The agenda that day addressed the potential for a dense yellow smog cloud to roll into Pittsburgh, as it did in in 1948 in Donora, a small mill town 24 miles southeast of Pittsburgh. That October day, a weather inversion event trapped poisonous gas from nearby steel plants into the atmosphere, awakening residents to burning sensations in their throats, eyes and noses. Patients overwhelmed local hospitals in respiratory distress.
By the time the air cleared five days later, 20 people had died. The next time Pittsburgh encounters a severe natural disaster, local emergency response officials will rely on predictive data software to detect early warning signs and guide responsive efforts.
Emergency medical services personnel are increasingly adopting predictive modeling software that identifies patterns of geospatial data to predict future events. The automated platforms can call in extra ambulances from neighboring counties and direct the vehicles into areas with the most vulnerable populations before floods of 911 calls begin. Early detection also allows time to set up emergency shelters and overflow hospital rooms.
The technology is moving from descriptive analytics (what has happened) and diagnostic analytics (why it happened) to predictive analytics (what will happen). Data industry practitioners say the next wide-scale adoption will be prescriptive analytics, or data that can simulate the outcome of different tweaks to emergency response systems.
“We are really just at the cusp — not just as a company but as a sector. We’re getting to a point where we can match those technologies to get to a desired outcome,” said Justin Schaper, senior vice president of analytics with Intermedix, a Nashville-based predictive analytics company.
Intermedix produces Optima Predict, a suite of software that collects input from emergency response systems in order to optimize ambulance routes and, once the vehicles have entered areas of large medical demand, reduce the response time for checking patients into hospitals.
The data can be synced with FirstWatch, an online dashboard for EMS personnel to read information about 911 callers and responder systems in real time. The software can alert key officials to viral outbreaks or crime activity by spotting geographic clusters of reported incidents before humans notice the trend.
“Just like when grocery stores stock up on flu medicine, it’s about predicting supply and demand. But this is harder, with a geographic component and people can die,” said Todd Stout, founder and president of FirstWatch.
The Pittsburgh emergency preparedness workshop in mid-April provided scenario planning to EMS personnel, academics, IT professionals, and local and federal government officials. The seminar centered around emergency response technology was aimed at preparing the region for an equivalent event to the 1948 Donora smog.
The University of Pittsburgh Graduate School of Public Health showcased an online tool that day that uses public health data to simulate the spread of infection across populations within a specific geography.
The system, FRED (short for a Framework for Reconstructing Epidemiological Dynamics) pinpoints how different neighborhoods would be affected amid a wave of air pollution, extreme heat or viral pandemic. Researchers have used FRED modeling for influenza, measles, mosquito-borne diseases, HIV and hepatitis C.
“We can follow the chain of infection from one person to another, from schools into households and neighborhoods,” said John Grefenstette, a health policy and management professor at the University of Pittsburgh.
FRED data is open source and can map the population of any geographic region in the nation, Grefenstette said. The system is a currently a research tool, but university researchers are building a Web interface and hope to release the website this year.
Embracing data analytics may rely on a cultural shift in some public departments that have grown to trust anecdotal experience over computerized systems, said Chris Callsen, a vice president at Intermedix. Putting data to use can represent a difficult shift for experienced industry practitioners, he said.
“Every organization we encounter in emergency services has huge amounts of data. They’ve been archiving for years — most have not made the transition to using it,” Callsen said.
Anecdotal “experience can be augmented by data and you have to trust in that,” he added.
The value of the day-long Pittsburgh workshop was to bring lots of unlikely actors together around a challenge, said Grant Ervin, chief resilience officer of Pittsburgh.
Relationship-building is just as essential as the software on display, he said. Optimal emergency response relies on a talented, cohesive team who can stay nimble and adapt as numbers change on the screen in front of them.
“The predictive analytics stoke the thinking and guide the mission, but at the end of the day it’s up to practitioners to work well together,” he said.