The COVID-19 pandemic has fundamentally changed life in the United States. As of May 2020, more than 100,000 Americans had lost their lives and the unemployment rate was near 25 percent.

The response to the crisis has taken shape in several initiatives—many led by statisticians—including epidemiological studies to understand immunity and infectivity, models to predict when there will be a next wave, clinical trials to identify safe and effective treatments and vaccines, and studies of the impact of various policies—including social distancing policies.

The ASA established the COVID-19 Data, Statistics, Research, and Discussion Community to facilitate collaborative science, and more than 600 recently attended an ASA and National Institute for Statistical Sciences informational webinar that focused on the role of modeling this disease. As highlighted in the May issue of Amstat News, statisticians across the spectrum of the ASA community are diligently working to provide the necessary expertise during this uncertain time.

Applied and practicing statisticians in the ASA Statistical Consulting Section are engaged and playing key roles in COVID-19 research. Here, we highlight members of the section who are responding to COVID-19. We also laud those we did not discover, as well as the many others throughout the ASA doing similar work.

Epidemiological Studies of Immunity and Infectivity

A number of epidemiological studies of immunity and infectivity are underway. In addition to The Johns Hopkins dashboard by Wang, et al. highlighted in the June issue, Jason Wilson, associate professor of statistics at Biola University, has highlighted sex as a risk factor in his recently posted paper, “Systematic Review and Meta-Analysis of Sex-Specific COVID-19 Clinical Outcomes.” The paper, posted on medrxiv and accepted for publication in Frontiers of Medicine, shows the incidence of COVID-19 in males is higher than females, particularly as symptoms grow worse.

Edward Boone, professor of statistics at Virginia Commonwealth University (VCU), is working with Ryad Ghanam, professor of mathematics at VCU Qatar, and Abdel-Salam Gomaa Abdel-Salam, associate professor of statistics at Qatar University, to build a Susceptibles, Exposed, Infected, Recovered, and Deaths (SIERD) model for predicting the next wave of COVID-19 in Qatar. The SIERD model allows for quantifying the impact of various government interventions to slow the spread of the virus and predictions of when peak active infections will occur.

To better understand the genetic component of the disease, Jeet Mozumdar, statistical geneticist and bioinformatician, is performing analyses on the DNA of SARS-CoV-2.

Operating Characteristics of Diagnostic Tests for Infection and Immunity

Other researchers are focused on understanding the operating characteristics of diagnostic tests for COVID-19 and providing insight into their interpretation. Naomi Brownstein, assistant member at Moffitt Cancer Center, is working with her colleague Ann Chen on an analysis of positive and negative predictive value (i.e., false negative and positive rates) of antibodies tests under an Emergency Use Authorization from the US Food and Drug Administration. The working title of their paper is “Are Antibodies Tests Accurate? Understanding Predictive Values and Uncertainty of Serology Tests for the Novel Coronavirus.”

Isabel Allen, professor of epidemiology and biostatistics at the University of California at San Francisco (UCSF), worked with colleagues at the University of Pennsylvania and UCSF on a meta-analysis of the sensitivity of CT scans vs. the RT_PCR tests for COVID-19. The paper found lots of biased studies and few with data on both tests. Sensitivity for both tests was close to 70 percent. The paper is currently in press in Investigative Radiology.

Impact of Mitigation Strategies

Isabel Allen additionally conducted a survey of faculty and administrators in higher education (US and Canada) on transition plans and teaching online with Bay View Analytics and sponsored by the Gates Foundation and others. The study found that faculty are concerned about their students and administrators are concerned about financial implications. Also, both groups provided interesting findings on remote education versus online education. Lack of training materials for moving courses online was identified as the biggest issue. A follow-up survey will be conducted in late summer to assess institutions’ plans for the fall semester and what types of support faculty are getting for moving courses online. Findings have been featured in a webinar with Inside Higher Ed (1,500 participants) and in The New York Times.

Clinical Trials and Clinical Studies

Many of our members are engaged in identifying effective treatments for COVID-19 and a vaccine for the novel coronavirus. Frank Harrell and Chris Lindsell, professors of biostatistics at Vanderbilt University, led the novel design of “Outcomes Related to COVID-19 Treated with Hydroxychloroquine Among In-Patients with Symptomatic Disease” (ORCHID). ORCHID is a Bayesian sequential parallel-group randomized clinical trial for COVID-19 that allows for continuous learning from data through the computation of probabilities that trigger go/no-go decisions about how to proceed with the trial. This work—done closely with David Schoenfeld, professor of biostatistics at Harvard University—involves developing a Bayesian sequential design and simulating its Bayesian operating characteristics in addition to developing Bayesian ordinal regression models so the ordinal outcome can be covariate-adjusted and the model extended to serially collected outcomes. Harrell and Lindsell have provided a detailed blueprint for our community, along with software, so others can borrow principles in designing similar studies.

Manisha Desai, professor of medicine and biomedical data science at Stanford University and chair of the ASA Statistical Consulting Section, has worked with her team at the Stanford University Quantitative Sciences Unit to create a shared infrastructure for facilitating randomized clinical trials (RCTs). This includes an international data and safety monitoring board registry of experts willing to serve on COVID trials, searchable by expertise and hosted on the Society of Clinical Trials (SCT) COVID-19 Research Resources Hub.

Also hosted on the SCT hub is the COVID-19 Endpoint Registry, which lists all approved endpoints in the US. Notably, coming up with the right COVID-19 endpoint has been a huge challenge, as there are no validated endpoints. In this registry, one can search for endpoints for a particular patient population or study phase and one can graphically view the diversity of endpoints and changes in endpoint choices over time. The most important component of the shared infrastructure is an adaptive master trial, which allows evaluation of multiple agents to be studied simultaneously. Such a trial allows arriving at answers about the safety and efficacy of drugs faster and with fewer resources than a traditional fixed trial design. Efficiency is gained as one does not need to reinvent the wheel each time a trial is being launched, and consolidating into one trial means less competition among individual trials that need the same patient population to address questions about a particular drug.

Robert Podolsky, director of informatics and biostatistics at Beaumont Health, has helped design the initial sampling strategy for the Beaumont employee serology study. This study focuses on determining antibody responses to COVID infection, estimating the prevalence of antibodies among all Beaumont Health employees, the ability to identify patients who never experienced symptoms, and the sustainability of antibody production. He has also designed a clinical trial, the SINK study, to evaluate the efficacy of using Naltrexone and Ketamine as immunomodulatory agents in trying to control the inflammatory response to the COVID virus.

In addition to clinical trials, members have designed studies to evaluate the effectiveness of various quality improvement initiatives in the clinical setting. Joseph Rigdon, assistant professor of biostatistics and data science at Wake Forest School of Medicine, is working with colleagues in infectious diseases to study the impact of wearing a mask on the amount of virus present in the air and on surfaces in hospital rooms using data from hospital patients with COVID-19. This study will guide best mask practices for providers during this and future similar pandemics. Additional work with colleagues in pediatric nephrology and epidemiology focuses on guidance for the design and analysis of observational studies of rates of COVID-19 and related complications in patients with hypertension who take ACE inhibitors.

Through collaborative efforts relying in part on statistical practice, we have learned much in the past six months about COVID-19. At least one treatment has emerged as promising for the severe patient. We are observing a number of disparities among subgroups of the population—with LatinX and African American populations more vulnerable than other groups—and mitigation strategies such as shelter-in-place are effective in reducing case burden (doi:10.1001/jama.2020.8598).

Importantly, there is much more we need to learn. We need to better understand transmission, immunity, and the impact of various reopening strategies. We statisticians have trained for moments like this one and are being called to action. The work highlighted here shows how many of us have pivoted from our usual work to address this crisis. It also provides hope that, with so much talent and energy devoted to the crisis, we will emerge with solutions for this and future pandemics.