KAUST Professor of statistics Hernando Ombao. File photo.
The global, multifarious challenge posed by the COVID-19 pandemic has scientists tapping their wide-ranging fields of expertise to attack the problem on many fronts. Answering the call from KAUST President Tony Chan, and coordinated by the University's leadership team, KAUST researchers making up the Rapid Research Response Team (R3T) are turning this crisis into an opportunity to innovate.
At the center of public health challenges posed by the pandemic is the stress that the spread of the virus has put on hospitals and clinical staff. Some researchers are using forecasting models to help prepare by projecting future waves of hospitalization.
"Preparedness is key to reducing fatality," said KAUST Professor of statistics Hernando Ombao. "If a country, or if a community, is able to anticipate the need for hospitalization, then maybe they can reduce fatality rates. One of the main reasons for high fatality rates is because the hospitals are just completely overwhelmed."
Professor Ombao heads up the Biostatistics Group within the University's Computer, Electrical and Mathematical Sciences and Engineering Division (CEMSE). The KAUST Biostatistics Group develops novel statistical methods and models for biological processes with complex dependence structures. They use tools like time series analysis, spectral analysis, computational statistics and data visualization.
One of the areas Ombao and his group have been focused on is how the brain responds to stimuli. A sound statistical approach is needed to understand how the different regions of the brain communicate with each other. The research team observe brain signals of rats and attempt to predict their decisions.
Specifically, the statistical analysis is being used to understand and model how a rat's brain can respond to certain types of events such as a stroke. Using techniques like stationary subspace analysis (SSA), researchers might be able to reduce the dimension of the data and improve prediction performance.
"For me, I've been very interested in how a pandemic does a shock to the entire system," Ombao explained. "So with the entire system, I can focus on the impact of hospitalization."
Ombao aims to develop statistical models that are geared towards obtaining accurate estimates of these projections. "If we are able to give our public health officials and the hospital administrators an accurate set of scenarios, from the optimistic to the pessimistic, then they'll be able to use it for preparedness," he said.
Models can be adjusted by tweaking mitigation factors such as infection rate, susceptibility rate, as well as recovery rate. Then, based on these inputs, it's possible to generate curves of plausible scenarios. The data can be visualized, replicated in graphs and saved as inputs into further analysis.
"Knowing about the mathematical models for epidemics is new, so we're learning as we go," Ombao said.
Ombao and his team of researchers, however, are not alone in navigating these new waters of mathematically modeling pandemics like COVID-19. The Biostatistics Group is very grateful for the input of David Ketcheson, an associate professor of applied mathematics and computational science at KAUST.
An SIR model is essentially an epidemiological model used to compute a theoretical number of people infected with a virus in a closed population over time in order to better understand the spread of the infectious disease (Susceptible/Infected/Recovered). Toolbox development is in an integral part of the research and is used to improve the impact of the work.
The R Shiny toolbox is quite interactive. An SIR simulator is available on Ombao's website. It provides a rough estimate of factors such as transmission rate and infection period.
"We can input that into our simulator and then it will give you a broad range of plausibilities with regards to what is the expected number of infection on each day," as he outlines.
Ombao points to the example of the Imperial College COVID-19 response team who has made available useful data to the UK government and is advising the prime minister's office about intervention effectiveness and estimates on hospitalizations and fatalities.
Ombao's team is specifically focused on collecting and analyzing data to predict patient admissions and hospitalization rates, based on daily infections data, at the hyperlocal level on the KAUST campus.
"Of course, applying the SIR model, really you have to look at large populations. At KAUST, we only have about 7,300 members of the community. Nevertheless, it is a good step," he explained. "So, the Biostatistics Group must also use broader data sets from other countries and adapt the demographic parameters to fit the KAUST community context."
This information is useful for the KAUST Health emergency team to forecast individual patient length of hospitalizations and, on any given day, assess bed occupancy and manage resources both at the clinic and in Jeddah hospitals.
KAUST scientists are also collaborating to provide insightful data to policymakers and government officials at the national level in Saudi Arabia. Faculty members, including Carlos Duarte, David Ketcheson, Paula Moraga, Xin Gao, Arnab Pain, Xiangliang Zhang, Takashi Gojobori, Ombao and others, are involved in developing and maintaining a dashboard accessible to public health authorities and also to the general public.
The COVID Compass, a global task force that the KAUST researchers have teamed up with, offers a real-time data dashboard "publishing relevant, relatable and actionable data around the COVID-19 pandemic."
"We plan to do in-Kingdom forecasts and we have to adapt our proportions of hospitalization so that they are more reflective of the Kingdom's demographics rather than the demographics in China and the U.K.," said Ombao, who contributes to the Analytics and Modeling portion of the COVID Compass team.
While the scientists aren't enjoying the benefits of a controlled environment for these experiments, it's interesting that results can be obtained by running statistical and mathematical simulations where certain settings, like curfews or lockdowns, are introduced.
"It is possible to do these kinds of modeling and put some weight on the different kinds of interventions that have been imposed—then one can fit regression models to assess the impact of them," Ombao said.
But, he cautioned, "it's not always straightforward because we are dealing with observational data rather than real experimental variables within a controlled environment."
Certain datasets with more empirical value would be geolocation data via mobile devices. Information that could be gleaned from cell phone data includes whether or not social distancing regulations are being followed. But privacy and logistical issues pose challenges to this method.
A critical aspect which statistical models can shed light on is how many lives can potentially be saved through interventions.
"We looked at the hospitalization [data from the U.S.] at [its] peak, and what would have happened if there was no initial intervention," said Ombao. "We estimated that the number of hospital beds demand were reduced in the hundreds of thousands."
When observing data from various countries and cities, there's a clear correlation between the introduction of interventions as a mitigating factor and the reduction in the effective reproduction of the infectious disease.
"What's more important is not really the transmission rate but rather the transmission rate multiplied by the length of the infection period. And so this product is equal to the effective reproduction number," Ombao said. "So here is where the intervention can really have an impact. If the intervention is to shut down the schools and workplaces, then you essentially reduce your number of contacts per day."
But as Ombao acknowledged, such necessary measures will undoubtedly leave their mark.
"A lot of us will survive the virus as a result of this physical distancing, but on the other hand, a lot of us are going to be mentally scarred with this experience. So I think that's another aspect. It would be very interesting to measure—while we measure physical distancing—what kind of social interactions people make even in the physical distancing setting."