Amanda Kimball

Do we have enough good data to make evidence based choices for smoke alarms?

Blog Post created by Amanda Kimball Employee on Mar 4, 2016

Joseph M. Fleming, Boston Fire Department, argues that we don’t.  In his presentation at SUPDET 2016, “Improving Data Collection of Smoke Alarms in Fires,” Fleming points out the investigators may be prone to bias, assuming fire alarms did or didn’t go off based on their expectations regarding survival likelihood.  If no one escapes, they might simply assume that the alarm was not triggered.  Investigators may even code the NFIRS Detector Operation Field as “failed to operate” or “undetermined” based on their assumptions despite witness testimony to the contrary.  Regarding the “Detector Failure Reason” coding, he states, “Most fire chiefs are not trained to make this determination. When this form is being filled out right after a fire, how would the fire chief know if the victim was impaired or deaf?  What if the victim was alerted but alerted after it was too late?”  All of these may lead to bad data and thus to incorrect conclusions.

 

6a00d8351b9f3453ef01bb08c29114970d-550wi.jpgFleming recommends working to improve our data and the consistency/accuracy of coding, as well as applying data analysis to fire fatalities in different areas, keeping in mind their different requirements, to shed light on the number of flaming vs. smoldering fires, the number of furniture fires, the efficacy of smoke detectors, and other issues.

 

This echoes the calls we’ve heard throughout the week, most particularly at Workshop on Big Data and Fire Protection Systems, for better data combined with robust algorithms to help us tease out patterns and solutions which will allow us to make the best evidence based recommendations.

To access the SUPDET 2016 presentations and papers, please visit the proceedings website.

Outcomes