This story appears as a part of Open Data PGH, a joint reporting project by Technical.ly and PublicSource on open data trends in Pittsburgh, underwritten by Heinz Endowments. Learn more here and get updates here.
Allegheny County has been host to a consistently high fire fatality rate over the years. In 2016 alone, 22 people died in fires in the county, tied for the unfortunate top spot statewide with the city of Philadelphia.
Local officials wanted to figure out how to better prevent such fatalities, and have since used a remarkable blend of local and U.S. Census data to create a picture for municipal fire departments of the houses most at risk for not having working smoke detectors. The profile created by the CountyStat team could be a game-changer for the Red Cross and local fire companies’ door-to-door smoke detector campaigns.
“We were collecting a lot of data [about fires] and we really stretch that data in every direction you can think of: any fatalities, of course; injuries; municipalities; time of day; day of the week,” said Allegheny County Emergency Services Chief Fire Marshal Matt Brown. “With CountyStat getting involved, we’ve done a better job of finding ways to use that data.”
It’s worth repeating what fire officials have been saying for many years: In a fire, it’s usually the smoke that will kill you. So, getting working smoke detectors in as many homes as possible is a no-brainer. But fire officials and the Red Cross may have limited supplies of smoke detectors. So, how to decide where to distribute them? Which homes are least likely to have them?
Working through this question is a formidable task: The EMS agency advises all of the 179 fire companies in Allegheny County’s 130 municipalities. Its primary role is to investigate the cause of fires (but only when called upon to do so by municipal officials). Brown estimates the agency investigates 450 to 500 fires a year.
“There were a multitude of things we could potentially do, but we wanted to figure out: What’s really going to work?” Brown said.
So Ellie Newman, an analyst with CountyStat, was tasked with creating a model that could help ascertain where these at-risk buildings were.
The fire fatality risk model was inspired by a field trip the CountyStat team took to New Orleans. Local officials in that city decided they needed to try to be more proactive about preventing fatal fires after five family members died in a 2014 fire. The home in that fire did not have working smoke detectors.
The New Orleans Fire Department had a smoke detector installation program in place in which firefighters would visit a home and install a free smoke detector, but it was based on requests — they only went where they were called. So its leadership worked with the local Office of Performance and Accountability to analyze several datasets that would allow them to determine which homes were most likely not to have smoke detectors.
The program, called Smoke Signals, was based largely on U.S. Census data. (The American Housing Survey asks participants, “Do you have working smoke detectors in your house?”)
Newman said CountyStat began with Census data, and compared county data to several different cities.
“We wanted to look at what were the predictive factors that seemed to correlate with people saying they didn’t have working smoke detectors,” Newman said. Poverty levels played a significant role, as did property characteristics: Row houses and three-story houses were among the property types that most strongly correlated with the non-smoke detector homes. “And Pittsburgh’s housing stock has many of those,” she noted.
CountyStat was able, with Brown’s help and contacts at local fire companies, to create a heat map showing the highest-risk municipalities. And they have refined the data over the past year to create a list for the highest-risk municipalities — Pittsburgh, Duquesne, McKeesport and Wilkinsburg — of the homes in their communities most at risk for not having smoke detectors. This helps smoke detector programs become more efficient.
“This helps however they distribute them. Sometimes it is door to door, Newman said of the different fire departments. “But if they don’t have any idea who needs them, they’re going in blind.”
The Red Cross will begin acting on the list this fall for its smoke detector distribution program, Newman said.
Brown added that he understands not all fire companies have the bandwidth to go door to door installing smoke detectors, but likes to see that happen whenever possible.
“Just being there and making sure it’s installed properly, then you have that interaction with the homeowner where you can talk about general fire safety,” Brown said, “it allows fire departments to make those connections to the people they serve.”
Newman said that beyond fine-tuning its smoke detector distribution program, the Red Cross wanted to take the project a step further. The organization provides services to displaced fire victims, and it wanted to know if it could focus its efforts better, Newman said, since it was only going to fire sites when summoned. The theory was that not everyone was aware of the services the Red Cross could provide to fire victims.
“They wanted to know how many fires were in occupied residential buildings,” she explained. “I talked to the 911 center and the fire marshal and it turns out we don’t have a dataset that tracks the number of fires like that.”
So Newman got to work creating a database with the information she did have. But once again, Allegheny County’s fragmented system of government made it challenging to create a complete picture; not all fire departments keep fire records in the same way. Undaunted, Newman did the painstaking work of collecting data from 911 calls and cross-referencing them with the county’s real estate assessment records. All 911 calls where someone reports a structure fire get a code, she explained.
Newman took data from each call to determine its point of origin and tried to match it as closely as possible with the real estate assessment department’s database of residential structures. That helped eliminate about 1,000 properties.
She then cross-referenced the remaining list of about 3,100 fires to the Red Cross’ database of areas where it had assisted with post-fire services.
“There was a lot of discrepancy, and it showed that northern Allegheny County communities were not getting as many Red Cross services,” Newman said. “It seems that maybe they are not as aware of the services the Red Cross provides as some other communities.”
While the smoke detector risk program is a model other communities may want to pursue, Newman said it’s not a cookie-cutter program, since the housing stock plays a big role in creating a complete picture of at-risk properties.
“It would be foolish to apply our model to Phoenix, Arizona, for example,” she said, since the housing stock there is entirely different. Older cities like Cincinnati or Boston may be a closer match, but will still have to factor in their own property profiles, Newman added.
She said the fire safety profile is an example of Allegheny County leadership finding innovative ways to use county datasets to solve problems, sometimes through finding the data’s hidden values.
“Our county database is designed to track services and programs the county administers, so we have all this data designed just to serve one purpose,” she said. “But along the way, these datasets generate all kinds of information we can use to answer other questions.”-30-
3 technologists’ hopes for a more open Pittsburgh
How to build a civic tech community, according to Pittsburgh leaders
5 ways Pittsburgh’s public servants are using human-centered design
Why Deacom’s team prioritizes collaboration and continuous improvement
7 reasons you should attend Open Data PGH’s Future of Pittsburgh Civic Tech event
In Pittsburgh, Harrisburg and Philly, Code4PA takes aim at the opioid crisis
Here’s why Idea Foundry thinks increasing access to entrepreneurship is like civic tech
How SmartLogic accelerated these startups’ product growth trajectories
Sign-up for daily news updates from Technical.ly