution and health outcomes, and then quantifying tradeoffs associated with the different data gained. We first present a study of long-term mortality from fine particulate matter (PM2.5) based on simulated data, followed by a real-world application of short-term PM2.5-related mortality in an urban area. We employ information yield curves to identify which variables more effectively reduce mortality uncertainty when increasing information. Our methodology can be used to explore how specific pollution scenarios will impact mortality and thus improve decision-making. The proposed framework is general and can be applied to any real case-scenario where knowledge in pollution, demographics, or health outcomes can be augmented through data acquisition or model improvements to generate more robust risk assessments.

Spatial modeling of mid-infrared spectral data with thermal compensation using integrated nested Laplace approximation.

The problem of analyzing substances using low-cost sensors with a low signal-to-noise ratio (SNR) remains challenging. Using accurate models for the spectral data is paramount for the success of any classification task. We demonstrate that the thermal compensation of sample heating and spatial variability analysis yield lower modeling errors than non-spatial modeling. Then, we obtain the inference of the spectral data probability density functions using the integrated nested Laplace approximation (INLA) on a Bayesian hierarchical model. To achieve this goal, we use the fast and user-friendly R-INLA package in R for the computation. This approach allows affordable and real-time substance identification with fewer SNR sensor measurements, thereby potentially increasing throughput and lowering costs.

Short-term and long-term health impacts of air pollution reductions from COVID-19 lockdowns in China and Europe: a modelling study.

BACKGROUND: Exposure to poor air quality leads to increased premature mortality from cardiovascular and respiratory diseases. Among the far-reaching implications of the ongoing COVID-19 pandemic, a substantial improvement in air quality was observed worldwide after the lockdowns imposed by many countries. We aimed to assess the implications of different lockdown measures on air pollution levels in Europe and China, as well as the short-term and long-term health impact. METHODS: For this modelling study, observations of fine particulate matter (PM2·5) concentrations from more than 2500 stations in Europe and China during 2016-20 were integrated with chemical transport model simulations to reconstruct PM2·5 fields at high spatiotemporal resolution. The health benefits, expressed as short-term and long-term avoided mortality from PM2·5 exposure associated with the interventions imposed to control the COVID-19 pandemic, were quantified on the basis of the latest epidemiological studies. To explore the long-term variability in air quality and associated premature mortality, we built different scenarios of economic recovery (immediate or gradual resumption of activities, a second outbreak in autumn, and permanent lockdown for the whole of 2020). FINDINGS: The lockdown interventions led to a reduction in population-weighted PM2·5 of 14·5 µg m-3 across China (-29·7%) and 2·2 µg m-3 across Europe (-17·1%), with unprecedented reductions of 40 µg m-3 in bimonthly mean PM2·5 in the areas most affected by COVID-19 in China. In the short term, an estimated 24 200 (95% CI 22 380-26 010) premature deaths were averted throughout China between Feb 1 and March 31, and an estimated 2190 (1960-2420) deaths were averted in Europe between Feb 21 and May 17. We also estimated a positive number of long-term avoided premature fatalities due to reduced PM2·5 concentrations, ranging from 76 400 (95% CI 62 600-86 900) to 287 000 (233 700-328 300) for China, and from 13 600 (11 900-15 300) to 29 500 (25 800-33 300) for Europe, depending on the future scenarios of economic recovery adopted. INTERPRETATION: These results indicate that lockdown interventions led to substantial reductions in PM2·5 concentrations in China and Europe. We estimated that tens of thousands of premature deaths from air pollution were avoided, although with significant differences observed in Europe and China. Our findings suggest that considerable improvements in air quality are achievable in both China and Europe when stringent emission control policies are adopted. FUNDING: None.

A scalable multi-resolution spatio-temporal model for brain activation and connectivity in fMRI data.

Functional Magnetic Resonance Imaging (fMRI) is a primary modality for studying brain activity. Modeling spatial dependence of imaging data at different spatial scales is one of the main challenges of contemporary neuroimaging, and it could allow for accurate testing for significance in neural activity. The high dimensionality of this type of data (on the order of hundreds of thousands of voxels) poses serious modeling challenges and considerable computational constraints. For the sake of feasibility, standard models typically reduce dimensionality by modeling covariance among regions of interest (ROIs)-coarser or larger spatial units-rather than among voxels. However, ignoring spatial dependence at different scales could drastically reduce our ability to detect activation patterns in the brain and hence produce misleading results. We introduce a multi-resolution spatio-temporal model and a computationally efficient methodology to estimate cognitive control related activation and whole-brain connectivity. The proposed model allows for testing voxel-specific activation while accounting for non-stationary local spatial dependence within anatomically defined ROIs, as well as regional dependence (between-ROIs). The model is used in a motor-task fMRI study to investigate brain activation and connectivity patterns aimed at identifying associations between these patterns and regaining motor functionality following a stroke.