Air pollution is associated with various adverse health impacts and is identified as one of the leading risk factors for global disease burden. Further, air pollution is one of the pathways through which climate change could negatively impact health. Field studies have shown that air pollution has high spatiotemporal variability and pollutant concentrations vary substantially within neighborhoods. Characterizing air pollution at a fine-grained level is essential for accurately estimating human exposure, assessing its impact to human health, and further aiding localized air pollution policy. Air quality models are developed to estimate air pollution at locations and time periods without monitors, and these estimates are commonly used for exposure and health effects studies. Traditional land use regression [LUR] models are one of the cost-effective empirical air quality models. LUR typically relies on fixed-site measurements, GIS-derived variables with limited spatial resolution, and captures linear relationships. In recent years, innovative open-source imagery datasets and their associated features (e.g., street view imagery, remote sensing imagery) have emerged and show potential to augment or replace traditional LUR predictors. Such imagery data sources embody abundant information of natural and built environment features. Advanced computer vision techniques enable feature extraction and quantification through these extensive imagery datasets. The overarching objective of this dissertation is to investigate the feasibility of leveraging open-source imagery datasets (i.e., Google Street View [GSV] imagery, Landsat imagery, etc.) and advanced machine learning algorithms to develop image-based empirical air quality models at both local and national scale. The first study of this work established a pipeline of feature extraction through street view imagery sematic segmentation. The resulting street view features were used to predict street-level particulate air pollution for a single city. The results showed that solely using GSV-derived features can achieve comparable model fits as using traditional GIS-derived variables. Feature engineering improved model stability and interpretability through reducing spurious variables from potential misclassifications from computer vision algorithms. The second study further developed GSV-based models at national scale across multiple years. Random forest models were developed to capture the nonlinear relationship between air pollution and its impacting factors. The results showed that with sufficient street view images, GSV imagery alone may explain the variation of long-term national NO2 concentrations. Adding satellite-derived aerosol estimates (i.e., OMI column density) can significantly boost model performance when GSV images are insufficient, but the addition narrows when more GSV images are available. Our systematic assessment of the impact of image availability on model performance suggested that a parsimonious image sampling strategy (i.e., one GSV image per 100m grid) may be sufficient and most cost-effective for model development and application. Our third study explored the feasibility of combining street view and remote sensing derived features for national NO2 and PM2.5 modeling and projection at high spatial resolution. We found that GSV-based models captured both the highest and lowest pollutant concentrations while remote sensing features tended to smooth the air pollution variations. The results suggested that GSV features may have the capability to better capture fine-scale air pollution variability. The resulting air pollution prediction product may serve a variety of applications, including providing new insights into environmental justice and epidemiological studies due to its high spatial resolution (i.e., street level). Collectively, the result of this dissertation suggests that GSV imagery, processed with computer vision techniques, is a promising data source to develop empirical air quality models with high spatial resolution and consistent predictor variables processing protocol. Image-based features assisted with advanced ML approaches have the potential to greatly improve air quality modeling estimates, and successfully show comparable and even superior model performance than other modeling studies. Moreover, the ever-growing public imagery data sources are particularly promising for remote or less developed areas where traditional curated geodatabases are sparse or nonexistent. / Doctor of Philosophy / Air pollution is detrimental to human health and well-being. Further, air pollutants concentrations can change rapidly within a short distance and temporal frame. Monitoring air pollution with high spatial-temporal resolution is important. Traditional air quality monitoring networks are expensive and sparsely distributed, leading to gaps in capturing the air pollution at small spatial scales. Air quality models are developed to estimate air pollution at locations and time periods without monitors. Empirical air quality models often use air measurements from stationary sites and GIS-derived features (e.g., traffic, population density, land use types, etc.) to develop regression models and use the regression formula to estimate air pollutant concentrations in unmonitored areas. However, GIS-derived features are often collected from curated GIS databases, which often have coarse resolution when available across large geography. Street view imagery and remote sensing imagery contains rich information of natural and built environments. Computer vision techniques can be applied to extract such information to replace or augment traditional GIS-derived features. Combined with advanced machine learning algorithms, features derived from open-access images are promising to develop air quality models with a consistent image collection and processing protocol. This dissertation examines the feasibility of using street view imagery (i.e., Google Street View [GSV] Imagery) and remote sensing imagery to develop air quality models at both local and national scales. Our results found that solely using GSV features to build local and national models can achieve good model performance, which is consistent or even better than other models using traditional GIS-derived variables. For areas without sufficient GSV images, adding satellite observations for air pollution can significantly enhance model performance. Remote sensing features tend to smooth air pollution variation while GSV features tend to better capture fine-scale intra-urban air pollution variation. In conclusion, leveraging open-source imagery datasets with advanced machine learning methods are promising for estimating air pollution at high spatial resolution with good model fits.

L’exposition prolongée par inhalation à des poussières de taille respirable contenant de la silice cristalline est reconnue pour causer des maladies respiratoires dont le cancer du poumon et la silicose. De nombreuses études ont relevé la surexposition des travailleurs de la construction à la silice cristalline, puisque ce composé est présent dans de nombreux matériaux utilisés sur les chantiers. L’évaluation de l’exposition à la silice cristalline dans cette industrie constitue un défi en raison de la multitude de conditions de travail et de la nature éphémère des chantiers. Afin de mieux cerner cette problématique, une banque de données d’exposition professionnelle compilée à partir de la littérature a été réalisée par une équipe de l’Université de Montréal et de l’IRSST, et constitue le point de départ de ce travail. Les données présentes dans la banque ont été divisées en fonction de la stratégie d’échantillonnage, résultant en deux analyses complémentaires ayant pour objectif d’estimer les niveaux d’exposition sur le quart de travail en fonction du titre d’emploi, et selon la nature de la tâche exécutée. La méthode de Monte Carlo a été utilisée pour recréer les échantillons provenant de données rapportées sous forme de paramètres de synthèse. Des modèles Tobit comprenant les variables de titre d’emploi, tâche exécutée, durée, année et stratégie d’échantillonnage, type de projet, secteur d’activité, environnement et moyens de maîtrise ont été développés et interprétés par inférence multimodèle. L’analyse basée sur le quart de travail a été réalisée à partir de 1346 données d’exposition couvrant 11 catégories de titre d’emploi. Le modèle contenant toutes les variables a expliqué 22% de la variabilité des mesures et la durée, l’année et la stratégie d’échantillonnage étaient d’importants prédicteurs de l’exposition. Les chantiers de génie civil et les projets de nouvelle construction étaient associés à des expositions plus faibles, alors que l’utilisation de moyens de maîtrise diminuait les concentrations de 18% à l’extérieur et de 24% à l’intérieur. Les moyennes géométriques les plus élevées prédites pour l’année 1999 sur 8 heures étaient retrouvées chez les foreurs (0.214 mg/m3), les travailleurs souterrains (0.191 mg/m3), les couvreurs (0.146 mg/m3) et les cimentiers-applicateurs (0.125 mg/m3). 1566 mesures réparties en 27 catégories de tâches étaient contenues dans la seconde analyse. Le modèle contenant toutes les variables a expliqué 59% des niveaux d’exposition, et l’ensemble des variables contextuelles étaient fortement prédictives. Les moyennes géométriques prédites pour l’année 1998 et selon la durée médiane par tâche dans la banque de données étaient plus élevées lors du bouchardage du béton (1.446 mg/m3), du cassage de pièces de maçonnerie avec autres outils (0.354 mg/m3), du décapage au jet de sable (0.349 mg/m3) et du meulage de joints de brique (0.200 mg/m3). Une diminution importante des concentrations a été observée avec les systèmes d’arrosage (-80%) et d’aspiration des poussières (-64%) intégrés aux outils. L’analyse en fonction des titres d’emploi a montré une surexposition généralisée à la valeur guide de l’ACGIH et à la norme québécoise, indiquant un risque à long terme de maladies professionnelles chez ces travailleurs. Les résultats obtenus pour l’évaluation en fonction de la tâche exécutée montrent que cette stratégie permet une meilleure caractérisation des facteurs associés à l’exposition et ainsi de mieux cibler les priorités d’intervention pour contrôler les niveaux d’exposition à la silice cristalline sur les chantiers de construction durant un quart de travail. / Chronic inhalation of respirable inorganic dusts containing crystalline silica is linked to occupational respiratory diseases such as lung cancer and silicosis. Several studies have documented the overexposure to respirable crystalline silica in construction workers, as this compound is found in several building materials and many operations can release fine suspended particles. Comprehensive quantitative exposure assessment in this industry is challenging due to the variety in tasks performed, materials used and environmental conditions between work sites, the mobile workforce and the temporary nature of construction sites. An occupational exposure database of silica exposure was compiled from the literature by a research group from the Université de Montréal and the Institut de recherche en santé et en sécurité du travail (IRSST) to address this issue. The exposure data contained in the database were separated on the basis of sampling strategies, which resulted in two separate – but complementary – analyses. The first analysis was restricted to samples collected to compare levels with an occupational exposure limit, in order to estimate work-shift respirable crystalline silica exposure by construction trade. The second analysis used measurements collected under a task-based sampling strategy in order to estimate the exposure levels associated with specific activities. Monte-Carlo simulation was used to recreate individual exposures from measurements reported as summary statistics. Modeling was performed using Tobit models within a multimodel inference framework, with construction trade, task, sampling duration, year and strategy, project type, construction sector, workspace and control methods as potential predictors. The dataset for the analysis by construction trade was comprised of 1346 exposure measurements and included 11 trade categories. The model containing all the variables explained 22% of the exposure variability and the sampling duration, year and strategy were identified as important predictors. Civil engineering and roadwork sites as well as new construction projects were associated with lower exposure levels, while the use of control methods reduced silica concentrations by 18% outdoors and 24% indoors. Predicted geometric means (GM) for year 1999 were the highest for drillers (0.214 mg/m3), underground workers (0.191 mg/m3), roofers (0.146 mg/m3) and cement grinders/finishers (0.125 mg/m3), based on a 8-hour shift. Heavy equipment operators (0.041 mg/m3) and foremen (0.047 mg/m3) had the lowest predicted GMs. 1566 task-based measurements, representing 27 task categories, were included in the activity-specific dataset. The proportion of variance explained by the model containing all the variables was 59%, and all the variables investigated had a strong influence on the exposure levels. Predictions were made based on the year 1998 and the median duration by task in the dataset. The largest predicted GMs were associated with the following operations: scabbling concrete (1.446 mg/m3), chipping with other tools (0.354 mg/m3), abrasive blasting (0.349 mg/m3) and tuck point grinding (0.200 mg/m3). Important reductions in exposure levels were found with the use of tool-integrated water sprays (-80%) and local exhaust ventilation (-64%). Important overexposure to the ACGIH Threshold Limit Value and the Québec exposure limit was found for all the trades investigated, indicating a long-term risk of silica-related occupational diseases. The results of the task-based analysis suggest that this sampling strategy provides a better characterization of the factors affecting exposure and the impacts of engineering dust control methods to control long-term exposure levels.

Doctor of Philosophy / Food Science / Justin Kastner / Many Chinese aquaculture farmers use unapproved chemicals to treat their fish, many of which are diseased as a result of the country’s poor waste management and environmental practices. During 2006-2007, the United States (U.S.), the European Union, and Japan rejected large amounts of Chinese seafood imports due to the presence of unapproved chemicals or the presence of approved chemicals at concentrations that exceeded permitted levels. This dissertation examines the sources of environmental health and food safety problems in China; it also examines how effective the U.S. and Chinese governments’ regulations are in protecting consumers from hazards in Chinese aquaculture products. The study looks at specific chemical contaminants found in Chinese aquaculture imports, explores their potential toxicity or carcinogenicity, and examines the reasons for their prohibition from human food. The study exploits the available U.S. seafood consumption patterns (courtesy of the National Health and Nutrition Examination Survey—NHANES—database) and then uses probabilistic modeling (courtesy of CREMe Software Limited) to determine the extent to which specific sensitive U.S. consumer subpopulations were exposed to aquaculture chemical contaminants in the food supply in a contrived scenario using real consumption data (from NHANES) and actual contamination data (from the FDA). The study compares exposure between children and adult consumers, and also looks at exposure among women aged 18 years and older and the elderly aged 60 years and older. This study suggests a strong likelihood that NHANES children, as well as female consumers aged 18 years and older and elderly consumers aged 60 years and older, were (in the contrived scenario) all exposed to violative intake levels of chemical contaminants from Chinese aquaculture imports. Children forming the 99.5th and 99.9th percentiles of NHANES seafood consumers were exposed to higher levels of nitrofuran, gentian violet, and malachite green contaminants per kilogram of body weight than were their adult counterparts. Conversely, children were exposed to lower levels of fluoroquinolone contaminants per kilogram of body weight than were their adult counterparts. The 50th, 95th, and 99.9th percentiles of female consumers aged 18 and older and elderly consumers aged 60 years and older were exposed to violative daily intake levels of contaminants in Chinese aquaculture. The study concludes by examining what the U.S. and Chinese governments should do to address aquaculture safety.

En France, deux bases de données d’exposition professionnelle, COLCHIC et SCOLA, coexistent avec des objectifs différents (prévention et réglementation). Leur représentativité par rapport à la population générale est cependant inconnue, et fait l’objet de ce travail. Après avoir effectué une analyse descriptive comparative, l’étude de l’association entre les niveaux d’exposition et les éléments contextuels a été réalisée par modélisation statistique pour chaque agent, séparément pour COLCHIC et SCOLA, puis dans un jeu de données commun. La synthèse à travers les agents s’est faite par Méta analyse. COLCHIC et SCOLA contiennent respectivement 929 700 (670 agents chimiques) et 429 104 données (105). Trois forts prédicteurs « Durée de prélèvement », « Equipement de protection individuelle » et « Année » sont systématiquement associés aux niveaux dans les deux bases et 3 autres sont spécifiques à chacune d’elles. Avec des niveaux deux fois plus élevés dans COLCHIC comparés à SCOLA en 2007, leurs concentrations deviennent comparables entre 2012 et 2015. COLCHIC et SCOLA représentent une source importante d’informations. La prise en compte des descripteurs associés aux mesures et l’utilisation de méthodes prédictives permettront d’en améliorer l’interprétation / Two occupational exposure databases of occupational exposures to chemicals, COLCHIC and SCOLA, coexist in France with different objectives (prevention and compliance). Little is known about their representativeness of exposures in the general population. We explored to what extent COLCHIC and SCOLA adequately reflect occupational exposures in France. After performing a descriptive and comparative analysis, associations between exposure levels and ancillary information were explored for each agent, separately for COLCHIC and SCOLA and in a common dataset, using statistical modelling. Modelling results were synthesized across agents using Meta analysis. COLCHIC and SCOLA contain, respectively, 929 700 (670 chemicals) and 429 104 records (105). Three predictors "Sample Time", "Personal protective equipment" and "Year" are strongly associated with exposure levels across a large majority of chemicals in both databases, and 3 others are specific to each one. Exposure levels are in average twice higher in COLCHIC compared to SCOLA in 2007, but become comparable from 2012-2015. COLCHIC and SCOLA are an important source of information. Inclusion of descriptors associated with exposure levels in our study and the use of predictive methods should help to improve their interpretation.

La veille sanitaire dans le champ de la santé au travail est cruciale pour détecter l’apparition de nouveaux risques. Le réseau National de Vigilance et de Prévention des Pathologies Professionnelles (RNV3P), réseau national d’experts opérationnel depuis 2011, y participe activement. Pour cela, le RNV3P a déployé plusieurs méthodes pour détecter les nouveaux risques : émergence clinique et émergence statistique. La méthode des mesures de disproportion couramment utilisée dans le domaine de la pharmacovigilance a déjà été appliquée avec succès aux données du RNV3P. Néanmoins, une limite importante peut être soulevée. Les professionnels de santé déclarant les expositions qui leur semble liées à la pathologie déclarée, nous pouvons légitimement suspecter l'existence d'un biais de sélection de ces expositions : l'exposition la plus reconnue par le monde professionnel sera forcément déclarée. Pour tenter de maîtriser ce biais, nous avons proposé d’appliquer les matrices emploi / exposition (MEE) sur les données du RNV3P et d’examiner leur impact sur les mesures de disproportion. Dans un premier temps, nous avons testé la faisabilité de l’application des MEE au RNV3P et leur apport. Pour cela, 3 MEE issues du programme Matgéné ont été appliquées (benzène, poussières respirables de silice cristalline libre et solvants chlorés). Les données d’exposition rapportées respectivement par les experts et par la MEE ont été comparées en particulier pour des associations bien connues comme la silicose/silice ou les maladies hématopoïétiques/benzène, et pour des associations moins connues ou douteuses comme la sclérose systémique/silice et les maladies hématopoïétiques/solvants chlorés (trichloréthylène). Dans un second temps, nous avons appliqué les mesures de disproportion de type fréquentielle et bayésienne en comparant les résultats obtenus par les experts et les MEE pour trois nuisances : silice, trichloroéthylène et tétrachloroéthylène. Au final, les MEE permettent de rapporter plus d’expositions que les experts pour chaque nuisance testée et ce d’autant plus que l’association est peu connue. Cet apport d’information permet la génération de nouveaux signaux avec les mesures de disproportion qu’il est intéressant de discuter et d’explorer avec des experts / Health surveillance in the field of occupational health is crucial to detect the emergence of new risks The National Network of Vigilance and Prevention of Occupational Pathologies (RNV3P), a national network of experts operating since 2011, actively participates in. The RNV3P has deployed several methods to detect new risks: clinical emergence and statistical emergence. In the latter case, the disproportionate measures used in the field of pharmacovigilance have been successfully applied to RNV3P. Nevertheless, an important limitation can be raised. As health professionals report exposures associated with the declared pathology, we can legitimately suspect the existence of a selection bias: the most well known exposures of the professional community inevitably are most liable to be more frequently declared. To try to control this bias, we proposed to apply job / exposure matrices (JEM) on the data of the RNV3P and to examine their impact on the measures of disproportion. Initially, we tested the feasibility of applying MEE to the RNV3P data and their contribution. For this, 3 JEMs from the Matgéné program were applied (benzene, free crystalline silica breathing dusts and chlorinated solvents). The exposure data reported by the experts and the JEM were compared in particular for well-known associations such as silicosis / silica or hematopoietic / benzene diseases, and for less known or doubtful associations such as systemic sclerosis and hematopoietic diseases / chlorinated solvents (trichloroethylene). Secondly, we applied the frequency and Bayesian disproportion measures by comparing the results obtained for experts and JEM and for three nuisances: silica, trichloroethylene and tetrachloroethylene. In summary, MEE help to provide more exposure than the experts for each tested chemicals and especially if the association is little known. These new data allow the generation of new signals with the disproportion measures that it is interesting to discuss and explore with experts

Introdução São Paulo é uma das maiores cidades da América Latina, com quase 11 milhões de habitantes e cerca de 6 milhões de veículos. Embora o tráfego seja a mais importante fonte de poluição atmosférica, poucos estudos investigaram a relação da poluição veicular com a saúde da população. A maioria dos estudos analisou os efeitos da poluição do ar à saúde utilizando valores médios de poluição ambiental para toda a área da cidade, os quais não evidenciam os gradientes de exposição na área intra-urbana. Objetivos- Avaliar a exposição da população à poluição relacionada ao tráfego e sua associação com as internações por doenças respiratórias de crianças e adolescentes de 0 a 18 anos. Métodos As concentrações de CO, NOx e PM10, foram calculadas a partir do modelo de dispersão CALINE-4 para os períodos de verão (Janeiro) e Inverno (Julho). Os casos de internação hospitalar por doenças respiratórias (AIH + CIH) foram georreferenciados por local de residência. O setor censitário do IBG foi considerado como unidade de análise. Um modelo de regressão logística foi usado, para estimar a associação entre exposição à poluição relacionada ao tráfego e hospitalização por doenças respiratórias, e controlado pelo IDH como indicador sócio econômico. Resultados Do ponto de vista espacial, os poluentes veiculares estudados tiveram maior concentração na área central do centro expandido de São Paulo. Do ponto de vista temporal, as maiores concentrações foram observadas no inverno. A poluição veicular estava diretamente associada às internações de crianças e adolescentes (0-18) anos por doenças respiratórias. Crianças e adolescentes que moravam em setores censitários classificados no 4º quartil de CO no período de inverno tiveram chance 80 por cento maior de serem internadas por doenças respiratórias. Conclusões Os poluentes veiculares analisados aumentam a chance de crianças e adolescentes (0-18 anos) serem internadas por doenças respiratórias. As condições socioeconômicas, avaliadas pelo IDH, também aumentam as chances de internação. O método usado neste estudo é importante para avaliações em micro-escala da relação entre os poluentes veiculares e a saúde da população. Outras cidades brasileiras ou cidades de países em desenvolvimento podem se beneficiar desta abordagem, dado que modelos são mais baratos e rápidos que campanhas de amostragem de poluentes atmosféricos ou aquisição/manutenção de estações de monitoramento da qualidade do ar / Introduction- Sao Paulo is one of the largest cities in Latin America, with almost 11 million inhabitants and around 6 million vehicles. Although traffic is the main source of air pollution, few studies investigated the relationship between vehicle pollution and health outcomes. Most studies analyzed health effects using average concentrations of environmental pollution for the whole city, which cannot give evidence for intra-urban gradients of exposure. Objectives- To evaluate the populations exposure to traffic-related air pollution and its association with hospital admission for respiratory diseases among children and adolescents aged 0-18 years. Methods - Concentrations of CO, NOx and PM10 were modeled using CALINE-4 dispersion model, in two periods: summer (January) and winter (June). Hospitalizations due to respiratory diseases (by private and public assistance) were geocoded by the residence address. IBGEs census sectors were considered as unit of analysis. A logistic regression model was used to estimate the association between exposure to traffic-related air pollution and hospitalization for respiratory disease which, allowing for HDI as a socioeconomical indicator. Results- Spatially, pollutants presented higher concentration at the central area of the Expanded Center of Sao Paulo city. Temporally, higher concentrations were observed at winter periods. Traffic-related pollutants was directly associated with hospitalization for respiratory disease among children and adolescents aged 0-18 years. Children and adolescents who lived in census sectors ranked in the 4º quartile of CO in the winter period had 80 per cent greater chance of being hospitalized for respiratory diseases. Conclusions- Traffic-related pollutants increase the chance of children being hospitalized for respiratory diseases. Socioeconomic conditions (evaluated by the HDI) also raised the chance of hospitalization. The method used in this study is important for micro-scale evaluations of the relationship between vehicular pollutants and population health. Other Brazilian cities or cities from developing countries may benefit from this approach, since models are less expensive and faster than air quality monitoring campaigns or acquisition/maintenance of air quality monitoring stations

Introdução São Paulo é uma das maiores cidades da América Latina, com quase 11 milhões de habitantes e cerca de 6 milhões de veículos. Embora o tráfego seja a mais importante fonte de poluição atmosférica, poucos estudos investigaram a relação da poluição veicular com a saúde da população. A maioria dos estudos analisou os efeitos da poluição do ar à saúde utilizando valores médios de poluição ambiental para toda a área da cidade, os quais não evidenciam os gradientes de exposição na área intra-urbana. Objetivos- Avaliar a exposição da população à poluição relacionada ao tráfego e sua associação com as internações por doenças respiratórias de crianças e adolescentes de 0 a 18 anos. Métodos As concentrações de CO, NOx e PM10, foram calculadas a partir do modelo de dispersão CALINE-4 para os períodos de verão (Janeiro) e Inverno (Julho). Os casos de internação hospitalar por doenças respiratórias (AIH + CIH) foram georreferenciados por local de residência. O setor censitário do IBG foi considerado como unidade de análise. Um modelo de regressão logística foi usado, para estimar a associação entre exposição à poluição relacionada ao tráfego e hospitalização por doenças respiratórias, e controlado pelo IDH como indicador sócio econômico. Resultados Do ponto de vista espacial, os poluentes veiculares estudados tiveram maior concentração na área central do centro expandido de São Paulo. Do ponto de vista temporal, as maiores concentrações foram observadas no inverno. A poluição veicular estava diretamente associada às internações de crianças e adolescentes (0-18) anos por doenças respiratórias. Crianças e adolescentes que moravam em setores censitários classificados no 4º quartil de CO no período de inverno tiveram chance 80 por cento maior de serem internadas por doenças respiratórias. Conclusões Os poluentes veiculares analisados aumentam a chance de crianças e adolescentes (0-18 anos) serem internadas por doenças respiratórias. As condições socioeconômicas, avaliadas pelo IDH, também aumentam as chances de internação. O método usado neste estudo é importante para avaliações em micro-escala da relação entre os poluentes veiculares e a saúde da população. Outras cidades brasileiras ou cidades de países em desenvolvimento podem se beneficiar desta abordagem, dado que modelos são mais baratos e rápidos que campanhas de amostragem de poluentes atmosféricos ou aquisição/manutenção de estações de monitoramento da qualidade do ar / Introduction- Sao Paulo is one of the largest cities in Latin America, with almost 11 million inhabitants and around 6 million vehicles. Although traffic is the main source of air pollution, few studies investigated the relationship between vehicle pollution and health outcomes. Most studies analyzed health effects using average concentrations of environmental pollution for the whole city, which cannot give evidence for intra-urban gradients of exposure. Objectives- To evaluate the populations exposure to traffic-related air pollution and its association with hospital admission for respiratory diseases among children and adolescents aged 0-18 years. Methods - Concentrations of CO, NOx and PM10 were modeled using CALINE-4 dispersion model, in two periods: summer (January) and winter (June). Hospitalizations due to respiratory diseases (by private and public assistance) were geocoded by the residence address. IBGEs census sectors were considered as unit of analysis. A logistic regression model was used to estimate the association between exposure to traffic-related air pollution and hospitalization for respiratory disease which, allowing for HDI as a socioeconomical indicator. Results- Spatially, pollutants presented higher concentration at the central area of the Expanded Center of Sao Paulo city. Temporally, higher concentrations were observed at winter periods. Traffic-related pollutants was directly associated with hospitalization for respiratory disease among children and adolescents aged 0-18 years. Children and adolescents who lived in census sectors ranked in the 4º quartile of CO in the winter period had 80 per cent greater chance of being hospitalized for respiratory diseases. Conclusions- Traffic-related pollutants increase the chance of children being hospitalized for respiratory diseases. Socioeconomic conditions (evaluated by the HDI) also raised the chance of hospitalization. The method used in this study is important for micro-scale evaluations of the relationship between vehicular pollutants and population health. Other Brazilian cities or cities from developing countries may benefit from this approach, since models are less expensive and faster than air quality monitoring campaigns or acquisition/maintenance of air quality monitoring stations

La contaminació atmosfèrica és un problema de salut pública, causa més de 380 000 morts a la Unió Europea. La present tesis té per objectius avaluar l'exposició i efectes sobre el tracte respiratori de la contaminació provinent del tràfic. Les concentracions exteriors y personals de sulfurs i de carbó són bons indicadors de exposició personal en una ciutat mediterrània; per a PM2.5 hi ha altres fonts a considerar. La molèstia deguda a la contaminació no és un bon indicador d'exposició, però reflecteix la percepció del subjecte. La contaminació que prové del tràfic augmenta els símptomes d'asma y probablement també causa asma en adults. El PM2.5 provinent de la combustió augmenta la permeabilitat de la barrera epitelial pulmonar. El tràfic és una font important de contaminació. Es requereixen eines adequades per a mesurar la seva exposició. La contaminació del tràfic es un factor de risc important per a la salut respiratòria. / La contaminación atmosférica es un problema de salud pública, causa 380 000 muertes anuales en la Unión Europea. Esta tesis tiene como objetivo evaluar la exposición a la contaminación debida al tráfico y sus efectos en la salud respiratoria. Los niveles diarios de carbón y sulfuro medidos centralmente son buenos indicadores de exposición personal en una ciudad mediterránea, para PM2.5 fuentes de emisión alternas se tienen que considerar. La molestia debida a la contaminación no es un marcador de exposición, pero es importante porque refleja las percepciones individuales. La contaminación proveniente del tráfico aumento los síntomas del asma, y probablemente también causa asma en adultos. El PM2.5 proveniente de la combustión aumenta la permeabilidad de la barrera epitelial pulmonar. El tráfico es una fuente importante de la contaminación. Herramientas adecuadas para medir su exposición son requeridas. La contaminación del tráfico es un factor de riesgo importante para la salud respiratoria. / Air pollution is a major public health concern causing annually 380 000 deaths in the European Union. This thesis aims to study traffic-related air pollution exposure assessment and its association with respiratory effects. Daily levels of carbon and sulphur of outdoor central measurements are good surrogates for personal exposure in a Mediterranean setting; for PM2.5 other sources have to be taken into account. Annoyance due to air pollution is not a valid maker of air pollution exposure but is valuable in its own right as it integrates individual perceptions. Traffic-related air pollution increases asthma symptoms in adults and an association with new asthma onset is suggested. Furthermore, PM2.5 from combustion might lead to an increase in the lung's epithelial barrier permeability. Traffic-related air pollution is a major source of pollution. Adequate tools to assess its exposure are still needed. Traffic-related air pollution is an important risk factor for respiratory morbidity.

In order to investigate the etiology of and evaluate interventions for work-related back injuries, researchers need efficient, accurate occupational exposure assessment methods suitable for large samples. The chapters in this thesis examine critical decisions using electromyography (EMG): How should exposure be measured? For what duration? Who should be measured, and how many times? Low-back EMG, or muscle activity data, was collected during 138 full-shift field measurements over 30 different job titles at 50 different worksites in 5 heavy industries: forestry, transportation, wood products, construction, and warehousing. Observations and self-reports of posture, manual materials handling (MMH), and driving exposures were collected concurrently. 1) Variability of EMG calibration measurements was investigated on right/left sides, multiple trials, 4 positions, and pre/post-shift. Position accounts for the majority of explained variability; there is little to gain by measuring multiple trials or pre- and post-shift, but measuring both sides and multiple positions is worthwhile. 2) Observation and self-report data were easier to collect and cheaper than the EMG direct measure. Costs and successful field performance need to be weighed against the added data detail when making choices about exposure assessment techniques for epidemiological studies. 3) Observed and self-reported exposures were used to predict EMG exposure metrics using mixed multiple linear regression models. Regression models using observed variables predicted 43-50% of the variability in the EMG metrics, while self-reported variables predicted 21%-36%. The observation exposure model provides a low-cost alternative to direct measurement. The self-reported exposure model should be considered with more caution. 4) Full-shift EMG data was resampled for 4, 2, and 1 hour, and for 10 and 2 minute durations to determine the optimal sampling duration. Bias was consistently low, but shorter durations had higher absolute error, percentage error, and limits of agreement. Durations of 4 and 2 hours may be acceptable but those less than 1 hour had large errors. 5) Components of EMG variance were calculated between- and within-subject, and between- industry, company, job, and post hoc grouping. Resolution, contrast, and exposure-response relationship attenuation were calculated for each grouping scheme. The post hoc scheme had the highest contrast and lowest resolution.

L’exposition prolongée par inhalation à des poussières de taille respirable contenant de la silice cristalline est reconnue pour causer des maladies respiratoires dont le cancer du poumon et la silicose. De nombreuses études ont relevé la surexposition des travailleurs de la construction à la silice cristalline, puisque ce composé est présent dans de nombreux matériaux utilisés sur les chantiers. L’évaluation de l’exposition à la silice cristalline dans cette industrie constitue un défi en raison de la multitude de conditions de travail et de la nature éphémère des chantiers. Afin de mieux cerner cette problématique, une banque de données d’exposition professionnelle compilée à partir de la littérature a été réalisée par une équipe de l’Université de Montréal et de l’IRSST, et constitue le point de départ de ce travail. Les données présentes dans la banque ont été divisées en fonction de la stratégie d’échantillonnage, résultant en deux analyses complémentaires ayant pour objectif d’estimer les niveaux d’exposition sur le quart de travail en fonction du titre d’emploi, et selon la nature de la tâche exécutée. La méthode de Monte Carlo a été utilisée pour recréer les échantillons provenant de données rapportées sous forme de paramètres de synthèse. Des modèles Tobit comprenant les variables de titre d’emploi, tâche exécutée, durée, année et stratégie d’échantillonnage, type de projet, secteur d’activité, environnement et moyens de maîtrise ont été développés et interprétés par inférence multimodèle. L’analyse basée sur le quart de travail a été réalisée à partir de 1346 données d’exposition couvrant 11 catégories de titre d’emploi. Le modèle contenant toutes les variables a expliqué 22% de la variabilité des mesures et la durée, l’année et la stratégie d’échantillonnage étaient d’importants prédicteurs de l’exposition. Les chantiers de génie civil et les projets de nouvelle construction étaient associés à des expositions plus faibles, alors que l’utilisation de moyens de maîtrise diminuait les concentrations de 18% à l’extérieur et de 24% à l’intérieur. Les moyennes géométriques les plus élevées prédites pour l’année 1999 sur 8 heures étaient retrouvées chez les foreurs (0.214 mg/m3), les travailleurs souterrains (0.191 mg/m3), les couvreurs (0.146 mg/m3) et les cimentiers-applicateurs (0.125 mg/m3). 1566 mesures réparties en 27 catégories de tâches étaient contenues dans la seconde analyse. Le modèle contenant toutes les variables a expliqué 59% des niveaux d’exposition, et l’ensemble des variables contextuelles étaient fortement prédictives. Les moyennes géométriques prédites pour l’année 1998 et selon la durée médiane par tâche dans la banque de données étaient plus élevées lors du bouchardage du béton (1.446 mg/m3), du cassage de pièces de maçonnerie avec autres outils (0.354 mg/m3), du décapage au jet de sable (0.349 mg/m3) et du meulage de joints de brique (0.200 mg/m3). Une diminution importante des concentrations a été observée avec les systèmes d’arrosage (-80%) et d’aspiration des poussières (-64%) intégrés aux outils. L’analyse en fonction des titres d’emploi a montré une surexposition généralisée à la valeur guide de l’ACGIH et à la norme québécoise, indiquant un risque à long terme de maladies professionnelles chez ces travailleurs. Les résultats obtenus pour l’évaluation en fonction de la tâche exécutée montrent que cette stratégie permet une meilleure caractérisation des facteurs associés à l’exposition et ainsi de mieux cibler les priorités d’intervention pour contrôler les niveaux d’exposition à la silice cristalline sur les chantiers de construction durant un quart de travail. / Chronic inhalation of respirable inorganic dusts containing crystalline silica is linked to occupational respiratory diseases such as lung cancer and silicosis. Several studies have documented the overexposure to respirable crystalline silica in construction workers, as this compound is found in several building materials and many operations can release fine suspended particles. Comprehensive quantitative exposure assessment in this industry is challenging due to the variety in tasks performed, materials used and environmental conditions between work sites, the mobile workforce and the temporary nature of construction sites. An occupational exposure database of silica exposure was compiled from the literature by a research group from the Université de Montréal and the Institut de recherche en santé et en sécurité du travail (IRSST) to address this issue. The exposure data contained in the database were separated on the basis of sampling strategies, which resulted in two separate – but complementary – analyses. The first analysis was restricted to samples collected to compare levels with an occupational exposure limit, in order to estimate work-shift respirable crystalline silica exposure by construction trade. The second analysis used measurements collected under a task-based sampling strategy in order to estimate the exposure levels associated with specific activities. Monte-Carlo simulation was used to recreate individual exposures from measurements reported as summary statistics. Modeling was performed using Tobit models within a multimodel inference framework, with construction trade, task, sampling duration, year and strategy, project type, construction sector, workspace and control methods as potential predictors. The dataset for the analysis by construction trade was comprised of 1346 exposure measurements and included 11 trade categories. The model containing all the variables explained 22% of the exposure variability and the sampling duration, year and strategy were identified as important predictors. Civil engineering and roadwork sites as well as new construction projects were associated with lower exposure levels, while the use of control methods reduced silica concentrations by 18% outdoors and 24% indoors. Predicted geometric means (GM) for year 1999 were the highest for drillers (0.214 mg/m3), underground workers (0.191 mg/m3), roofers (0.146 mg/m3) and cement grinders/finishers (0.125 mg/m3), based on a 8-hour shift. Heavy equipment operators (0.041 mg/m3) and foremen (0.047 mg/m3) had the lowest predicted GMs. 1566 task-based measurements, representing 27 task categories, were included in the activity-specific dataset. The proportion of variance explained by the model containing all the variables was 59%, and all the variables investigated had a strong influence on the exposure levels. Predictions were made based on the year 1998 and the median duration by task in the dataset. The largest predicted GMs were associated with the following operations: scabbling concrete (1.446 mg/m3), chipping with other tools (0.354 mg/m3), abrasive blasting (0.349 mg/m3) and tuck point grinding (0.200 mg/m3). Important reductions in exposure levels were found with the use of tool-integrated water sprays (-80%) and local exhaust ventilation (-64%). Important overexposure to the ACGIH Threshold Limit Value and the Québec exposure limit was found for all the trades investigated, indicating a long-term risk of silica-related occupational diseases. The results of the task-based analysis suggest that this sampling strategy provides a better characterization of the factors affecting exposure and the impacts of engineering dust control methods to control long-term exposure levels.