Comparison of machine learning approaches and statistical methods for estimation of airborne chemical exposures

Accurate quantitative estimation of exposure using retrospective data has been one of the most challenging tasks in the exposure assessment field. To improve these estimates, some models have been developed using published exposure databases with their corresponding exposure determinants. These models are designed to be applied to reported exposure determinants obtained from study subjects or exposure levels assigned by an industrial hygienist, so quantitative exposure estimates can be obtained. ^ In an effort to improve the prediction accuracy and generalizability of these models, and taking into account that the limitations encountered in previous studies might be due to limitations in the applicability of traditional statistical methods and concepts, the use of computer science- derived data analysis methods, predominantly machine learning approaches, were proposed and explored in this study. ^ The goal of this study was to develop a set of models using decision trees/ensemble and neural networks methods to predict occupational outcomes based on literature-derived databases, and compare, using cross-validation and data splitting techniques, the resulting prediction capacity to that of traditional regression models. Two cases were addressed: the categorical case, where the exposure level was measured as an exposure rating following the American Industrial Hygiene Association guidelines and the continuous case, where the result of the exposure is expressed as a concentration value. Previously developed literature-based exposure databases for 1,1,1 trichloroethane, methylene dichloride and, trichloroethylene were used. ^ When compared to regression estimations, results showed better accuracy of decision trees/ensemble techniques for the categorical case while neural networks were better for estimation of continuous exposure values. Overrepresentation of classes and overfitting were the main causes for poor neural network performance and accuracy. Estimations based on literature-based databases using machine learning techniques might provide an advantage when they are applied to other methodologies that combine `expert inputs' with current exposure measurements, like the Bayesian Decision Analysis tool. The use of machine learning techniques to more accurately estimate exposures from literature-based exposure databases might represent the starting point for the independence from the expert judgment.^

Providing health information to the general public: a survey of current practices in academic health sciences libraries*

A questionnaire was mailed to 148 publicly and privately supported academic health sciences libraries affiliated with Association of American Medical Colleges (AAMC)–accredited medical schools in the United States and Canada to determine level of access and services provided to the general public. For purposes of this study, “general public” was defined as nonaffiliated students or health care professionals, attorneys and other nonhealth-related professionals, patients from affiliated or other hospitals or clinics, and general consumers. One hundred five (71%) libraries responded. Results showed 98% of publicly supported libraries and 88% of privately supported libraries provided access to some or all of the general public. Publicly supported libraries saw greater numbers of public patrons, often provided more services, and were more likely to circulate materials from their collections than were privately supported libraries. A significant number of academic health sciences libraries housed a collection of consumer-oriented materials and many provided some level of document delivery service, usually for a fee. Most allowed the public to use some or all library computers. Results of this study indicated that academic health sciences libraries played a significant role in serving the information-seeking public and suggested a need to develop written policies or guidelines covering the services that will be provided to minimize the impact of this service on primary clientele.

Accounting for residential and non-residential environments to measure contextual effects on health behavior : the case of recreational walking behavior

Thèse réalisée en cotutelle entre l'Université de Montréal et l'Université Pierre et Marie Curie, Paris 06, Sorbonne Universités.

Frieze -- Health Sciences

Frieze located in front of building. Smith, Hinchman & Grylls, architects. W.B. Wood Co., construction

Coronary heart disease risk factors among tri-ethnic students at Florida International University

The present study identified and compared Coronary Heart Disease (CHD) risk factors quantified as “CHD risk point standards” (CHDRPS) among tri-ethnic (White non-Hispanic [WNH], Hispanic [H], and Black non-Hispanic [BNH]) college students. All 300 tri-ethnic subjects completed the Cardiovascular Risk Assessment Instruments and had blood pressure readings recorded on three occasions. The Bioelectrical Impedance Analysis (BIA) was used to measure body composition. Students' knowledge of CHD risk factors was also measured. In addition, a 15 ml fasting blood sample was collected from 180 subjects and blood lipids and Homocysteine (tHcy) levels were measured. Data were analyzed by gender and ethnicity using one-way Analysis of Variance (ANOVA) with Bonferroni's pairwise mean comparison procedure, Pearson correlation, and Chi-square test with follow-up Bonferroni's Chi-square tests. ^ The mean score of CHDRPS for all subjects was 19.15 ± 6.79. Assigned to the CHD risk category, college students were below-average risk of developing CHD. Males scored significantly (p < 0.013) higher for CHD risk than females, and BNHs scored significantly (p < 0.033) higher than WNHs. High consumption of dietary fat saturated fat and cholesterol resulted in a high CHDRPS among H males and females and WNH females. High alcohol consumption resulted in a high CHDRPS among all subjects. Mean tHcy ± SD of all subjects was 6.33 ± 3. 15 μmol/L. Males had significantly (p < 0.001) higher tHcy than females. Black non-Hispanic females and H females had significantly (p < 0.003) lower tHcy than WNH females. Positive associations were found between tHcy levels and CHDRPS among females (p < 0.001), Hs (p < 0.001), H males (p < 0.049), H females (p < 0.009), and BNH females (p < 0.005). Significant positive correlations were found between BMI levels and CHDRPS in males (p < 0.001), females (p < 0.001), WNHs (p < 0.008), Hs (p < 0.001), WNH males (p < 0.024), H males (p < 0.004) and H females (p < 0.001). The mean knowledge of CHD questions of all subjects was 71.70 ± 7.92 out of 100. The mean knowledge of CHD was significantly higher for WNH males (p < 0.039) than BNH males. A significant inverse correlation (r = 0.392, p < 0.032) was found between the CHD knowledge and CHDRPS in WNH females. The researcher's findings indicate strong gender and ethnic differences in CHD risk factors among the college-age population. ^

Risk factors for type 2 diabetes mellitus and cardiovascular diseases in Hispanic adolescents

This cross-sectional study evaluated risk factors (RF) for type 2 diabetes (T2DM) and cardiovascular diseases (CVD) in 100 Hispanic adolescents(50 overweight, 50 non-overweight) aged 12-16 years, and their associations with body mass index (BMI), diet, physical activity (PA), gender, and birth weight (BW). The RF studied were fasting plasma glucose (FPG), insulin sensitivity (IS), total cholesterol (TC), triacylglycerols (TG), low-density lipoprotein cholesterol (LDL), high-density lipoprotein cholesterol (HDL), acanthosis nigricans (AN), and blood pressure (BP). Dietary intakes were assessed using the Block Kids Questionnaire, fat-related intake behavior (FB) using the Fat-Related Diet Habits Questionnaire, and PA using the Modifiable Activity Questionnaire for Adolescents. Blood was collected after an overnight fast of 12 hours. All statistical analyses used SPSS 14.0. Overweight adolescents had presence of AN, higher BP, TC, TG, and LDL, and lower IS, ps < .001, as compared to non-overweight adolescents. Overweight adolescents were more likely to have 1 and 2 RF for T2DM and CVD as compared to having 0, ps < .001, and 2 RF as compared to having 1, p =.033. Adolescents with kilocalorie (Kcal) intake above requirements for age gender, and PA level were 4.6 times more likely to be overweight, p = .005. Overweight adolescents had worse FB, p = .011, and lower PA, p < .001. Adolescents with worse FB had higher BP, p = .016. Fiber below recommendations (14g/1,000 Kcal) was associated with being overweight, p = .012, and lower IS, p = .040. Adolescents with higher BW had higher FPG, p = .013. Our findings point to an association between being overweight and RF for T2DM and CVD, suggesting that overweight during adolescence may have serious health consequences for Hispanic adolescents. Also, our results indicate that Hispanic overweight adolescents eat more Kcal and less fiber than required, have worst FB, and less PA levels than their non-overweight counterparts. In addition, high BW and dietary habits of Hispanic adolescents, such as low fiber and FB, increase their risk for T2DM and CVD. We conclude that BMI can serve as a useful tool to identify Hispanic adolescents at risk for T2DM and CVD.

The experiences of Cuban American women attending a Hispanic serving institution and the influences on identity development

The purpose of this study was to gain an understanding and gather insight into the experiences of Cuban American women attending a 4-year, public, Hispanic Serving Institution and how those experiences influenced their identity development. This was accomplished by conducting in-depth interviews and focus groups with 12 self-identified Cuban American women who were classified as sophomores, juniors, seniors, or graduate students. All of the participants had attended Florida International University for at least 1 year. The women had varying degrees of on and off campus academic and campus involvement activities. Participants were asked about six topics: (a) family, (b) cultural influences, (c) gender, (d) ethical and moral development, (e) education, and (f) ethnic identity. Based on the coding of the data provided by the participants, several interconnected themes emerged including the importance of family, familial support, cultural pride, expected gender roles, core values, decision making, biculturalism, and the value of attending a Hispanic Serving Institution. These themes were found to be all related to the identity development of the participants. It was found that looking at identity through a multidimensional lens is essential. Looking at personal growth and development through anthropological, sociological, and psychosocial lenses gave greater insight to a population of students who have been largely underrepresented in the literature. The findings of this case study are that culture is contextual and identity development is complex for first and second generation Cuban American women attending a Hispanic Serving Institution in a majority minority city. It was found that several factors, including the importance of family and gender roles, were not found to be more important than one another; rather they supported each other in regards to the participants' identity development. The notion of biculturalism as it has been presented in the literature was challenged in this study as it was found that the participants' experiences living and attending a school in a majority minority city presented a new way of understanding what it might mean to be bicultural. For professionals in the field, the findings of this study may lead to a broader understanding of nuances within the Hispanic community and a better understanding of the distinctiveness of what it means to be a Cuban American woman.

Relationships among dietary intake, dietary patterns, dietary quality and nutrient adequacy and substance use in Latina mother/daughter dyads

This study examined the relationships among dietary intake, substance use, socioeconomic and acculturation-related factors among Latinas in Miami-Dade County. Substance abuse is rising among Latinas. A fuller understanding of this problem is needed given the rise of the Hispanic population and the role of women in Latin society. A better understanding between substance use and dietary intake can guide nutrition interventions to reduce negative substance-related health consequences. A purposeful sample of 320 Latina mother/daughter dyads were recruited and interviewed face-to-face as part of the Latino Women's Study. Dietary intake was collected via a 24-hour recall and examined by (1) nutrient intake, (2) dietary patterns using cluster analysis, (3) quality of diet using the Healthy Eating Index (HEI) and (4) the Dietary Reference Intakes to determine nutrient adequacy. Substance use was measured with the Drug Use Frequency and the Healthy and Daily Living Form. Acculturation was measured with the Cultural Identity Scale. Three dietary patterns emerged based on the number of servings from the food groups established in MyPyramid. None were associated with substance use. Latinas who reported using cannabis, cocaine, sedatives without prescription and/or more than five alcoholic drinks on an occasion at least once a month during the previous twelve months had significantly lower HEI scores (64 vs. 60; F = 7.8, p = .005) and consumed fewer fruits (F = 16, p < .001) than non-users. Latinas classified as mothers whom reported consuming cannabis at least 1-7 times a week had significantly lower HEI scores (F = 4.23, p = .015, η2 = .027) than daughters with the same frequency of substance use. One dimension of acculturation, greater familiarity with Latin culture, was associated with good dietary quality (β = .142, p = .012) regardless of any type of substance used or income level. There was a high prevalence of inadequacy of folic acid intake (50-75%) regardless of substance use. Substance users consumed significantly more energy (1,798 vs. 1,615; p = .027) than non-users. Although effect sizes were small, associations between dietary intake and substance use among Latinas deserve further exploration while acknowledging the combined association with acculturation. ^

The civic achievement gap: A study on the civic knowledge, skills, and attitudes of Hispanic students in Miami-Dade County Public Schools

This study assessed the civic knowledge, skills, and attitudes of Hispanic eighth grade students in Miami-Dade County Public Schools (M-DCPS), Florida. Three hundred sixty one Hispanic students of Cuban (253), Colombian (57), and Nicaraguan (51) ancestry from 10 middle schools participated in the study. Two hundred twenty eight students were from low socio-economic status (SES) background, and 133 were of middle SES background. There were 136 boys and 225 girls. The International Association for the Evaluation of Educational Achievement Civic Education Student Questionnaire was used to collect data. The instrument assessed the students’ civic knowledge, skills, and attitudes. Multivariate analysis of variance was used to test for differences in the civic knowledge, skills, and attitudes of participants based on ancestry, SES, and gender. ^ The findings indicated that there was no significant difference in the civic knowledge, skills, and attitudes of Hispanic eighth grade students that were of Cuban, Colombian, and Nicaraguan ancestry. There was no significant difference in the civic skills and in five of the civic attitude scales for students from low SES families compared to those from middle SES families. However, there was a significant difference in the civic knowledge and in the civic attitude concerning classroom discussions and participation based on SES. The civic knowledge of middle SES students was higher than that of low SES students. Furthermore, middle SES Hispanic students displayed a higher mean score for the civic attitude of classroom discussions and participation than low SES students. There was no significant difference in the civic knowledge and in five of the civic attitude scales between boys and girls. However, there was a significant difference in the civic skills and the civic attitude of support for women’s rights between boys and girls. Hispanic girls displayed a higher mean score in civic skills than Hispanic boys. Furthermore, the mean score of civic attitude of support for women’s rights for Hispanic girls was higher than that of Hispanic boys. ^ It was concluded that Cuban, Colombian, or Nicaraguan participants did not demonstrate differences in civic attitudes and levels of civic knowledge and skills that eighth grade students possessed. In addition, when compared to boys, girls demonstrated a higher level of civic skills and a greater support for women’s rights and participation in politics and their roles in politics. Moreover, SES was demonstrated to be a key factor in the acquisition of civic knowledge, regardless of ancestry.^

Risk factors and associations for hepatitis C infection among Hispanic/Latino intravenous drug users in Miami-Dade County, Florida

Hepatitis C infection (HCV) continues to disproportionately affect Hispanics/Latinos in the United States. Hispanic/Latino intravenous drug users (IDUs), because of their risky injection and sexual behaviors, are prone to HCV infection and rapid transmission of the virus to others via several routes. With a prevalence rate of approximately 75% among IDUs, it is imperative that transmission of HCV be prevented in this population. This study aims to examine the associations between demographic, injection and sexual risk factors to HCV infection in a group Hispanic/Latino IDUs in Miami-Dade County, Florida. Preliminary unadjusted results in this sample reveal that age (OR=4.592, p=0.004), weekly injection (OR=5.171, p=0.000), daily injection frequency (OR=3.856, p=0.000) and use of a dirty needle (OR=2.320, p= 0.006) were all significantly associated with HCV infection. Being born outside the U.S. was significantly negatively associated with HCV infection (OR=0.349, p=0.004). Additionally, having two or more sex partners in the past three months (OR=0.472, p=0.014) was negatively associated with HCV infection. After adjusting for all other variables, older age (AOR=7.470, p=0.006), weekly injection (AOR=3.238, p=0.007) and daily injection frequency (AOR=2.625, p=0.010) were all significantly associated with HCV infection. Being born outside the U.S. (AOR=0.369, p=0.019) was a significant protective factor for HCV infection, along with having two or more sex partners in the past three months (AOR=0.481, p=0.037). When analyzing the significant variables in a backward regression model, having 2 or more sex partners in the past three months was not significant at the p

An investigation of the relationship between maltreatment and alcohol and other drug (AOD) problems in African -American and Hispanic adolescent girls

Maltreatment experienced in childhood or adolescence is a known risk factor for later problem alcohol and/or other drug (AOD) use (Bailey & McCloskey, 2005; Shin, Edwards, Heeren, 2009). A growing body of empirical work has found significant associations between adolescent girls’ AOD use and maltreatment experiences. However, questions remain as to how this relation unfolds with African-American and Hispanic adolescent girls. Guided by four relational models that have been proposed in the literature, this study examined the links between maltreatment, trauma symptoms, and alcohol and/or other drug (AOD) problems in a sample of 170 African-American and Hispanic adolescent girls who were participants in a school-based AOD use intervention. Results of this study revealed that maltreatment experiences (physical and emotional abuse) were positively related to trauma symptoms, which were positively related to AOD problem severity, alcohol abuse, alcohol dependency, drug abuse, and drug dependency. Perceived discrimination moderated this relation between sexual abuse and trauma symptoms, such that more perceived discrimination resulted in a stronger effect of sexual abuse on trauma symptoms. Ethnic identity moderated the relation between sexual abuse and AOD problem severity, such that ethnic identity demonstrated protective properties in the relation between sexual abuse and AOD problem severity. My research adds to extant knowledge on the relation between maltreatment and AOD use in adolescent girls and suggests the importance of developing interventions targeting maltreatment and AOD use concurrently.

Accounting for residential and non-residential environments to measure contextual effects on health behavior : the case of recreational walking behavior

Thèse réalisée en cotutelle entre l'Université de Montréal et l'Université Pierre et Marie Curie, Paris 06, Sorbonne Universités.

The Relationship of Diet Quality and Blood Serum Lipid Levels in a Population at High Risk for Diabetes: The Strong Heart Family Study

Thesis (Master's)--University of Washington, 2016-08

Subject response rates in case-control studies of cancer : time trends, study design determinants, and quality of reporting

Objectifs: Examiner les tendances temporelles, les déterminants en lien avec le design des études et la qualité des taux de réponse rapportés dans des études cas-témoins sur le cancer publiées lors des 30 dernières années. Méthodes: Une revue des études cas-témoins sur le cancer a été menée. Les critères d'inclusion étaient la publication (i) dans l’un de 15 grands périodiques ciblés et (ii) lors de quatre périodes de publication (1984-1986, 1995, 2005 et 2013) couvrant trois décennies. 370 études ont été sélectionnées et examinées. La méthodologie en lien avec le recrutement des sujets et la collecte de données, les caractéristiques de la population, les taux de participation et les raisons de la non-participation ont été extraites de ces études. Des statistiques descriptives ont été utilisées pour résumer la qualité des taux de réponse rapportés (en fonction de la quantité d’information disponible), les tendances temporelles et les déterminants des taux de réponse; des modèles de régression linéaire ont été utilisés pour analyser les tendances temporelles et les déterminants des taux de participation. Résultats: Dans l'ensemble, les qualités des taux de réponse rapportés et des raisons de non-participation étaient très faible, particulièrement chez les témoins. La participation a diminué au cours des 30 dernières années, et cette baisse est plus marquée dans les études menées après 2000. Lorsque l'on compare les taux de réponse dans les études récentes a ceux des études menées au cours de 1971 à 1980, il y a une plus grande baisse chez les témoins sélectionnés en population générale ( -17,04%, IC 95%: -23,17%, -10,91%) que chez les cas (-5,99%, IC 95%: -11,50%, -0,48%). Les déterminants statistiquement significatifs du taux de réponse chez les cas étaient: le type de cancer examiné, la localisation géographique de la population de l'étude, et le mode de collecte des données. Le seul déterminant statistiquement significatif du taux de réponse chez les témoins hospitaliers était leur localisation géographique. Le seul déterminant statistiquement significatif du taux de participation chez les témoins sélectionnés en population générale était le type de répondant (sujet uniquement ou accompagné d’une tierce personne). Conclusion: Le taux de participation dans les études cas-témoins sur le cancer semble avoir diminué au cours des 30 dernières années et cette baisse serait plus marquée dans les études récentes. Afin d'évaluer le niveau réel de non-participation et ses déterminants, ainsi que l'impact de la non-participation sur la validité des études, il est nécessaire que les études publiées utilisent une approche normalisée pour calculer leurs taux de participation et qu’elles rapportent ceux-ci de façon transparente.

Subject response rates in case-control studies of cancer : time trends, study design determinants, and quality of reporting

Objectifs: Examiner les tendances temporelles, les déterminants en lien avec le design des études et la qualité des taux de réponse rapportés dans des études cas-témoins sur le cancer publiées lors des 30 dernières années. Méthodes: Une revue des études cas-témoins sur le cancer a été menée. Les critères d'inclusion étaient la publication (i) dans l’un de 15 grands périodiques ciblés et (ii) lors de quatre périodes de publication (1984-1986, 1995, 2005 et 2013) couvrant trois décennies. 370 études ont été sélectionnées et examinées. La méthodologie en lien avec le recrutement des sujets et la collecte de données, les caractéristiques de la population, les taux de participation et les raisons de la non-participation ont été extraites de ces études. Des statistiques descriptives ont été utilisées pour résumer la qualité des taux de réponse rapportés (en fonction de la quantité d’information disponible), les tendances temporelles et les déterminants des taux de réponse; des modèles de régression linéaire ont été utilisés pour analyser les tendances temporelles et les déterminants des taux de participation. Résultats: Dans l'ensemble, les qualités des taux de réponse rapportés et des raisons de non-participation étaient très faible, particulièrement chez les témoins. La participation a diminué au cours des 30 dernières années, et cette baisse est plus marquée dans les études menées après 2000. Lorsque l'on compare les taux de réponse dans les études récentes a ceux des études menées au cours de 1971 à 1980, il y a une plus grande baisse chez les témoins sélectionnés en population générale ( -17,04%, IC 95%: -23,17%, -10,91%) que chez les cas (-5,99%, IC 95%: -11,50%, -0,48%). Les déterminants statistiquement significatifs du taux de réponse chez les cas étaient: le type de cancer examiné, la localisation géographique de la population de l'étude, et le mode de collecte des données. Le seul déterminant statistiquement significatif du taux de réponse chez les témoins hospitaliers était leur localisation géographique. Le seul déterminant statistiquement significatif du taux de participation chez les témoins sélectionnés en population générale était le type de répondant (sujet uniquement ou accompagné d’une tierce personne). Conclusion: Le taux de participation dans les études cas-témoins sur le cancer semble avoir diminué au cours des 30 dernières années et cette baisse serait plus marquée dans les études récentes. Afin d'évaluer le niveau réel de non-participation et ses déterminants, ainsi que l'impact de la non-participation sur la validité des études, il est nécessaire que les études publiées utilisent une approche normalisée pour calculer leurs taux de participation et qu’elles rapportent ceux-ci de façon transparente.