Data-Smart City Solutions

Search

By Katherine Hillenbrand

When a fire broke out in a New Orleans home in October 2015, safety procedures worked exactly as they should – smoke alarms sounded, the occupants were alerted to the danger, and all 11 residents safely escaped from the house. Earlier that year, a new program from the New Orleans Fire Department (NOFD) and the city’s analytics team, the Office of Performance and Accountability (OPA), had distributed free smoke alarms to the house based on a calculated assessment of its risk. If the fire had occurred just a few months earlier, the outcome could have been far more tragic.

 

Residents in New Orleans, like in many U.S. cities, have long been able to receive free smoke alarms from their local fire stations. However, relatively few residents actually took the time to schedule smoke alarm installation appointments. In the wake of tragic fire in which three children, their mother, and grandmother died in a home that lacked smoke alarms, NOFD Superintendent Tim McConnell wanted his department to start more proactive, door-to-door smoke alarm installations. However, determining which houses needed alarms “was like finding a needle in a haystack,” according to Oliver Wise, director of OPA.

 

The OPA’s charge is to lead data-driven decision-making and improvements in the city by managing accountability efforts, including ResultsNOLA performance reporting, and working with departments such as NOFD to solve ad-hoc problems with data. Over a period of four months starting in late 2014, OPA worked to identify data and develop models that could identify fire risk.

 

DEVELOPING A MODEL

 

OPA first tried analyzing historical permitting data but quickly discovered that its low quality limited its usefulness. The next strategy the team considered was to conduct a new survey of which structures had smoke alarms in order to create predictions based on the survey results. However, this would have required significant upfront work by firefighters to do door-to-door outreach. The breakthrough idea happened by chance – Melissa Schigoda, an OPA team member working on another project, discovered that the U.S. Census Bureau’s American Housing Survey (AHS) includes a question about smoke alarms. Although the AHS data for New Orleans is only available at the parish level, it has questions that are also asked on the American Community Survey (ACS), which has more granular data available. As a result, the city was able to use the AHS data to identify variables— income as a proportion of poverty level, age of the structure, and length of tenancy— that could predict the presence of a smoke alarm.

“Nothing we did required big data or fancy machines or big tech investments. We are a city of less than 400,000 [residents] and we are strained for resources. If we can do it, anyone can do it.”

OPA then used the ACS data to apply that prediction to individual Census Block Groups. Finally, they combined the smoke alarm prediction with the likelihood of fire fatality risk, which is based on both historical fire data from the NOFD and Census data on the presence of young and elderly residents. The resulting model is a map-based risk assessment of individual blocks in the city.  

Using the tool, the fire department is now going door to door in the highest-risk areas to distribute free alarms. They have installed over 8,000 to date. Wise reflected, “Nothing we did required big data or fancy machines or big tech investments. We are a city of less than 400,000 [residents] and we are strained for resources. If we can do it, anyone can do it.”

 

REACHING NATIONAL SCALE

 

Enigma, a data analytics startup, helped New Orleans with methodological review and strategic support for the effort. Enigma knew that fire safety was a problem reaching far beyond New Orleans. The National Fire Protection Association (NFPA) estimates that there are over 386,500 residential fires per year in the United States.[1] Fire fatalities are twice as high in homes without smoke alarms.[2]  The American Red Cross also has a national Home Fire Campaign, which is working to reduce fire deaths and injuries through smoke alarm distribution and other education and prevention efforts.

 

Enigma saw the opportunity to help address this larger issue in partnership with the Red Cross and DataKind, a pro-bono data science organization. Because the New Orleans risk model was primarily based on federal data, they were able to quickly scale the analysis based on AHS, ACS, and Census data for other cities across the country.  They produced Smoke Signals, a tool that offers block-level risk assessment for 178 American cities and is available as an interactive map and downloadable CSV files. Cities can also upload their own historical fire incident data to improve the model for their area. Marc DaCosta, co-founder of Enigma, said, “For us, one of the main goals throughout the process was to be extremely open and transparent in all of the components: the geocoder, the algorithm, and other tools, with the hope that anyone with an interest can go on GitHub and get them.”

 

This data-driven approach is already saving lives in New Orleans and across the country. Refining a solution to a common city challenge and then scaling it nationwide with the help of private and nonprofit partners is a promising one for the future of civic data.

 

[1] Haynes, Hylton J.G. Fire Loss in the United States During 2014. National Fire Protection Association. September 2015. http://www.nfpa.org/research/reports-and-statistics/fires-in-the-us/overall-fire-problem/fire-loss-in-the-united-states

 

[2] Ahrens, Marty. Smoke Alarms in U.S. Home Fires. National Fire Protection Association. September 2015. http://www.nfpa.org/research/reports-and-statistics/fire-safety-equipment/smoke-alarms-in-us-home-fires.

 

About the Author

Katherine Hillenbrand

Katherine Hillenbrand is the Project Manager for Data-Smart City Solutions. She has been working on the project in different roles at Harvard Kennedy School for over four years. She holds a degree from Amherst College. 

Connect

Comments

comments powered by Disqus