When COVID-19 was threatening Connecticut in early 2020, state policymakers turned to leading scientists for guidance. Among those responding to the call was Yale School of Public Health Associate Professor of Biostatistics Forrest Crawford.
Crawford and a multidisciplinary team of scientists created a computer model capable of estimating potential COVID-19 infections, hospitalizations, and deaths in every county in Connecticut as COVID-19 transmission progressed in the state over time. The information proved vital in helping Connecticut Governor Ned Lamont and the Connecticut Department of Public Health make informed decisions to protect the state’s 3.6 million residents.
“Forecasting these outcomes involved building a mathematical model of disease transmission that we used to learn about the dynamics of transmission across the state,” said Crawford. “We also collected a new kind of data to monitor social distancing behavior using mobile device data.”
Called a contact metric, the new data leveraged people’s mobile device location information to measure interpersonal contact. In order to protect people’s privacy, all personal identifying information was removed from the passively collected data before it was provided to the researchers. Previously, scientists relied on other mobility metrics such as the distance people traveled or the amount of time they spent away from home to estimate interpersonal contact rates. The new contact metric substantially improved the model’s accuracy in predicting infections in Connecticut, Crawford said.
The research team believes the contact metric created for their COVID-19 study could be applied in other research investigating the transmission dynamics of a disease.
“We have focused in this study on the U.S. state of Connecticut, but the usefulness of anonymized and passively collected contact data could be generalized to other settings,” the researchers wrote in their study.
Crawford is part of the Public Health Modeling Unit at the Yale School of Public Health, an interdisciplinary group of faculty engaged in mathematical and statistical research in biostatistics, epidemiology, and health policy.
Mathematical modeling involves synthesizing and analyzing large amounts of data to help public health professionals identify effective interventions and strategies to address complex public health issues. Creating mathematical models and then applying them through computer simulation and analysis is a practical means of assembling and analyzing public health data in situations where more traditional forms of investigation are difficult due to financial, logistical, or temporal restraints and other obstacles.
“We are pioneering a new direction for public health investigation, one that complements the more traditional approaches of observational data analysis and experimentation,” said Crawford. “By focusing on the explicit portrayal of real-world processes and deducing how intervening in those processes may affect the future health of populations, we are generating new evidence that could not otherwise be obtained.”
Crawford attributes the modeling group’s success to its broad range of scholarly disciplines and public health outlooks, including operations researchers, economists, epidemiologists, doctors, toxicologists, biostatisticians, evolutionary biologists, and decision scientists.
“Our greatest strength is our people: their dedication, interests, and expertise,” said Crawford. “We aren’t wedded to a single operational method for identifying and interpreting the complex mechanisms that drive the health of populations.”
Cross-disciplinary Contributions
Crawford’s methodological interests include epidemiological models, networks and graphs, stochastic processes, semi-parametric inference, causal inference, computational statistics, optimization, and algorithms. His applied work is focused on solving the most challenging inferential problems in epidemiology, public health, and biomedical science.
“Throughout my career, the major theme in my work has remained constant: I seek to apply advanced computational and mathematical tools to solve the world’s most challenging problems in biology and public health,” said Crawford.
Crawford joined the Yale School of Public Health in 2012 as an assistant professor of Biostatistics. Over the past decade, he has taken on several new positions. He is currently an associate professor of Biostatistics, Statistics & Data Science, Management (Operations), and Ecology & Evolutionary Biology at Yale. He is also affiliated with the Center for Interdisciplinary Research on AIDS, the Institute for Network Science, and the Computational Biology and Bioinformatics program.
“My career has benefitted from diverse training and experience in biology, epidemiology, medicine, public health, mathematics, and statistics,” said Crawford. “This has resulted in important contributions to innovative cross-disciplinary research in biology, medicine, and public health.”
Another member of the Department of Biostatistics modeling group, Associate Professor Joshua Warren, has also made significant contributions to public health research, including important work during the COVID-19 pandemic. Warren co-authored a study led by Yale MD/PhD student Emmanuella Ngozi Asabor that found that, due to structural racism, Black and Hispanic neighborhoods had access to fewer COVID-19 testing sites during the early stages of the pandemic.
Warren helped analyze census data used in the study that, when compared to testing site locations in some of the country’s largest cities, revealed glaring gaps in testing access.
“Public health data can be messy, and as a result, average analysis tools aren’t appropriate,” said Warren. “New biostatistical methods allow us to overcome some of those limitations and investigate interesting and important questions.”
Warren also helped Crawford in his study using mobile phone data to predict COVID-19 exposures. To validate the contact metric, they developed multiple models to determine the association between the mobility data and case counts.
“When we observe similar findings across different approaches, it gives us more confidence in the robustness of the findings,” said Warren.
Both Warren and Crawford agree that the integrative nature of modeling is best-suited to address complex public health issues.
“Models serve three purposes for us: to formally represent our knowledge about the mechanisms and dynamics of a process, to help us learn about those mechanisms using data, and to allow us to predict what would happen to that process if circumstances were different,” said Crawford. “Statisticians in the Public Health Modeling Unit engage in all aspects of modeling.”