22689 research outputs found
Sort by
STRUCTURAL RACISM, HEALTHCARE INFRASTRUCTURE, AND RACIAL AND GEOGRAPHIC DISPARITIES IN DECEASED ORGAN DONATION IN THE UNITED STATES
No full textOrgan donation remains a lifesaving intervention in the United States, yet disparities in donation rates and donor eligibility persist. Because donation typically requires a hospital-based death, place of death represents a critical determinant of donor potential. This dissertation uses national mortality, healthcare system, and census-based data to examine how structural racism and healthcare infrastructure shape racial and geographic disparities in deaths consistent with organ donation.
First, we used national mortality data to examine racial and ethnic differences in the likelihood of dying in outpatient settings among individuals whose deaths were consistent with organ donation (Chapter 2). We found that Black, American Indian, Asian, and Hispanic individuals had a lower risk of outpatient death compared to non-Hispanic White individuals.
Second, we evaluated whether healthcare infrastructure—including primary care provider (PCP) density, hospital density, and rurality—was associated with the likelihood of outpatient deaths consistent with organ donation (Chapter 3). Using county-level data linked to national mortality records, we found that lower PCP density and rural status were associated with a modestly higher risk of outpatient death. Hospital density showed no consistent association.
Third, we explored how racial and economic segregation relate to outpatient deaths consistent with organ donation using the Index of Concentration at the Extremes (ICE) (Chapter 4). Counties with greater racialized economic segregation had a lower risk of outpatient deaths, as did counties with higher concentrations of Black residents or low-income households.
Finally, we conducted a nationally representative survey to assess public attitudes toward organ donation and identify factors associated with registration (Chapter 5). Trust in the healthcare system, familiarity with the donation process, and discussions with family or providers were all significantly associated with donor registration status. Black respondents had lower rates of registration than White respondents, but higher levels of willingness to register in the future.
This dissertation broadens our understanding of how place of death and structural inequities intersect to shape organ donation outcomes. By incorporating geographic, systemic, and attitudinal perspectives, these findings provide a foundation for more equitable approaches to donor identification and public health strategies
AI-Driven Policy Analysis: NLP And Multi-Level Modeling for Complex Systems
No full textPolicymakers today face an increasingly complex landscape marked by limited resources, fragmented information, and interdependent socio-environmental challenges. This dissertation develops a series of complementary methodologies that leverage natural language processing (NLP) and multi-level modeling to enhance different phases of the policy analysis cycle, from policy problem definition to intervention simulation.
The first part of the dissertation focuses on NLP for policy analysis. Using case studies such as the TRIPS waiver debate and the digital health policy discourse in India, it demonstrates how real-time text mining, topic modeling, and co-occurrence analysis can uncover shifts in stakeholder positions, identify emergent policy windows, and surface hidden dynamics within public narratives. Results show that NLP pipelines enable timely, structured insights into political framing but require careful attention to preprocessing choices and validation frameworks to avoid bias and misrepresentation.
The second part of the dissertation develops methods to simulate policy interventions in complex systems. To assess community resilience during health crises, we develop a pandemic-focused COPEWELL model using system dynamics to generate a community resilience index. For suicide prevention among Indigenous American youth, we construct a multi-level model that combines system dynamics with microsimulation to evaluate how variations in leadership commitment, funding levels, and program management affect the sustainability of interventions and the reduction of youth suicide risk. To explore adolescent dating relationships and substance use, we employ an attention-based neural network augmented with SHAP value analysis, identifying which behavioral and contextual factors most strongly predict relationship disruptions. To quantify the economic effects of climate anomalies, we build a multi-level ENSO-climate model that traces how El Niño and La Niña events propagate through agricultural production and influence broader macroeconomic indicators, revealing key feedback loops between environmental shocks and economic performance. These models provide insight into the co-evolution of human behavior, environmental shocks, and systemic feedback, while maintaining transparency and stakeholder relevance.
This thesis presents a broad applicability of NLP and multi-level modeling for policy analysis. It demonstrates how real-time text based policy insights and ,multi-level system simulations can support adaptive and transparent policymaking in complex socio environmental settings
Investigating Receptor Tyrosine Kinase Signal Transduction Utilizing An Interdisciplinary Biophysical Approach
No full textReceptor tyrosine kinases (RTKs) are single-pass membrane receptors that mediate diverse physiological and pathological signaling events. This dissertation investigates the molecular mechanisms of RTK activation and signaling specificity using an integrated approach that combines protein purification, quantitative fluorescence imaging and dose-dependent cellular signaling assays. Full-length and extracellular domain constructs of the EphA4 receptor were expressed and purified for structural studies, establishing a scalable protocol suitable for cryo-electron microscopy. In parallel, Number and Brightness (N&B) analysis paired with surface receptor concentration calibration was used to evaluate EphA4 and its cancer-associated mutant’s oligomerization in living cells, revealing mutation-dependent differences in ligand-induced clustering. To explore signaling output, the FGFR3 receptor and its disease-associated mutants were assessed for pathway preference using quantitative immunoblotting of the MAPK/ERK and PI3K/AKT pathways. These analyses demonstrated that pathogenic mutations bias FGFR3 signaling toward specific downstream cascades, supporting the concept of mutation-induced signaling bias. Collectively, this work improves the methodology for studying membrane receptor behavior, advances our understanding of RTK signaling specificity and informs future therapeutic strategies targeting biased RTK activity
Preventing Child Sexual Abuse in Youth-serving Organizations
Get PDFChild sexual abuse (CSA) is a complex public health issue that has devastating public health impacts for victims and for institutions where abuse cases occur. This type of abuse is a unique form of childhood trauma that can result in poorer health outcomes over one’s lifetime, including risk of substance abuse, mental health issues, and less economic success. CSA has not always been treated as preventable phenomenon, but research over the last several decades shows that implementing prevention strategies within youth-serving organizations (YSOs) may help to reduce the risk of victimization of children in those settings.
This dissertation explores the issue of CSA prevention in YSOs through three papers, with an introductory chapter. The first paper includes background information on best practices in CSA prevention, and an integration of concepts from implementation and management sciences to better frame the issue within institutions. This paper was informed by key informant interviews with leaders who were involved in early implementation of CSA prevention within one national YSO setting, with a focus on how they got started with their prevention approach and shared it with others. The second paper explores CSA prevention in YSOs on the local level in two states: Massachusetts and Rhode Island. Key informant interviews were conducted with local YSO leaders to gain their perspectives on CSA prevention within their organizations, with a focus on if/how they use existing best practices and what barriers they face. The third and final paper includes recommendations for YSO leaders approaching the issue of CSA prevention.
The results/conclusions of these studies include the influence of public CSA incidents in motivating leaders to make changes in their approach to CSA prevention, the importance of leadership commitment to prevention, and challenges including lack of time, expertise, and resources. Leaders did indicate an awareness of risk of this issue, but interviews indicated that it may be challenging to keep the issue top of mind, especially for those YSOs not connected to a national YSO organization
COUNTERINSURGENT URBANISM: WEAPONIZING LAND AND HERITAGE IN THE KURDISH REGION OF TURKEY
No full textEmerging technologies of war and counterinsurgency such as drone-delivered ordinance increasingly target urban landscapes, populations, and infrastructures. To achieve military control, states also expand their repertoires of counterinsurgency to include bureaucratic techniques, such as leveraging courts, via the reformulation of property rights. Focusing on the state’s frontier, where sovereignty is still under question, this dissertation explores the role of the built environment in counterinsurgency campaigns. This project asks: If civil wars are territorially based armed challenges between states and insurgents, why do states expend the time and resources to simultaneously engage in developmental projects targeting urban spaces in the midst of counterinsurgency campaigns? What is the spectrum of tactics so used? How do inhabitants whose self-identity is shaped, in part, through attachments to urban landscapes conceive, adopt, or repurpose the land-heritage-military nexus in the everyday? Focusing on the civil war between the Kurdistan Workers’ Party and the Turkish state, this dissertation argues that states intervene in the everyday environment of ostensibly unruly populations as part of a broader security strategy of establishing spatial control over contested territory. Spatial control refers to the construction and projection of military and political authority as it materializes through the built environment. I term this set of processes “counterinsurgent urbanism.” Counterinsurgent urbanism is a layered group of policies and interventionist actions aimed at annihilating existing insurgencies and foreclosing future ones through the remaking and subsequent control of the built environment. I identify three core mechanisms of counterinsurgent urbanism: legal-institutional restructuring of land; policing of the urban landscape; and selective development. I draw on a multi-method approach that includes ethnography, in-depth interviews, neighborhood mapping, photography, tourism surveys, and archival research to demonstrate how everyday, non-spectacular forms of violence operate in global counterinsurgencies. Shifting beyond the delimited spaces of battlefields as the dominant site of analysis in war and counterinsurgency, it presents destruction, (re)construction, and decay in the built environment as distinct yet relational modalities of counterinsurgent urbanism. This dissertation contributes to key literature on political violence, counterinsurgency, and urban politics
“BRICKS TO STAND ON”: BEYOND BELONGING, LAYING THE FOUNDATION FOR BLACK TEACHER CANDIDATE SUCCESS
Get PDFThis mixed methods dissertation investigated the persistent racialized performance gap experienced by aspiring teachers in an online teacher preparation program. Utilizing Bronfenbrenner’s Ecological Systems Theory (EST) and framed within the transformative paradigm, the connection between culturally responsive pedagogy (CRP) and academic engagement between Black and White teacher residents (TRs) was interrogated and analyzed. The study answered three questions: (1) How does academic engagement differ between Black and White TRs? (2) What CRP practices do faculty report using, and how do Black TRs perceive these practices? (3) What is the relationship between CRP practices and the academic engagement of Black TRs?
Quantitative data was collected from both TRs and faculty using Likert-scale survey instruments and analyzed using descriptive statistics, t-tests, and a series of regression models. Although no statistically significant differences were revealed in engagement scores by race in this study, Black TRs consistently reported higher engagement across behavioral, emotional, and cognitive domains. These findings, alongside persistent academic disparities, surface points of fragility within EDU’s capacity to achieve equitable outcomes for all students. Notably, both faculty and Black TRs reported frequent use of CRP, though their definitions and perceived enactment of CRP revealed important divergences. Regression analysis revealed that diverse teaching practices significantly predicted behavioral and overall engagement among Black TRs.
Qualitative data, gathered through semi-structured interviews with faculty and Black TRs, were thematically coded and integrated using joint displays. Qualitative findings identified four core themes of CRP at EDU: a) relational engagement, b) instructional flexibility, c) critical professional skill-building, and d) reflective praxis. Black TRs communicated a strong perception of CRP, especially in relation to equity commitments and a building community. However, TRs and faculty differed in their views of CRP enactment. Faculty pointed to cultural acknowledgment, while Black TRs described these gestures as superficial, noting their lived experiences and funds of knowledge were seldom affirmed in meaningful ways.
Conducted in a period of accelerating backlash against identity-affirming practices, this research underscores the importance of examining how such practices influence engagement and equity, particularly when policy shifts risk erasing them without understanding their impact
Sharecropping and Surplus Life in US Southern Literature
No full text“Sharecropping and Surplus Life in US Southern Literature” shows how writers in the US South in the first half of the twentieth century represented the history of sharecropping, the predominant form of labor exploitation in Southern agriculture in the late nineteenth and early twentieth centuries. As the onset of the Great Depression and increased mechanization in agriculture led to the disintegration of the sharecropping system, a range of writers working across genres turned their focus toward the deepening poverty of the South’s rural population. While literary works on sharecropping are commonly interpreted as nostalgic records of a vanishing peasantry, I argue that these works are better understood as a contested set of literary responses to an emergent form of rural unemployment. This literary response to crisis was central to the institutional history of Southern literature, a subfield that emerged from a contestation of regional cultural identity. By situating sharecropping within the dynamics of capitalist accumulation and the history of the wage relation, I advance a new reading of the political and representational problems of race, labor, and land in Southern literature
COMPUTABLE PHENOTYPING IN PRECISION ONCOLOGY
Get PDFPrecision oncology aims to personalize cancer treatments by integrating diverse data sources, including genomic and clinical information, but the sheer volume and complexity of unstructured text present significant barriers to efficient knowledge discovery and its application in clinical decision-making, such as in Molecular Tumor Boards (MTBs). This research focuses on developing and rigorously evaluating Natural Language Processing (NLP) methodologies to automate the retrieval of essential information, thereby advancing computable phenotyping to support enhanced, data-driven decision-making in cancer care.
My research is structured across three primary investigations. First, I conducted a comprehensive scoping review to map current trends, methodologies, and opportunities in computable clinical phenotyping, specifically identifying gaps in precision oncology applications and highlighting the need for approaches that integrate diverse data sources and undergo real-world evaluation. Second, I developed and evaluated AI-driven systems for knowledge extraction from biomedical literature, focusing on Named Entity Recognition (NER) and Relation Extraction (RE) of biomarkers in cancer study such as genes, variants, diseases, and chemicals. This work demonstrated that fine-tuned BioBERT models can achieve strong performance and may offer a significant step towards AI-assisted knowledge discovery for MTBs. Third, I addressed the challenge of extracting critical clinical endpoints from pathology reports by comparing customized rule-based systems, fine-tuned BioBERT models, and instruct-tuned LLMs for the automated retrieval of numerical tumor sizes. This investigation found that BERT models achieved superior accuracy for this task, which can help for cancer staging and treatment response assessment.
Overall, this dissertation makes significant contributions to the field of precision oncology by providing novel, validated NLP methods for deep phenotyping and automated information extraction from complex biomedical texts. The developed methodologies demonstrate the potential to reduce manual effort, improve the accuracy and consistency of extracted data, and streamline the synthesis of information critical for clinical decision-making. Continued research and collaboration will be crucial to fully realize the transformative potential of these technologies in advancing personalized and effective cancer care
Identifying and addressing the implications of bursty network traffic
Get PDFMeasurements of network traffic unveil periods of high link utilization that can exhaust switch buffers and lead to packet loss. Unfortunately, the diversity and ever-changing nature of networks, workloads and traffic patterns, and even host networking deployments pose a challenge in attributing the sources of bursty traffic. Additionally, bursts are usually short-lived. Hence, existing proactive countermeasures struggle to detect and address them in time. In this study, we first revisit the definition of burstiness in the context of network traffic and introduce a high-resolution network traffic measurement framework to accurately identify and analyze the egress traffic from host machines.
Our findings hint that existing traffic shaping solutions, e.g., packet schedulers, are either ineffective in preventing bursts or do not take into account the bursty nature of the traffic. We demonstrate that Deficit Round Robin (DRR), the de facto fair packet scheduler in the Internet, can perform poorly because of its design assumptions. Concretely, our study unveils that DRR performs best if (1) packet size distributions are known in advance, and (2) all bursts are long and create backlogged queues.
We show that neither of these assumptions holds in today's Internet: packet size distributions are varied and dynamic, and Internet traffic consists of many short, latency-sensitive flows, creating small bursts. These flows can experience high latency under DRR as it serves a potentially large number of flows in a round-robin fashion.
To address these shortcomings, we introduce Self-Clocked Round-Robin Scheduling (SCRR), a parameter-less, low-latency, and scalable packet scheduler that boosts short latency-sensitive flows through careful adjustments to their virtual times without violating their fair share guarantees.
Finally, we investigate what happens when short-term bursts hit the switch buffers at the core of the network and packet loss becomes imminent. We introduce selective packet deflection, a reactive measure to prevent microbursts from causing expensive packet drops in the event of extreme buffer congestion. Packet deflection aims to use the available buffer in neighboring switches in managed environments to temporarily host packets that arrive at full buffers, significantly reducing the packet loss and improving application performance
CHARACTERIZATION OF RWPE-1 PROSTATE ORGANOIDS: EFFECTS OF CULTURE MEDIUM COMPOSITION, SEEDING DENSITY, AND STROMAL CELL CO-CULTURE
No full textAs the most commonly diagnosed cancer in men in the United States, prostate cancer imposes a substantial public health burden. Despite the generally favorable prognosis following primary treatment, a subset of patients initially present with or eventually progress to the metastatic form of prostate cancer, which is associated with poor prognosis and remains therapeutically challenging. Although substantial progress has been made in prostate cancer research to date, several key aspects of the cancer’s life cycle, including tumorigenesis, progression, and therapeutic responses, remain to be fully elucidated.
In order to study the earlier stages of tumorigenesis, an in vitro model of the healthy prostate epithelium is needed. Such a model has traditionally been challenging to construct, as normal prostate luminal cells are non-proliferative and thus exhibit poor growth in flask-based cell culture systems. Organoid cultures, in which cells with stem cell properties are seeded on bioactive hydrogels under appropriate conditions to induce differentiation into the desirable lineages and the formation of physiologically relevant structures, serve as a potential solution to this problem. In this work, we provide a comprehensive characterization of prostate organoids derived from the immortalized prostate epithelial cell line RWPE-1. We characterize RWPE-1 organoids grown under different culture conditions, such as the inclusion of serum and androgen in the medium, the initial seeding density, and co-culture with stromal cells. This study evaluates the potential of RWPE-1 prostate organoids to serve as an in vitro normal prostate epithelium model for studies aimed at determining factors that contribute to early prostate cancer development. This study also evaluates the physiological relevance of the RWPE-1 organoid model, describing key behaviors of the organoids, including acinar morphogenesis, branching morphogenesis, and extracellular matrix deposition under various culture conditions