Epidemiology & Biostatistics Theses and Dissertations

Permanent URI for this collection


Recent Submissions

Now showing 1 - 20 of 71
  • Item
    (2023) Lai, Jianyu; Milton, Donald K; Epidemiology and Biostatistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)
    Acute respiratory infections (ARIs), which usually appear in the form of common colds and influenza, as well as outbreak brought on by emerging viruses such as SARS-CoV-2, result in millions of deaths and hospitalizations each year. Aerosols being exhaled by infected population and inhaled by susceptible population has been identified as an important transmission route for ARIs; yet few studies have compared the viral load in exhaled breath aerosol (EBA) in naturally and experimentally infected cases, as well as among various infections. The specific aims of this dissertation were: 1) examine the comparability of EBA viral shedding between experimentally and a full range of natural ambulatory influenza cases; 2) compare seasonal coronavirus, influenza, SARS-CoV-2 Omicron, and other SARS-CoV-2 variants in terms of viral loads in exhaled breath aerosols; and 3) examine the relative efficacy of face masks, KN95, and N95 respirators as source control for SARS-CoV-2. We analyzed data from four studies that collected 30-minute fine (≤5 μm) and coarse (>5 μm) EBA samples using a Gesundheit-II sampler. Viral RNA load in EBA was quantified using real-time RT-PCR. Nasal inoculation of influenza virus A/Wisconsin/67/2005 showed lower EBA viral shedding compared to the natural influenza A H3 infections. Among the viruses studied, SARS-CoV-2 Omicron variants demonstrated the highest viral RNA loads in both EBA size fractions, emphasizing its superior spread capability via inhalation. Furthermore, while all masks and respirators showed significant reductions in viral RNA load in exhaled aerosols, the duckbill N95 respirators stood out, providing reductions of up to 99% and outperforming both surgical and cloth masks, and KN95 respirators. Given the evident transmission risk via inhalation for the studied viruses, measures such as masking and indoor air hygiene are crucial. The pronounced efficacy of N95 respirators highlights their importance in healthcare settings and places with vulnerable populations, especially during periods of heightened respiratory viral infections.
  • Item
    (2023) Ogbenna, Bethany Townsend; Dallal, Cher; Epidemiology and Biostatistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)
    Background: Observed racial and ethnic disparities in breast cancer are complex, in part, due to biological and behavioral factors at the individual and neighborhood level. Cardiometabolic factors such as the use of cholesterol-lowering drugs and engaging in healthy lifestyle behaviors may reduce breast cancer risk, however, the current understanding of these factors among diverse racial and ethnic populations remains limited. Moreover, at the neighborhood-level, the extent to which neighborhood socioeconomic status (nSES) influences inflammatory profiles among racially and ethnically diverse populations remains unclear. Using data from the Multiethnic Cohort Study (MEC), this dissertation investigates cholesterol-lowering drug use (Aim I) and a Healthy Lifestyle Index (HLI) (Aim II) in relation to postmenopausal breast cancer risk by race and ethnicity; and, assesses associations between nSES and inflammatory biomarkers among adults (Aim III). Methods: Prospective cohort analyses were conducted among postmenopausal women who completed the third MEC follow-up questionnaire in 2003 (Aim 1, n=41,394) or the baseline questionnaire in 1993-1996 (Aim 2, n=65,561) and were followed until 2017 for invasive breast cancer diagnoses (n=1,681 and 4,555 cases, respectively). Multivariable adjusted hazard ratios [HR] and 95% confidence intervals [95% CI] were estimated using Cox proportional hazards regression. For Aim III, multivariable linear regression assessed cross-sectional associations between nSES and inflammatory serum biomarkers (adiponectin, leptin and C-reactive protein) among adults residing in California (n=6,919) and Hawaii (n=6,899) (2000-2017). Results: Cholesterol-lowering drug use (Aim 1) and duration was not associated with breast cancer risk among all women with no statistically significant heterogeneity in associations by race and ethnicity (p-interaction >0.05). In Aim 2 analyses, women with a higher HLI score (Tertile (T)) had a reduced risk of breast cancer (HRT3 vs T1: 0.76; 95% CI: 0.69, 0.84; HRT2 vs T1: 0.88; 95% CI: 0.79, 0.97) compared to women in the lowest HLI tertile with a significant dose-response observed (p-trend <0.01). Similar patterns were observed across all racial and ethnic groups of women. In California and Hawaii, individuals living in low nSES neighborhoods had higher serum levels of CRP (p-trend <0.001; p-trend = 0.02, respectively) and leptin (p-trend <0.001) while adiponectin levels were lower (p-trend <0.01; p-trend = 0.03, respectively) compared to individuals living in neighborhoods with high nSES. Additional adjustment for body mass index attenuated these associations (p-trend >0.05) (Aim III). Public Health Impact: Findings from this dissertation further support engaging in healthy lifestyle behaviors as a preventative strategy for breast cancer reduction among multiethnic populations of postmenopausal women whereas cholesterol-lowering drug use was not associated with reductions in risk. In addition, residing in low nSES neighborhoods was associated with less favorable inflammatory biomarkers levels.
  • Item
    (2023) Yu, Weijun; Nguyen, Quynh C.; Epidemiology and Biostatistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)
    This dissertation investigates the impact of international and local mobility of asymptomatic versus symptomatic COVID-19 cases on the pandemic in Hong Kong.The first manuscript analyzed empirical data from 11,775 confirmed COVID-19 cases in Hong Kong from January 2020 to April 2021, building a retrospective cohort. The results indicated that COVID-19 asymptomatic airport or flight crew were ten times more likely to have inbound air travel history than symptomatic airport or flight crew (adjusted RR=10.00, 95% CI: 4.00–25.00), and the median flight duration of asymptomatic cases was 4.6 person-hours shorter than that of symptomatic cases (p<0.01). The second manuscript presented a social network analysis study that build networks for the three peaks of COVID-19 diagnosis in Hong Kong. The results showed that asymptomatic cases were 1.33 times more likely to be presented in the inbound flight cabin or airport with other COVID-19 cases simultaneously than symptomatic cases (95%CI: 1.21-1.45) at the early stage of the pandemic. Additionally, the study found that network percolation simulation targeted attacks were more efficient than random failures in dismantling networks with a low level of connectedness. The third manuscript used geocoded COVID-19 cases’ travel records in Hong Kong to conduct a spatial analysis study. The findings indicated that asymptomatic cases visited locations mostly clustered in the southern part of Hong Kong, while symptomatic cases visited locations mostly clustered in the middle and southern parts of Hong Kong. This study also found that Geographically Weighted Regression models performed better among symptomatic cases than asymptomatic cases, and median local travel time was higher (p<0.01) among asymptomatic (68.09 person-minutes) than symptomatic cases (59.46 person-minutes) based on 19,568 Origin-Destination Cost Matrix least-cost paths. Overall, this dissertation highlights the importance of promoting public health prevention strategies to contain future infectious disease pandemics at the early stage, regardless of the presence of symptoms. Moreover, it suggests that travel restriction may not be effective in dismantling networks with a low-level of connectedness. Local health authorities and policymakers should tailor detection and containment strategies based on spatial variability in different areas.
  • Item
    Understanding the Risk of Drug Overdose and Alcohol-Induced Deaths Among Adults with Different Types of Disabilities
    (2023) Aram, Jonathan; Dallal, Cher M; Epidemiology and Biostatistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)
    Background: Disability is associated with drug and alcohol morbidity and mortality, which have reached high levels in recent years. Previous disability studies often combine all disabilities into a single category or focus on a single type of limitation. This dissertation characterizes different types of disabilities among U.S. adults and assesses associations with drug and alcohol morbidity and mortality. Methods: Using the 2018-2019 National Survey on Drug Use and Health (NSDUH, Aim 1), a nationally representative cross-sectional survey (n=83,485), different individual disabilities and co-occurring disabilities were identified. Multivariable logistic regression was used to estimate adjusted odds ratios (aORs) and 95% confidence intervals (CIs) for associations between disability type and prevalence of drug and alcohol use disorders. Disabilities were also characterized within the Mortality Disparities in American Communities Study (Aims 2 and 3), a nationally-representative prospective cohort with baseline data collected in 2008 and mortality follow-up through 2019 (n=3,324,000). Multivariable Cox proportional hazards regression was used to estimate adjusted hazard ratios (aHRs) and CIs for associations between disability type and drug overdose death (OD, Aim 2) and alcohol-induced death (AID, Aim 3). Results: Within the NSDUH analysis, adults with cognitive disability had increased odds of drug (aOR=3.3; 95% CI=2.9–3.8), and alcohol use disorder (aOR=2.3; 95% CI=2.0–2.6), compared to adults without disability. Positive associations of lesser magnitude were observed between hearing/seeing and ambulatory disabilities and drug use disorder. In MDAC analyses, OD risk was elevated among adults with cognitive (aHR=2.6; 95% CI=2.4–2.9), ambulatory (aHR=2.8; 95% CI=2.6–3.1), ambulatory and hearing/seeing (aHR=2.5; 95% CI=2.0–3.1), and hearing/seeing disability (aHR=1.6; 95% CI=1.4–1.9), compared to adults without disability. The risk of AID was elevated for adults with co-occurring ambulatory and hearing/seeing disability (aHR=1.8; 95% CI=1.5–2.2), ambulatory disability only (aHR=1.5; 95% CI=1.3–1.7), and hearing/seeing disability only (aHR=1.2; 95% CI=1.0–1.4). Conclusions: The examination of specific disability categories reveals unique associations that are not apparent when all disabilities are combined. These findings can be used to improve access to recovery support services. Expansion of educational and occupational opportunities for adults with disabilities should be considered as strategies to reduce drug and alcohol morbidity and mortality.
  • Item
    Effects of Tamoxifen Therapy on Breast Carcinogenesis: Epidemiological Associations and Biomechanisms of Action
    (2022) Ghosh, Rajrupa; Dallal, Cher; Epidemiology and Biostatistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)
    Background: Tamoxifen, a key chemopreventive and adjuvant endocrine therapy (ET) for breast cancer, is suggested to alter breast cancer risk factors including circulating hormones (estrogen and insulin-like growth factors (IGFs)) and breast density. However, the biological underpinnings of tamoxifen’s effect on these factors remain unclear. This dissertation evaluated effects of tamoxifen on estrogen metabolites (EMs), explored associations between circulating IGFs (IGF-I and IGFBP-3) and volume average sound speed measures (VASS) of breast density, and synthesized evidence from real-world studies to meta-analyze adjuvant ET in relation to contralateral breast cancer (CBC) risk. Methods: Within the Ultrasound Study of Tamoxifen, serial serum samples collected prior to and 12 months after tamoxifen treatment were used to assess longitudinal changes (paired t-tests) in 15 circulating EMs among postmenopausal women (n=23) (Aim 1), and changes in IGFs and VASS (n=53) (Aim 2). Multivariable linear regression examined associations between metabolites of tamoxifen and estrogen among pre- (n=33) and postmenopausal women (n=27) (Aim 1) and concomitant changes in IGFs and VASS (n=53) (Aim 2), 12 months after treatment initiation. In Aim 3, a random effects meta-analysis of observational studies (n=17, 287,576 participants) estimated relative risks (RR) and 95% confidence intervals (CI) for associations between ET and CBC risk among primary breast cancer patients overall, by menopausal status and CBC estrogen receptor (ER)-subtype. Results: Circulating 2-OH and 16-OH pathway EMs, IGF-I, and IGF-I:IGFBP-3 decreased 12 months after tamoxifen initiation (p <0.05; Aims 1 and 2). No associations were observed between concomitant changes in IGFs and VASS among tamoxifen-treated patients (Aim 2). In meta-analyses (Aim 3), endocrine therapy was associated with reduced CBC risk (RR: 0.62, 95% CI: 0.53, 0.73), with a greater reduction observed among premenopausal (RR: 0.58, 95% CI: 0.43, 0.78) versus postmenopausal women (RR: 0.72, 95% CI: 0.60, 0.87). Endocrine therapy reduced the risk of ER-positive (RR: 0.55, 95% CI: 0.43, 0.70) but not ER-negative CBC. Conclusion: The tamoxifen mediated decline of circulating 2-OH, 16-OH and IGF-I provides etiologic insight into the biomechanisms of tamoxifen on breast carcinogenesis. Meta-analyses of observational studies further support a chemopreventive role of endocrine therapy on CBC risk, particularly, ER-positive CBC.
  • Item
    Intersectional stigma, self-efficacy, depression, and resilience: a Rasch analysis
    (2022) Reuben, Jacqueline; Liu, Hongjie; Turpin, Rodman; Epidemiology and Biostatistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)
    Background: Black men who have sex with men (BMSM) are disproportionally affected by negative health outcomes associated with stigma related to both their racial and sexual minority status. Foundational to understanding stigma is the ability to correctly measure this latent construct. The Rasch model is a probabilistic model for analyzing categorical data that was developed to improve the measurement of latent traits. This study, grounded in intersectional minority stress theory, reviewed the application of Rasch analysis in the HIV/AIDS literature (Aim 1) and used the Rasch model to calibrate person measures to assess the interrelationships among internalized stigma, resilience, self-efficacy, and psychologic well-being among BMSM (Aims 2 and 3). Methods: For Aim 1, we conducted a systematic review of the literature following PRISMA guidelines. Aims 2 and 3 used data from a cross-sectional online survey of 151 HIV-negative BMSM in 2020. For Aim 2, we conducted Rasch analysis to assess the psychometric properties of the internalized racism (IR), internalized homophobia (IH), self-efficacy, and resilience scales. For Aim 3, we used linear regression and path analysis of the Rasch-calibrated person measures to examine the mediating and modifying effects of self-efficacy and resilience on the relationship between intersectional stigma and depressive symptoms. Results: For Aim 1, after screening 183 articles, 45 articles were included in the analysis. Strengths and weaknesses of using the Rasch approach were summarized. For Aim 2, the final IR scale had a person reliability and separation of 0.91 and 3.13, respectively, and an item reliability and separation of 0.94 and 4.01, respectively. The final IH scale had a person reliability and separation of 0.88 and 2.72, respectively, and an item reliability and separation of 0.79 and 1.95, respectively. For Aim 3, IR (β=0.296, 95% CI [0.133, 0.458]) and IH (β=0.414, 95% CI [0.204,0.623]) were independently and positively associated with depression in multivariable models controlling for age, income, and relationship status. Resilience and self-efficacy modified the relationship between IH and depression (βIHxRES=-0.034, 95% CI [-0.060, -0.008] and βIHxSE=-.056, 95% CI [-0.113, 0.00], respectively), but there was no evidence of effect modification by resilience or self-efficacy on the association between IR and depression. Public health implications: Our findings suggest that interventions targeting multiply marginalized groups such as BMSM that address co-occurring forms of stigma and foster positive self-evaluation and coping skills may reduce the negative consequences of internalized stigma on mental health outcomes.
  • Item
    The effects of HIV criminalization laws (and their enforcement) on HIV risk among Black and Hispanic populations
    (2022) Keralis, Jessica Maciel; Nguyen, Quynh C; Epidemiology and Biostatistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)
    Background: In the U.S., 25 states have laws that explicitly criminalize the transmission or exposure of HIV. This study, grounded in Nancy Krieger's ecosocial theory, estimated the association between HIV criminalization laws and state- (Aim 1) and county-level (Aim 2) HIV incidence rates, as well as individual HIV testing history (Aim 3), and assessed effect modification by overpolicing, using incarceration rates as a proxy. Methods: The study uses data from state- and county-level HIV incidence data from AIDSVu (2010-2019), incarceration data from the Vera Institute of Justice (2010-2018), and HIV testing data from BRFSS (2016-2019). For Aim 1, a longitudinal analysis was conducted using multivariate marginal Poisson GEE models to estimate rate ratios. For Aim 2, count-rate hierarchical (multilevel) models were fitted to estimate rate ratios. For Aim 3, logistic regression models were fitted to estimate odds ratios. Results: The presence of a state HIV testing law was associated with a higher state HIV incidence in the general and Hispanic populations (aRR=1.48 and 1.68, respectively), but higher incarceration at the state level did not significantly modify the relationship between the law and HIV incidence. At the county level, being in a state with an HIV-specific criminalization statute was associated with a higher county-wide HIV incidence rate for all three populations (aRR=1.14, 1.30, and 1.32 for the general, Black, and Hispanic populations, respectively). Unlike the state-level analysis, this association was attenuated by a higher jailed population rate for the general and Black populations. The effect modification was statistically significant for the general population (p=0.01) and marginally significant for the Black population (p=0.06). Finally, the presence of a state HIV testing law (aOR=1.06) was associated with a greater likelihood of HIV testing history in the general population. However, in HIV criminalization states, heavier policing negatively modified the effect of the law on the likelihood of having ever received an HIV test (p<0.01). Public health implications: This study contributes to a gap in the literature by using recent data to estimate the association of HIV criminalization laws and their enforcement with HIV incidence rates and HIV testing behavior.
  • Item
    Metformin and statins in relation to ovarian cancer: implications for risk and potential etiologic pathways
    (2022) Irvin, Sarah Renee; Dallal, Cher; Epidemiology and Biostatistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)
    Background: Metabolic factors, including obesity, diabetes, and circulating lipids are associated with increased risk of ovarian cancer, the most fatal gynecologic malignancy. Drugs targeting these conditions have drawn interest as potential chemopreventive agents, but limited studies have examined relationships with ovarian cancer overall, by possible benefit groups or with regards to related etiologic pathways. This dissertation examined medications that target metabolic abnormalities with risk of ovarian cancer, and expression of a pro-tumorigenic lipid pathway in ovarian cancer tissue.Methods: Nested case-control studies of metformin (Aim 1) and statin (Aim 2) use in relation to ovarian cancer (cases n=9,207 and 10:1 frequency-matched controls n=92,070) were conducted in Clinical Practice Research Datalink (CPRD) (1987-2020), a primary care database in the United Kingdom. Aim 3 was a case-only analysis from The Polish Ovarian Cancer Study (POCS) (2001-2003, n=166) determining expression of sphingosine-1-phosphate receptors S1PR1 and S1PR3 in ovarian cancer tissue. Multivariable logistic regression estimated odds ratios (ORs) and 95% confidence intervals (CI) for associations between medication use and ovarian cancer in CPRD (Aims 1 and 2), and cancer characteristics with marker expression in POCS (Aim 3). Results: Within the CPRD analyses, metabolic conditions such as obesity (OR 1.19, 95% CI 1.11-1.28) and polycystic ovarian syndrome (OR 1.66, 95% CI 1.17-2.35) were significantly associated with increased ovarian cancer risk. However, neither metformin (OR 1.01, 95% CI 0.91-1.13) nor statin use (OR 1.02, 95% CI 0.96-1.08) was associated with ovarian cancer risk. Associations did not differ by drug indications (pint >0.05), individual statin or class. In Aim 3, strong expression of lipid markers in ovarian cancer tissue significantly differed by histotype (p<0.01). Lower marker expression was observed in non-serous tumors compared to serous, but expression was not associated with cancer risk factors or survival. Conclusion: Findings from this dissertation do not support use of metformin or statins as chemopreventive agents. However, as most CPRD users had an indication, it remains unclear whether findings are generalizable to the at-risk population for ovarian cancer. Strong expression of lipid markers in ovarian cancer tissue suggests involvement in carcinogenesis; additional molecular studies are needed to elucidate their role.
  • Item
    Exploring Psychosocial and Structural Syndemic Effects as Predictors for HIV-Related Outcomes among Black Women
    (2022) Watson, Lakeshia; Dyer, Typhanye; Epidemiology and Biostatistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)
    Black women continue to be disproportionately affected by HIV with marked disparities in HIV incidence, prevalence, and care outcomes. There is a critical need to explore the role of psychosocial and structural factors and the negative impact of these factors on HIV-related outcomes, including HIV risk behaviors, HIV medication adherence, and healthcare utilization. This research uses the syndemic theory to measure the effects of both psychosocial and structural syndemics on HIV-related outcomes among Black women. Syndemic theory is a theoretical framework, which posits there are multiple, overlapping factors that form a cumulative and synergistic effect on HIV disease burden. The substance abuse, violence and AIDS (SAVA) syndemic, clustering effects of substance use, intimate partner abuse, and HIV/AIDS, have been applied in studies exploring HIV outcomes among women; however, the SAVA syndemic framework does not include additional psychosocial and structural factors such as food insecurity, housing instability, incarceration, post -traumatic stress disorder, and depression to better understand the syndemic profiles of Black women. Using study data from the HIV Prevention Trials Network 061 Women’s Seroincidence study and the Women’s Interagency HIV Study, we tested whether a cumulative syndemic of psychosocial and structural factors contributed to sexual risk behaviors and HIV care outcomes. The studies will assess three parameterizations of syndemic factors: cumulative syndemic index, syndemic group indices reflecting the level of influence (psychosocial, participant-level, and neighborhood), and number of syndemic groups (0, 1, 2, or 3). We also tested whether HIV status modified the relationships between a cumulative syndemic of psychosocial and structural factors and sexual risk behaviors. In study 1, a higher syndemic score was significantly associated with increased prevalence of unknown HIV status of last male sex partner (aPR = 1.07, 95% CI: 1.04-1.10), engaging in exchange sex (aPR = 1.17, 95% CI: 1.14-1.20), and having multiple sex partners (aPR = 1.07, 95% CI: 1.06-1.09) among a sample of 1,347 Black women. In study 2, generalized linear mixed models found that being in two syndemic groups was associated with increased odds of reporting unknown HIV status of last male sex partners (aOR=3.04, 95% CI: 1.24-7.44) and having multiple sex partners (aOR=4.29, 95% CI: 1.81-10.18) among 1,364 Black women living with and without HIV across twelve follow-up visits. We also found that being in all three syndemic groups was associated with increased odds of reporting inconsistent condom use (aOR= 2.15, 95% CI: 1.28-3.61), unknown HIV status of last male sex partners (aOR=5.26, 95% CI: 1.94-14.25), and having multiple sex partners (aOR=7.47, 95% CI: 2.85-19.58). Among a total of 969 Black women living with HIV in study 3, a higher cumulative syndemic score and a higher neighborhood-level structural syndemic group score was associated with increased odds of reporting sub-optimal HIV medication adherence (aOR=1.04, 95% CI: 1.01-1.06 and 1.08, respectively). Black women in all three syndemic groups had increased odds of reporting sub-optimal HIV medication adherence (OR=2.88, 95% CI: 1.32-6.29) and missed HIV appointments (OR=3.39, 95% CI: 1.06-10.92). Results from these studies highlight the evidence of psychosocial and structural syndemic effects on multiple HIV risk and care outcomes among Black women.
  • Item
    Effects of the 2014-2016 Ebola Epidemic on Infectious Disease Prevention in Guinea
    (2021) Callaway, Julia; Dyer, Typhanye V; Epidemiology and Biostatistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)
    From 2014-2016, Ebola ravaged the three West African countries of Guinea, Sierra Leone, and Liberia. There were more than 28,000 cases and 11,000 deaths, with more than 10,600 additional deaths due to other medical conditions. We aimed to examine how utilization of infectious disease prevention services changed in Guinea over the course of the Ebola epidemic, using Andersen’s model of healthcare utilization as a theoretical framework. The specific aims of this study were: 1) to examine differences in measles vaccination rates among children ages 12-35 months before and after the Ebola epidemic by wealth quintile; 2) to estimate mosquito net possession at two time points post-epidemic compared to pre-epidemic among households with at least one child under five, nationally and regionally; and 3) to quantify how much of the association between exposure to the Ebola epidemic and HIV testing prevalence can be explained by changes in urban/rural residence rates among adults in Guinea. The studies use data from the 2012 and 2018 Guinea DHS, and the 2016 MICS. Studies 1 and 2 use quasi-Poisson regression models to estimate prevalence ratios, and study 3 uses log-binomial regression models in a mediation analysis. In study 1 (n = 2,573 children ages 12-35 months), the poorest children were 54% (95% CI = 58%-67%) as likely to be vaccinated for measles in 2018 compared to 2012, and the wealthiest children were 78% (95% CI = 69%-90%) as likely. In study 2 (n = 14,756 households with at least one child under five), mosquito net possession in 2016 was 72% (95% CI = 56%-90%) higher and in 2018 was 12% (95% CI = 8%-15%) higher than in 2012. In study 3 (n = 27,809 adults), of the 4.59% (95% CI = 4%-6%) increase in the log-likelihood of ever having been tested for HIV due to being in the 2018 cohort, an estimated 0.269% of the effect (95% CI = 0.04%-1%) could be attributed to differences in urban/rural residence. Understanding these changes gives a more complete picture of the effects of epidemics on infectious disease prevention and can help public health officials plan for future epidemics.
  • Item
    (2020) Li, Yuruo; Liu, Hongjie; Epidemiology and Biostatistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)
    This dissertation aims to assess the applicability of the social network method on HIV research among female sex workers (FSWs). Manuscript 1 reported the findings from a systematic literature review which examined the application of social network method in HIV studies focusing on FSWs. The majority of the identified studies were limited to local social networks or FSW establishments and did not use sophisticated statistical approaches to analyze sociocentric network data. The discrepancies in network definitions and data collections made it difficult in interpreting their findings and assessing validity. Most of the analytic plans for egocentric studies were limited to information at the individual level rather than that at the ego-alter ties. The project reported in manuscripts 2 and 3 used empirical data from a multi-center egocentric network study among mid-age FSWs in China to assess the extent to which social network components influence HIV testing behaviors (paper 2), and the associations between Chinese collectivist culture and FSWs’ social networks (paper 3). As reported in the manuscript 2, among 1,245 FSWs, 62.2% of them received an HIV test. HIV testing was positively associated with higher network transitivity (AOR: 1.77; 95% CI: 1.18-2.64) and inversely associated with network trust (AOR: 0.74; 95% CI: 0.56-0.97). Although social support was not associated with HIV testing, the increase in social cohesion may provide substantial support for HIV testing. As documented in manuscript 3, Chinese collectivism tendency was negatively associated with their perceived social support (95% CI: -0.33, -0.04), network effective size (95% CI: -0.30, -0.01), and network betweenness (95% CI: -0.33, -0.09). FSWs who had the highest level of collectivistic tendency and perceived a higher level of stigma are more likely to stay at a “bridging” position and connect with weak social ties rather than a strong cohesive group. This dissertation projects provide empirical evidence that social networks can be used to analyze the social environment of FSWs and its impact on HIV preventive behaviors among this HIV vulnerable population. The findings make additional contributions to the application of social network methods in social and behavioral research with a focus on FSWs.
  • Item
    Biomarker Categorization in Transcriptome Meta-analysis by Statistical significance, Biological Significance and Concordance
    (2020) Ye, Zhenyao; Ma, Tianzhou; Epidemiology and Biostatistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)
    With the advancement of high-throughput technology, transcriptomic studies have been accumulated in the public domain. Meta-analysis combines multiple studies on a related hypothesis and improves the statistical power and reproducibility of single studies. However, a majority of existing meta-analysis methods only consider the statistical significance. We propose a novel method to categorize biomarkers by simultaneously considering statistical significance, biological significance (large effect size), and concordance patterns across studies, accounting for the complex study heterogeneity that exists in most meta-analysis problems. We conducted simulation studies and applied our method to Gynecologic and breast cancer RNA-seq data from The Cancer Genome Atlas to show its strength as compared to adaptively-weighted Fisher’s method. We found several major biomarker categories according to their cross-study patterns, and these categories are enriched in very different sets of pathways, offering different biological functions for future precision medicine.
  • Item
    (2020) xu, yixi; Dyer, Typhanye; Epidemiology and Biostatistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)
    Hispanic/Latino gay and bisexual men (HLGBM) are especially vulnerable to HIV acquisition compared to their heterosexual counterparts. In an era of treatment as prevention, HIV testing is a crucial point to link individuals with HIV infection to treatment and healthcare services. In this dissertation, I addressed three specific aims: 1) to assess the prevalence of mental health conditions and socioeconomic risk factors among HLGBM comparing those who have and have not been tested for HIV; 2) to examine the applicability of syndemic theory to HIV testing among HLGBM; and, 3) to assess whether race/ethnicity modifies the association between health care access and socioeconomic factors with HIV testing in gay and bisexual men (GBM). In Aim 1, bivariate associations indicated that depression (prevalence ratio [PR]=1.36; 95% CI: 1.12, 1.64) and frequent high stress (PR=1.23; 95% CI: 1.02, 1.49) were associated with a higher prevalence of HIV testing, whereas poverty was associated with a lower prevalence of HIV testing (PR=0.64, 9% CI: 0.55, 0.75). In an adjusted model that included all mental health and demographic variables including age, marital status, health insurance status, access to a personal doctor, and education, only poverty status maintained an association with HIV testing at p<.05 (PR=0.77, 95% CI:0.65, 0.92). In Aim2, results from interaction tests supported the application of syndemic theory to HIV testing (p-values <.05 for all pair-wise interactions between risk factors). We used strata-specific estimates to display the synergistic relationships between combinations of risk factors, adjusted for demographic characteristics. For example, individuals who were poor and had a mental health condition (i.e., depressive disorder, heavy alcohol consumption, frequent high stress) had a lower prevalence of HIV testing relative to those with poverty or a mental health condition alone (e.g. among HLGBM living in poverty, those who were heavy drinkers had 0.16 (95% CI: 0.05, 0.54) times the prevalence of having an HIV test compared to those were not in poverty and not heavy drinkers). In Aim 3, analyses revealed that race/ethnicity modified the associations between health care access and socioeconomic factors with HIV testing (all p-values <.05). In adjusted models stratified by race/ethnicity, poverty was associated with HIV testing among Black GBM (PR=1.21; 95%CI 1.06, 1.38) and White GBM (PR=0.86; 95% CI: 0.80, 0.93) in opposite directions; and, having a personal doctor was associated with a higher prevalence of HIV test among Hispanic/Latino GBM only (PR=1.30; 95% CI: 1.10, 1.53). Taken together, results from these studies suggest that sociodemographic factors and mental health conditions facing HLGBM work in tandem and contribute to syndemic conditions; being White and having insurance, having a personal doctor, as well as higher household income were protective, which advance knowledge about HIV testing among GBM. Findings from this study further support addressing racial disparities in health care access and improving socioeconomic conditions, which together may promote HIV testing uptake among high-risk populations.
  • Item
    Self-reported discrimination among Asian Americans: An examination of its measurement and relationship with health-related quality of life
    (2019) Jung, Mary; Lee, Sunmin; Epidemiology and Biostatistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)
    Labeled as a “model minority,” Asian Americans have often been excluded from the dialogue on discrimination. Previous studies frequently used discrimination measures that were initially developed for African Americans and lack items related to language and nativity. Although discrimination adversely affects mental and physical health, its relationship to health-related quality of life (HRQOL) has been understudied in this population. This dissertation evaluated self-reported discrimination among Asian Americans by examining its measurement and relationship with HRQOL. Manuscript 1 assessed the psychometric properties of everyday (ERDS) and major racial discrimination scales (MRDS) that were developed for Asian Americans. In a sample of 569 Asian American immigrant adults, reliability was excellent for ERDS but moderate for MRDS (Cronbach’s α=0.94 and 0.60). The latter is likely low given that MRDS is an inventory of events. For construct validity, weak positive linear correlations with perceived stress and depressive symptoms were observed (r=0.22 and 0.28 for ERDS and r=0.11 and 0.14 for MRDS; p<0.001). Factor analysis confirmed the unidimensionality of both measures. Manuscript 2 cross-sectionally examined the association between the racial discrimination measures from Manuscript 1 and HRQOL among 524 foreign-born Asian American adults. Multivariable logistic regression and negative binomial regression were performed to examine self-rated health (poor vs. good) and days of poor physical health, mental health, and activity limitation. ERDS (IRR range: 1.04-1.08) and MRDS scores (IRR range: 1.19-1.61) were significantly and positively associated with worse HRQOL for all measures, except for self-rated health. Manuscript 3 examined discrimination trajectories and assessed their relationship with self-rated health in a longitudinal, multi-ethnic sample of 2,004 middle-aged women in the Study of Women’s Health Across the Nation (SWAN). Group-based trajectory analysis identified three distinct discrimination groups. The “high and decreasing” but not “moderate and decreasing” group was significantly associated with poor self-rated health compared to the “low and decreasing” group (OR=1.61; 95% CI: 1.12-2.31). This dissertation provides insight into discrimination measures and their relationships with HRQOL among Asian Americans. Proper measurement and health impact assessment of discrimination in this population can contribute to better monitoring and provide improved accountability and support for interventions and public policies.
  • Item
    (2019) Bueno de Mesquita, Paul Jacob; Milton, Donald K.; Epidemiology and Biostatistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)
    Pandemic preparedness is weakened by uncertainty about the relative importance of influenza transmission modes, particularly airborne droplet nuclei (aerosols). A human-challenge transmission trial in a controlled environment was conducted to address this uncertainty. Healthy, seronegative volunteer ‘Donors’ (N=52) were randomly selected for intranasal challenge with influenza A/Wisconsin/67/2005 (H3N2) and exposed to seronegative ‘Recipients’ randomized to intervention (N=40) or control (N=35) groups. Intervention recipients wore face shields and hand sanitized frequently to limit large droplet and contact transmission. A transmitted infection, confirmed by serology in a control recipient, yielded a 1.3% SAR overall. This was significantly less than the expected 16% SAR (p <0.001) based on a proof-of-concept study that used half as many Donors and exposure days. The main difference between these studies was mechanical building ventilation in the follow-on study, suggesting a possible role for aerosols. The extent to which Donor viral shedding was similar to that of mild, natural infections and may be useful for studying transmission was investigated. The only available aerosol shedding comparison data comes from a population of adults with influenza A H3 infection enrolled on the basis of febrile illness plus cough or sore throat, or positive Quidel QuickVue rapid test (N=83). Systematic differences in case selection compared with Donors yielded more severe cases and introduced bias. To account for differences in illness severity, propensity score matching, stratification, and inverse weighting ultimately demonstrated that the experimental and naturally infected groups were too different to compare without bias. While acknowledging the uncertainty in the generalizability of the current challenge model, observed aerosol shedding and CO2 were used in the rebreathed-air version of the Wells-Riley equation to compute average quantum generation rates (95% CI) 0.029 (0.027, 0.03) and 0.11 (0.088, 0.12) per hour for infected Donors and fine aerosol shedding Donors, respectively. Donors shed 1.4E+5 (1.0E+5, 1.8E+5) airborne viral RNA copies per quantum (ID63). This dissertation provides evidence for airborne transmission, presents a methodology for estimating an airborne dose, and suggests a role for building ventilation in reducing risk and the need for future observational studies to evaluate transmission modes in non-experimental settings with greater generalizability.
  • Item
    (2019) Al-Nassir, Marwa Fawzi; Dallal, Cher M; Epidemiology and Biostatistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)
    Trust is the foundation of the patient-physician relationship. Patients’ trust in a physician has been associated with patient satisfaction, adherence to treatment, continuity of care, and improved health outcomes. Trust in a physician is especially important for health-vulnerable populations, such as cancer survivors, as they tend to endure complex emotional needs related to re-acclimating to the new normal post cancer diagnosis. The patient-physician relationship also relies heavily on effective patient-centered communication (PCC), however, associations between PCC and trust in physician information overall and by cancer survivorship status is not well understood. Using nationally representative data (N = 2604) ascertained from Cycle 1 of the fifth iteration of the 2017 Health Information National Trends Survey (HINTS), a cross-sectional analysis was conducted to examine PCC in relation to trust in physician information. PCC was assessed on a 4-point Likert scale using responses from seven sub-questions that address the main functions of PCC: 1) fostering healing relationships, 2) exchanging clinical information, 3) responding to emotional needs, 4) managing uncertainty, 5) facilitating shared decision-making, and 6) enabling patient self-management. Trust in physician information was analyzed dichotomously (high versus low) based on responses from a single item question. PCC was analyzed as individual components (optimal versus sub-optimal) and as an overall score. Confounders included age, sex, race/ethnicity, education, and household annual income. Odds ratios (OR) and 95% confidence intervals (CI) for the relationship between PCC and trust in physician information were estimated using multivariable logistic regression. Analyses of cancer survivorship status (cancer survivor versus never had cancer) as an effect modifier of the relationship between PCC and trust in physician information was also conducted using an interaction term. Results from the weighted multivariable models revealed that for every one-unit increase in the overall PCC score (range 1 to 100), the odds of having high trust in physician information increased by 4% (adj OR = 1.04, 95% CI = 1.03–1.05). The odds of reporting high level of trust in physician information were significantly associated with each individual component of PCC when comparing those who felt their communication component was optimal versus sub-optimal (PCC components: exchanging clinical information (adj OR = 2.57, 95% CI = 1.82–3.62), responding to emotional needs (adj OR = 2.34, 95% CI = 1.65–3.30), facilitating in shared decision-making (adj OR = 2.35, 95% CI = 1.70–3.26), enabling patient self-management (adj OR = 2.88, 95% CI = 2.11–3.92), managing uncertainty (adj OR = 2.45, 95% CI = 1.74–3.44), fostering healing relationships (adj OR = 2.79, 95% CI = 2.18–3.57), and spending enough time with you (adj OR = 2.09, 95% CI = 1.49–2.93)). When examining relationships by cancer survivorship status, estimates among cancer survivors were of greater magnitude compared to persons who reported never having cancer, however, no significant interactions were observed in the weighted multivariable models (all p-interaction>0.05). These findings provide insight on how optimal experiences of PCC influence trust in physician information and can help inform the development of PCC strategies to ultimately improve health outcomes and reduce consequences related to poor patient-physician trust overall and among cancer survivors.
  • Item
    (2019) Jenkins, Emily J; Slopen, Natalie B; Epidemiology and Biostatistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)
    Tooth resorption (TR) in felines may present an increased risk for complications during dental procedures, and it is currently unknown whether oral examination is a valid diagnostic method for type 1 TR. Using existing data from 1,530 felines from a large veterinary hospital in Washington, D.C., I examined the association between type 1 TR and complication under anesthesia during a dental procedure, and the validity of oral examination as a diagnostic tool. Controlling for breed, weight, age, sex, hematocrit, total protein, technician, veterinarian, and presence of oral and systemic disease, type 1 TR was associated with a complication under anesthesia during a dental procedure (p<0.0001). Sensitivity (93.1%) and specificity (97.6%) support the reliability of oral examination for diagnosing type 1 TR. Veterinarians can proactively anticipate lower blood pressures when anesthetizing felines with type 1 TR. There are similarities between TR in humans and felines, and further research is needed on the pathophysiology and health implications of TR in both species.
  • Item
    (2019) Barrall, Angelica Lynne; Dyer, Typhanye V; Epidemiology and Biostatistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)
    Individuals with submicroscopic malaria infection are an important reservoir for transmission, but the clinical consequences of these low-density parasitemia infections are poorly understood. Using cross-sectional data from six household-based surveys conducted during the dry and rainy seasons in Malawi from 2012 to 2014, this study examined the association between submicroscopic infection and fever in children and adults. For each survey, 900 households were recruited from three distinct ecological settings in southern Malawi to participate in the study (N=22,145). Overall prevalence of submicroscopic infection in the analytic sample was 8.1%. In a generalized linear mixed model accounting for clustering at the household and neighborhood levels and controlling for age and survey number, submicroscopic infection predicted fever in the dry season only (OR=1.66; 95% CI: 1.04, 2.66). Therefore, fever might not be a consistent marker of submicroscopic infection, but identification and treatment of low parasitemia infections is necessary to eliminate malaria transmission.
  • Item
    Using the Index of Concentration at the Extremes to Examine the Impact of Air Pollution Exposure on Infant Mortality in the United States
    (2019) Ammons, Samantha; Nguyen, Quynh; Epidemiology and Biostatistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)
    Background: The concentration of privilege in a geographic area can determine how vital resources are distributed among certain groups in that area, thus influencing a community’s health. High air pollutant exposure is often concentrated in deprived neighborhoods with lack of vital resources. Objective: Determine whether states with a high concentration of air pollution exposure have higher infant mortality rates (IMR) than states with lower concentrations of air pollution exposure. Methods: The Index of Concentration of the Extreme was utilized to measure the concentration of air pollution exposure for each state. Incidence Rate Ratios and 95% Confidence Intervals for state infant mortality rate were computed using Poisson regression in Statistical Analysis Software. Results: States with high concentrations of air pollution exposure had 19% lower IMR than states with low air pollution exposure (95%CI:0.70 – 0.94). Conclusions: These findings can enable researchers to conduct census-tract research on adverse health outcomes and societal distributions.
  • Item
    Relationships of social and physical environmental factors with cardiometabolic outcomes
    (2019) Huang, Dina; Puett, Robin; Nguyen, Quynh C; Epidemiology and Biostatistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)
    The social and physical environmental factors impact health in general and have been linked with increased risks of cardiometabolic outcomes including obesity, diabetes, hypertension and cardiometabolic biomarkers. The dissertation added to important knowledge on this topic in two ways: 1) by leveraging innovative Twitter-derived characteristics to study the potential influence of social environment on cardiometabolic outcomes, 2) investigating the effects of air pollution exposures on cardiometabolic outcomes in youth living with type I diabetes. The first study investigated the associations between Twitter-derived area-level predictors (happiness, diet, physical activity) with cardiometabolic outcomes (obesity, diabetes, hypertension) using a nationally representative sample from National Health and Nutrition Examination Survey (NHANES). People living in neighborhoods with higher happiness, healthier diet and more physical activity had lower prevalence of obesity and hypertension but not diabetes. Twitter-derived social neighborhood characteristics can be used to identify communities with higher risk of cardiometabolic outcomes. We obtained data from SEARCH for Diabetes in Youth (SEARCH) study for the second and the third study. The second study examined the associations between chronic exposure to air pollution and glucose hemostasis (HbA1c) in youth living with type I diabetes. Particulate matter with aerodynamic diameter <2.5 (PM2.5), proximity to heavily trafficked roads and annual average daily traffic count were associated with higher HbA1c in study site South Carolina, Colorado and Washington, but not in study site Ohio and California. Differences in particulate matter compositions may explain the inconsistent results. The third study assessed the effect of acute exposure to air pollution on subclinical CVD markers including pulse wave velocity (PWV), augmentation index (AIx) and brachial distensibility (BrachD) using a repeated measures design. Reduction in PM2.5 on the day prior to assessment was associated with lower AIx, but not associated with either PWV or BrachD. In summary, exposure to air pollution may be associated with cardiometabolic outcomes and reducing air pollution may have implications in early prevention of cardiovascular complications for youth living with type I diabetes. Overall, reducing social stressors and reducing hazardous physical environmental factors may decrease the risk of cardiometabolic outcomes, providing possible directions for CVD prevention for public health practitioners.