Public Health Trans-Disciplinary Collaboration Pilot Awardees
The Public Health Trans-Disciplinary Collaboration Pilot Awards support collaborations between Graduate School of Public Health (GSPH) faculty and School of Medicine (SOM) faculty. The goal of this program is to establish and strengthen collaboration between GSPH and SOM faculty members with priority areas being a) climate change and health and b) precision public health. Funded projects can target vulnerable communities locally, regionally, and globally, and are expected to involve trans-disciplinary collaborations between at least one GPSH investigator and one SOM investigator.
A Translational Mouse Model of Climate Change and Chronic Kidney Disease of Unknown Origin (CKDu)
$45,000 Awardee
Team: Alison Sanders, MD (PI), Jacqueline Ho, MD (CO-PI)
Abstract: Chronic Kidney Disease of unknown origin (CKDu) is a public health emergency in Central America, regions of Sri Lanka, India and the U.S. Over the past four decades, disparate agricultural communities have observed dramatic increases in the incidence and deaths due to CKDu. CKDu is not explained by traditional risk factors for kidney disease like age, hypertension or diabetes. The communities most at risk for CKDu live and work in environments with substantial environmental toxicant co-exposures and hot conditions. Recent studies suggest that children living in CKDu endemic areas have early life renal damage, and we propose that CKDu has developmental origins via exposure to common nephrotoxicants exacerbated by heat stress. Identifying developmental factors that set individuals on a path for CKDu as adults is a major shift in understanding the origins of this disease. Windows of susceptibility can occur during times of rapid development that are vulnerable to adverse exposures. In this pilot study, we posit a fundamental question: does maternal heat stress following early gestation toxicant exposures prime the kidney for subsequent damage and/or exacerbate damage? Resolving this hypothesis will provide both mechanistic insight to kidney disease pathogenesis and foundational evidence for understanding CKDu risk. Our study aims to disentangle complex environmental risk factors, including climate change-related heat stress, experienced in communities with endemic CKDu. This work will characterize how nephrotoxic metals and heat stress synergistically disrupt nephron number and kidney function in a mouse model, using human-relevant exposures. Together, these efforts will produce an approach integrating mouse models with population relevant human exposure data, providing knowledge vital to the etiology of CKDu.
Genomic Epidemiology of Human Metapneumovirus
$45,000 Awardee
Team: Anna Wang-Erickson, PhD (PI), Mark Roberts, MD (CO-PI), John Williams, MD, Lee Harrison, MD, Vaughn Cooper, PhD
Abstract: Human metapneumovirus (HMPV) is a major cause of acute respiratory disease worldwide and the second most common cause of lower respiratory infection in US children. There is no licensed vaccine and no published large-scale population-based genomic epidemiologic study of HMPV using whole genome sequencing (WGS). Genomic analysis of HMPV would significantly contribute to the development of an effective vaccine. As illustrated by the COVID-19 pandemic, genomic data is crucial for identifying subtle mutations in viral RNA that give rise to variants that may have increased transmissibility and, importantly, immune escape. HMPV is also an RNA virus that mutates, but the public health implications of HMPV variants have not been extensively studied. We propose to use these data to generate a robust modeling approach to predict transmission and disease burden of HMPV. We propose a pilot study that will be the first large-scale population-based HMPV genomic epidemiologic study. We hypothesize that WGS analysis offers better resolution to understand transmission patterns. Using a large and robust collection of HMPV-positive clinical samples collected with corresponding electronic medical record data through the CDC-funded New Vaccine Surveillance Network, this study aims to describe the genomic diversity of HMPV in Pittsburgh from 2016-2020 in relation to patient demographic data. We will also identify whether a variant that became dominant and replaced the “classic” strain in Yokohama City by 2018 has circulated in Pittsburgh and possibly when it was introduced. Finally, we will develop a mathematical model of HPMV transmission that will incorporate demographic and epidemiologic data.
Precision Care in Childhood Asthma using EHR Analytics
$45,000 Awardee
Team: Ying Ding, PhD (PI), Erick Forno, MD (CO-PI)
Abstract: Asthma affects over 6 million children in the US. The main drivers of pediatric asthma morbidity are emergency department (ED) visits (760,000/year) and hospitalizations (74,000/year). Early identification of sub-populations at high risk for severe exacerbations, and understanding the factors that contribute to worse outcomes, will be crucial to improve asthma surveillance and management. The main objective of this proposal is to apply machine learning methods to analyze pediatric asthma EHR data to identify subgroups for early intervention, improve clinical outcomes, and help reduce asthma disparities. The comprehensive asthma dashboard at Children’s Hospital of Pittsburgh (CHP) includes EHR data from 12,000 unique patients 7,200 ED visits, and 3,600 hospitalizations. It constitutes an exceptional, real-world, real-time resource to examine and address challenges and unmet needs in managing childhood asthma. Specifically, we will: (1) Extract population- and individual-level CHP EHR records to build analyzable datasets with sustainable architecture. (2a) Identify subgroups with worse asthma care outcomes (e.g., ED visits and unscheduled returns, hospitalizations and readmissions) using tree-based and deep-neural-network-based machine learning methods; and (2b) characterize the key factors (including sociodemographic and environmental characteristics) that contribute to the observed heterogeneity. (3) As an immediate quasi-experimental application, we will assess the direct impact of COVID19 on acute care for childhood asthma using a counterfactual causal inference framework and characterize vulnerable asthma populations who have suffered more impacts from the pandemic. We will use the findings to develop a precision surveillance strategy to help mitigate adverse impacts on asthma care from future disruptive events.
Precision Public Health Methods for Alzheimer’s Disease
$45,000 Awardee
Team Members: Jiebiao Wang, PhD (PI), Victor Talisa, PhD (CO-PI)
Abstract: Alzheimer’s disease (AD) is the most common type of dementia in the aging population. However, AD is known to be diverse and heterogeneous, and its etiology is largely unknown. Although AD treatments targeting specific pathological features have been developed, their efficacy in the general population is debatable. Thus, access to non-invasive procedures for quantification of these features individually may be useful for screening and optimal treatment identification. Here we propose to develop a non-invasive machine learning (ML) model for AD brain pathology to inform precision public health (PPH) practices by improving the pairing of patients with effective targeted therapies. Existing studies have identified dozens of risk/protective factors from molecular, behavioral, social, psychological, environmental, and medical variables. Unlike most studies focusing on one data modality, we will study the interactions among multiple modalities to simultaneously predict levels of three AD pathological markers which are currently only measured post-mortem, but which characterize the gold-standard criteria for AD diagnosis: neuritic plaques, diffuse plaques, and neurofibrillary tangles. To achieve this goal, we will harmonize rich data from existing cohort studies with the Alzheimer’s Disease Neuroimaging Initiative, the Religious Orders Study and Memory and Aging Project, and multigenerational Framingham Heart Study. We will utilize cross-sectional and longitudinal ante- and post-mortem-biospecimen data and multiple layers of "omics" data (e.g., clinical, psychological, experiential, genomic, and blood transcriptomic). The large-scale multidimensional data with our novel multivariate ML approach will serve as the basis for future systems biology and gene-environment studies and the development of a new taxonomy for AD prevention.
Using a Precision Public Health Approach to Evaluate the Impacts of Structural Determinants and Neighborhood Conditions of Sepsis Risk
$45,000 Awardee
Team Members: Christina Mair, PhD (PI), Kristina Rudd, MD (CO-PI), Christopher Seymour, MD
Abstract: While sepsis is a tremendous challenge in nearly every location in the world, there is strong evidence that certain subgroups are at higher risk than others. Improved insight into relationships between sepsis and community-level risk factors will be critical to understanding future implications of changing community conditions on sepsis risk and inequities. The features of structural determinants and neighborhood conditions that can impact sepsis inequities encompass many constructs. Computational (machine learning) methods, such as causal modeling over high-dimensional data, can identify confounding variables, distinguish between direct (potentially “causal”) relations and simple correlates, and identify key combinations of environmental measures that may directly affect sepsis inequities. These methods provide innovative solutions to understanding which combinations of environmental features impact sepsis incidence and can help elucidate potentially mechanistic relationships between environmental conditions and inequities. The overall objective in this proposal is to estimate potential causal relationships and key combinations of structural determinants and neighborhood risks and supports that may affect sepsis risk, using a precision public health approach. We will accomplish this objective with two innovative specific aims: 1) evaluate spatial, social, and health systems based patterns of incident sepsis hospitalizations among adults in Allegheny County, Pennsylvania over a 4- year period using multilevel models that geographically link electronic health records, census data, and community data sources at the ZIP code level, and 2) apply machine learning approaches to the combined dataset to infer potential cause-effect associations between sepsis risk and key elements of the social, policy, and physical environments and novel composite predictors.
