Exploration of NFIRS protected populations using geocoded fire incidents

The NFIRS database is one of the primary resources for data relevant to fire statistics in the United States. This database is populated by self reported fire incidents by U.S. fire departments. A major criticism of this database for statistical analysis is the problem of reporting bias. One cannot use basic summary statistics and other useful, simple metrics to characterize the U.S. fire problem because the NFIRS database does not mandate reporting from the entire nation. Thus, it only contains data from a subset of fire departments across the nation. Additionally, no major attempts have been made to remedy this problem because U.S. fire data consumers rely on the only other data source (NFPA estimates) at the national level. This paper explores approaches to better characterize the NFIRS reporting population. Public data release records from the NFIRS database were geocoded, and then spatial data processes were applied to obtain a reasonable estimate of the protected populations of NFIRS reporting fire departments. The analysis indicates that between 71 and 83% of the American population is covered by NFIRS reporting fire departments. Additionally, estimates of fire incident, fatality, and injury rates from NFIRS reporting populations appear to be roughly 20%-50% lower than the national rates as reported by the NFPA. These rates are confirmed to be feasible by additional analyses that also supply estimates of the non-NFIRS-reporting population's relevant incident and fatality rates. Regional estimates of the NFIRS-reporting population's fire problem are also generated. Until now, there had been no means of generating an independent assessment of the US fire problem. This new modeling approach provides researchers with a publicly available toolset to conduct rigorous sensitivity analysis on the impact of fire prevention and mitigation approaches at the national level.