Perhaps the first rule that I must make for myself is
to pay attention to what I see. Italo Calvino
During the initial stages of the COVID-19 pandemic, public health communities around the world were facing a daunting challenge. The immense scale, dynamic nature, and severity of the pandemic required swift response to prevent infections and protect local communities. Yet, little was known about how the new SARS-CoV-2 virus was spreading among local residents, or the health risks it posed. I was an assistant professor at UTHealth School of Public Health in Houston, Texas, focusing on nutritional epidemiology research and teaching Epidemiology Methods to future public health professionals. As the pandemic continued to ramp up worldwide, I was invited to join my School of Public Health colleagues in tackling this pressing issue. The team featured experts in the fields of epidemiology, biostatistics, public health policy, geospatial analytics, and intervention research. We began to look at COVID-19 community and hospital data from Harris County, in southeast Texas, one of the largest and most diverse metropolitan areas in the United States.
Week after week, my colleagues and I met with local public health departments and health care leaders to share information on the growing outbreak in our region. Each week, new, updated data was processed, cleaned, and analyzed within a short turnaround time so that our findings could shed light on the magnitude and severity of the outbreak in as close to real time as we could get. We also developed methods to identify "testing deserts" and detect significant health inequities in the region. These findings were then presented to local public health partners, informing resource planning and rapidly evolving strategies to respond to the emergency in Harris County.
Once the first outbreak began to reduce in intensity, it was clear that we needed to prepare for a second wave. As the analyses continued, we left our weekly meetings with a shared sense of surprise at the resilience and unpredictable nature of the pandemic. Results from early models predicting the timing and magnitude of a new outbreak wave were not fully confirmed by the data on the ground; partially because of the complex, fast-changing nature of the pandemic, and the insufficient data on factors influencing viral transmission (e.g., mask use, human mobility, new variants). Was it possible to predict what was coming next?
At that time, a new idea emerged: looking at the problem through the lenses of classical physics. According to Newtonian mechanics, the greater the speed of a particle at a given time (i.e., its momentum), the greater the amount of force or longer the amount of time needed to stop the particle from moving. Similarly, the faster the change in its speed (i.e., the acceleration), the greater the response required to slow down a given object. If we could borrow Newton’s idea, the number of new cases detected daily (i.e., growth rate) could be seen as the velocity at which SARS-CoV-2 virus was traveling within a certain region. Both growth rate and its daily change (i.e., growth acceleration) could provide important clues to the dynamics of community transmission, and about the intensity of the response required to change its trajectory.
Around the same time, we came across a new article by Dr. Yuri Utsunomiya and colleagues at São Paulo State University in Brazil. Utsunomiya proposed the use of Hidden Markov Models to estimate COVID-19 growth rate and acceleration in real-time, and to identify outbreak stages (exponential growth, deceleration, stationary) using data from various countries. I reached out to Dr. Utsunomiya, who kindly shared the code developed as part of his work. We built on and expanded his methodology to estimate the dynamics of infection transmission among residents in the Harris County area, and the time period for an outbreak wave to propagate from the community to local hospitals. We quickly began to report current outbreak stage and acceleration rates among county residents, as well as among specific groups. This approach shed a new light on transmission patterns among individuals sharing common behaviors and/or health risks that could be tackled by different strategies. For example, a change in momentum among children, young adults, or senior residents helped inform discussions surrounding strategies to engage local schools, bars, restaurants, and allowed insights about resources needed to care for older adults at risk of serious illness.
Among many lessons learned from this ongoing work, I was particularly struck by the assumption underlying Markov models. The future is independent of the past and is determined only by the present. In a period when the entire world was frantically trying to predict what happens next week, next month, next summer, Markov models suggest that starting from the present moment is not only a clever idea, but an indispensable step. The present. This tiny instant is often thought of as unimportant, as many of us spend more time thinking about the past or worrying about the future. Yet, when we look at it carefully enough, the present is so full of all that has preceded us. The present moment is always in action. While many of us will continue to strive to understand life and grapple with its complexity, this experience suggests that the magnitude and direction of a changing instant contain the energy necessary to unleash the ingenuity, creativity, and compassion required to respond to public health challenges and hope for the future.