Have a UROP opening you would like to submit?
Please fill out the form.
Evaluating COVID-19-Induced Excess Mortality at the County-Level in the United States
HST: Health Sciences and Technology
Roger Greenwood Mark
Marie-Laure Charpignon: email@example.com
Keywords: biostatistics, causal inference, natural language processing COVID-19 has led to increased morbidity and mortality among older populations and those with underlying health conditions, but also among those who are unable to access outpatient and emergency services because of the lockdown and surge in acute care demand. As an example, in Colorado, the decrease in fatal road and work accidents associated with stay-at-home orders did not make up for the COVID-19-related deaths, and excess mortality from February through July was estimated at 60 per 100k (25 attributed to COVID-19), an 18% increase compared to 2015-2019. Explaining the underlying causes of death, and assessing the geographical heterogeneity of the excess death toll within the state, require careful retrospective analysis of death certificates, including job occupation, industry, and education level information. To that end, we have contacted departments of health across all 50 states and the District of Columbia. We have collected county-level all-cause mortality data for 15 states, and are about to conclude additional partnerships with local departments. Additionally, we compiled news articles from the Media Cloud state-level media aggregator that had specific county mentions since March 1, 2020. Deriving additional insights from infoveillance tools may help provide the narrative behind excess death calculation from COVID-19 at the county level in the US. The challenge is to develop a causal-mediation analysis framework to assess: (a) the direct effect of underlying demographics, population health, and county infrastructure onto COVID-19-induced excess mortality; (b) the indirect contribution of mobility on COVID-19 mortality and non-COVID-19 excess deaths, by integrating inter-county mobility data from PlaceIQ, Unacast, SafeGraph, and Facebook.
We are looking for students interested in interdisciplinary research, with a background in statistics, natural language processing, and/or network science. Knowledge of Python is required; experience in epidemiology is not required but a plus.