These are the stories of researchers, epidemiologists, and physicians racing to understand and predict the COVID-19 pandemic in their cities, told each week one city a time…
In a small, rural hospital in upstate New York, Dr. Asad Nasir sat in front of his computer adjusting a slider on a website that lets you quickly model different outbreak scenarios. Unlike the sprawling urban center of New York City a few hours to the south, there were no teams of data scientists and epidemiologists flocking to model the outbreak in his county. There were no tech junkies turned epidemiologists pouring over the much smaller numbers of incident cases where he lived. The numbers were insignificant to most people — less than 5 cases currently hospitalized. So who cares?
What terrified Dr. Nasir was that their hospital, serving a region of 250,000 people in and around Chemung Coounty, only had 20–40 staffed ICU beds depending on the progress of their surge planning. They could only handle 240 hospitalizations, and only 40 severe cases requiring invasive interventions like ventilation. Any more than that and they were toast.
Dr. Nasir connected with our team at flashpub.io through a facebook group for physicians and hospitalists discussing the looming surge in hospitalizations. With no one to help him model the local rural conditions he was facing, he hoped to join our community of modelers and epidemiologists.
Within a week, Jeffrey Morgan, a PhD candidate working on statistical modeling, Tim Snyder, a data scientist and software engineer, and Emily Eshraghian, a public health masters student at UCSD, were excited to join team Chemung!
“I appreciate all your help … This is sobering.” — Asad Nasir, MD (Medical Director in Chemung County)
The team quickly realized that the biggest source of variance was the nursing homes — if the Washington state nursing home scenario happened here in Chemung, it would be devastating.
This highlights one of the most challenging aspects of modeling outbreaks in an area with a small population. Minor fluctuations in the underlying data, sometimes even a single infection or death, can have a big impact on the model’s predictions.
This is why local models informed by the details of local conditions combined with direct communications with physicians and citizens on the ground can be the difference between an accurate model and a useless model. The widely available, large-scale models predicting state-wide trends do not have the resolution to account for critical local-level shifts in the epidemic. These general models cannot account for how a handful of infections at a single nursing home could devastate some rural county that has a fraction of the state’s population.
Residents and health professionals of rural counties need and deserve thoughtful, accurate models to predict their outbreak as much as those living in big cities. The information afforded by such models enable critical public health planning and can save lives. We’re hoping to replicate our efforts from Chemung in other rural counties that are currently under-resourced and need help.
If you work in a rural hospital and need help modeling your local COVID-19 outbreak, get in touch at firstname.lastname@example.org.