View this email in your browser

As scores of medical experts and institutions worldwide work to develop treatments and vaccines to combat COVID-19, research scientists are racing to build data-based predictive analytics to help medical professionals and policymakers make informed decisions to protect public health.

That effort is well underway in Tampa Bay, under the moniker of Project ASPeCT (Anonymized, Systematic, Population, e-Geofenced, Contagion, Training model).

What ASPeCT is finding is a direct correlation – with a 21-day delay – between the increased or decreased movement of the region’s residents, and the subsequent number of hospitalizations due to the virus.
“When we see a spike in movement, 21 days later we see an increase in COVID-19-related hospitalizations in the Tampa Bay region,” said Dr. Matthew Mullarkey, Ph.D., director of the Doctor of Business Administration Program for the USF Muma College of Business, who runs Project ASPeCT.
Armed with the ability to identify the timing of potential surges, Mullarkey believes that medical officials could appropriately prepare with staffing, personal protective equipment, ICU beds and other necessities.
"Policymakers, equipped with the same evidenced-based, Tampa Bay-specific information, could use these dynamic modeling insights to prevent more people from contracting COVID-19," added Dr. Marissa Levine, director of the USF Center for Leadership in Public Health Practice.
A region-specific model

Applying data from mobile devices is nothing new when it comes to measuring the spread of the virus. Johns Hopkins, the Centers for Disease Control and Prevention, and countless other research institutions provide continuous national analysis and predictions using similar aggregated mobility data.
What is new in the USF model, however, is the granularity with which Tampa Bay officials can predict outcomes based on behaviors specific to our region.
“Movement is a very localized phenomenon with different regions – even in the same state – exhibiting very different patterns. Knowing movement for Miami does not necessarily help healthcare officials in Tampa Bay anticipate demand,” said Mullarkey.
More than 3.5 million mobile devices are in use daily in the 13-county area considered, for Mullarkey’s purposes, as the Tampa Bay region. Working with a third-party mobile data aggregator, the project team receives aggregated data on total distance traveled, number of unique locations visited, and total number of “dense” locations visited (where more than 10 devices are co-located).

“We use these mobility indicators to build a machine learning model that allows us to extract the mobility patterns across the region,” said Dr. Wolfgang Jank, Professor SISM, who leads the data science effort of project ASPeCT. He then uses these patterns to forecast the number of hospital admissions for the 52 hospitals in the region.

Trends in New Daily Hospital COVID Admissions continue a steep, exponential rise from the second week of June.

Mobility patterns linked to hospital admissions

The way to contract COVID-19 is to come into contact with someone who has it. Chances of getting sick are very low if you don’t. But if you’re meeting multiple people, particularly in dense locations, your risk of becoming infected with the virus increases.
During the statewide stay-at-home order, movement data clearly demonstrated that Tampa Bay residents, for the most part, stayed home and avoided densely populated locations.
“In fact, measurable mobile device movement in the region fell by 50% from an already reduced rate at the start of March,” said Mullarkey.

Since local mask orders were implemented over the last two weeks, weekly mobility fell by 14-18% in total distance and unique locations visited and by 32% in number of dense locations visited.

Right around April 21, the ASPeCT team noticed that movement had started to grow significantly in the region. By the announcement of the Phase One re-opening the first week of June, mobile device movement had increased by 80% and movement to densely populated locations had increased by 140%. In direct correlation with the increased movement, 21 days after the first phase of reopening, daily hospital admissions grew from 20 to 40 across the region.
The Phase Two re-opening, immediately following Memorial Day weekend, sparked a significant spike in movement. Within three weeks, hospital admissions grew to 80 per day and are now reaching numbers in excess of 100 per day across the region. Overall, the data shows that total movement of mobile devices grew by 150% over mobility observed during the stay-at-home order and the average number of dense locations visited daily grew by 220% across the region.

New Daily Hospital COVID Admissions continue to be positively correlated with mobility that occurred in the three weeks prior.

“When we stayed at home, on average, no one was going to dense locations,” Mullarkey said. “But we can see how that changed very dramatically on May 25 and in the weeks that followed.”
Data shows that the most mobile people in the late-May period were traveling to two dense locations per day. “This is a massive increase in mobility,” said Mullarkey, “especially when you realize that COVID-19 cases in people aged 65 and over have stayed steady since late May. This increase in mobility is happening in a population that’s aged 20 to 50.”
And this mobile population is not just traveling farther distances, but visiting more unique locations, as well.
As Mullarkey explained: “Half our population has now gone from two to almost eight dense locations per day on average, putting themselves at significant risk of coming into contact with someone with the virus.” 

Curve could be trending in a healthier direction

Despite the recent alarming surge in cases and hospitalizations, Mullarkey’s team sees a positive trend in some of the most recent regional mobility data. In the two weeks since the region’s counties and several cities encouraged or imposed rules for mask use, the actual mobility in the region has dropped back to Phase One re-opening levels. If the models hold, over the following three weeks, hospitals could see a reduction in daily admissions and the curve could trend in a healthier direction.
Based on our region’s behavioral history, there are three potential trajectories, according to Mullarkey.

Predictive modeling suggests that hospitalization trends will continue through week 29 and then, potentially, begin to drop weekly back toward the Phase One opening levels IF mobility remains low, especially in numbers of Dense Locations. It is also possible that mass mask wearing and an increase in non-dense location mobility may also lead to the start of a lowering of hospital admissions beginning in week 30-33 (July 20-August 15). However, new mobility associated with the return to schools can be expected to return new cases and hospital admissions to at least Phase Two levels.

In the first scenario, if residents keep the same or higher levels of movement, and don’t take any new measures to prevent transmissions, then hospital admissions will grow exponentially from here and place overwhelming pressure on the system.
In the second scenario, if residents exercise real restraint in mobility and travel only as needed and significantly limit dense location visits, then hospital admissions will flatten out.
In the third and most hopeful scenario, combining movement discipline with personal transmission safeguards (i.e. face masks, social distancing) will result in a drop in hospital admissions.
The recent decline in mobility leads Mullarkey to believe the second scenario is most likely happening now.
“Prior to people reducing their mobility during the past two weeks, we were definitely stressing existing capacity, and on the way to exceeding available healthcare capacity in our region," said Mullarkey. "Unfortunately, we will not see the full benefit of that change in behavior for at least three weeks.”
Still, the overarching question for our region is clear: What level of risk are we willing to live with?
“The outcomes are up to us, depending on how aggressively we address behaviors,” Mullarkey said. “If enough of us take seriously our ability to limit our level of exposure, we will reduce the rate of spread of this virus dramatically in our region.”

Project ASPeCT is a unique collaboration, initiated by data scientists at the USF Muma College of Business and Tampa General Hospital (TGH), with a transdisciplinary steering committee from across the Colleges of Arts and Sciences, Public Health, Nursing and the Morsani College of Medicine at the University of South Florida.
The Project ASPeCT team is working together to better understand how and where COVID-19 is spreading in Tampa Bay, using anonymous GPS movement data from mobile devices across the region.
To insure privacy and protect the security of device owners, none of the GPS data contains any personally identifiable information or device identification for individuals.
Healthcare data – including COVID-19 related admissions, ICU cases, patients on ventilators, and hospital-recorded deaths – is provided by the Florida West Coast Regional Data Exchange, a collaboration between TGH, AdventHealth, BayCare Health System, HCA Healthcare, Lakeland Regional Health, Lee Health and Manatee Memorial Hospital.
COVID-19 Community Forum:
Can We Predict the Curve?

Wednesday, July 22 from 12:00 to 1:00 p.m.
Join us for a virtual community forum, as we explore the latest findings from Project ASPeCT, followed by a Q&A with data science and public health experts.
The COVID-19 Community Forum is a State of the Region event, produced by the Tampa Bay Partnership in collaboration with Community Foundation of Tampa Bay, United Way Suncoast and the USF Muma College of Business.
COVID-19: Regional Re-Opening Dashboard
View the latest data in the COVID-19 Regional Re-Opening Dashboard, where we track critical public health metrics that will guide the re-opening of our regional economy.

COVID-19: Regional Recovery Dashboard
View the latest data in the COVID-19 Regional Recovery Dashboard, where we track key economic measures and navigate the region's return to prosperity.
This email is produced by the
in collaboration with
Data reported in this email are believed to be the latest available at the time of production, accurate and from reliable sources. For more information, please contact Dave Sobush.


Want to change how you receive these emails?
You can update your preferences or unsubscribe from this list.