Making fire risk assessment real: Using data to inform fireground decisions

A case study highlights how data can impact risk assessment and resource deployment models


By Steven Knight, Ph.D.

The fire service often self-identifies as an “insurance policy.”

The origins of America’s fire service and fire insurance industry are attributed to Ben Franklin, who started the first volunteer fire company, the Union Fire Company, in Philadelphia in 1736. That later contributed to the formulation of the Philadelphia Contributionship for the Insuring of Houses from Loss by Fire in 1752. Interestingly, a “surveyor” was utilized to inspect buildings for insurability – an indication of early fire-related risk assessment.

As the purveyors of risk assessment, the insurance industry has a wealth of information for us to learn from and evolve our understanding of risk. (Photo/Getty Images)
As the purveyors of risk assessment, the insurance industry has a wealth of information for us to learn from and evolve our understanding of risk. (Photo/Getty Images)

I would suggest that in the modern fire service, the greatest influencer for utilizing community risk assessments in creating risk-based deployment strategies is the Center for Public Safety Excellence’s Commission on Fire Accreditation International (CPSE/CFAI).

The international fire accreditation process necessitates a community-wide risk assessment and a correlated decision model for how resources are allocated and deployed in an effort to match risk mitigation efforts to identified risks. By design, the CFAI model is not overly prescriptive in methodology, which allows local agencies the flexibility to fine-tune to their unique environments and experiences. Ultimately, the model requires reasonable and articulable methods that are defensible within the peer-driven process.

Case study: Illustrative application of occupancy-level risk assessment

An urban county with a population of approximately 1 million and nearly 100% buildout was utilized to evaluate the response history and, loosely, the efficacy of the occupancy-level risk assessment process.

At the time, 33% of the agencies were accredited by the CFAI and were familiar with the risk assessment process, although a uniform application for risk assessment was not utilized across the entire county. Comparably, the building stock is not “old” as compared to other areas of the country.

In total, 3,715 high-risk structures were identified and, as designed, resource allocation decisions were adjusted to ensure an appropriate correlation between the number and type of responding apparatus and personnel.

Over a five-year period between 2009-2013, 130,374 calls occurred in the 3,715 high-risk structures. At a high level, the large call volume is enticing that the risk assessment process was an accurate predictor of call activity. However, EMS activity accounted for 71% of the responses to these high-risk structures. In other words, the risk matrices were generated by and for fire risks – and the overwhelming response was for EMS.

An additional 12% were attributed to “other/good intent,” 10% to “dispatched and canceled en route,” and 7% to “fire alarms” (without fires).

During the five-year period, 360 building fires did occur in these high-risk structures (~0.3%), but were relatively low in severity, averaging less than $10,000 in damage per incident and 18-minutes per incident. Over the five years and 360 building fires, there were four incidents with fire injuries, and each was a singular injury. There were 19 civilian injuries that included three occurrences of two injuries each. In other words, 13 singular injuries and three incidents of two each for the remaining six injuries over five years. Interestingly, the frequency of non-fire incidents resulted in a total of 41 injuries to firefighters and 14 civilian injuries.

So, what are we trying to measure?

The illustrative case study is utilized to not only demonstrate that we have great opportunities for improvement in how we assess risk in our communities, but also how we are applying the information produced in our risk assessment processes.

The first question should be “What are we trying to accomplish?” For example, if the above case study, and associated risk assessment, is to identify fire risk and have an appropriate mitigating firefighting resource allocation assigned, then one could argue that the historical data may not be well aligned. I will refer to this as “Risk for Deployment.”

However, functionally, the extremely low frequency of building fires at 0.3%, having a built-in higher concentration of firefighter resources and apparatus dispatched at the beginning of the incident, is both proper and inconsequential to system draw-down. In other words, it would be entirely appropriate and acceptable to throw the kitchen sink at it if and when it occurs, because the frequency is sufficiently low not to draw down resources. Similarly, in the absence of active fires, such as fire alarms, a measured response should be utilized even in structures identified as high risk.

If the desired outcome is to identify which occupancies/structures should receive prioritized inspections, referred to as “Risk for Prevention,” then the results may begin to diverge. For example, in the case study, many of the high-risk structures were high-rises built up to reasonably modern codes, and therefore have passive mitigation strategies employed, such as monitored alarm systems, automatic sprinklers, smoke doors, etc. These same occupancies are generally on fire prevention divisions’ annual cycle already, which may serve to limit the impact of the risk assessment feedback.

What is missing is the remaining occupancies that may receive some form of a tiered inspection strategy but are not captured as high-risk from the firefighting lens. For example, a simple retail or small office may only be inspected biannually or triennially. Similarly, residential occupancies, where most fires occur, are outside of the legislative ability to inspect. So the lens for which the risk assessment is supposed to inform is important to identify at the beginning so a one-stop application isn’t applied to multiple competing lenses, even though all aspects are intended to reduce the frequency, loss of life, and loss of property, or some portion thereof. Of course, I would be remiss if I didn’t reinforce that the efficacy of prevention far exceeds the efficacy of the response forces.

Opportunities for the future

While the current strategies may be informative and serving us well during our evolution, there are ample opportunities for improvement. Dr. Hinds-Aldrich wrote a recent article that provided a great overview of the evolution of fire risk prediction efforts and models. While more advanced analytical approaches to risk assessments, and fire risk prediction, are coming of age, one notable element was universal in Dr. Hinds-Aldrich’s article – that it was unknown or unclear if the fire departments were utilizing the information. In other words, even though the world of analytics may be advancing, either their applicability or acceptance for fire departments were challenged.

The fire service as an insurer

If the fire service is to be viewed as an insurance policy, then the fire service should adopt a few elements from the insurance field.

As the purveyors of risk assessment, the insurance industry has a wealth of information for us to learn from and evolve our understanding of risk. First, the fire service has significant challenges in the robustness, accuracy, applicability and even digital access. Much of what we measure for risk is a variable of convenience through the lens of response-related fire suppression activity. By definition, I would suggest that incidents have fully transitioned from potential or prospective risk when the bell rings for an actual event. It is common for insurers of fire peril to focus on zip code areas, and may be more influenced by socioeconomic and demographic indicators, such as credit scores and other consumer-related information, than of what the fire service has historically valued.

In the absence of more accurate and dynamic data sources, the fire service may need to continue to be more reliant on retrospective or historical experiences as a surrogate predictor of future risks (at least for risks for deployment). The underlying socioeconomic, demographic and behavioral influences for risk do not change rapidly and can be easily understood over time.

From a policy perspective, it would be difficult to make resource allocation changes based on potential risks that aren’t evidenced. For example, our understanding of the impacts of being a renter in the 1970s, when much of the socioeconomic and demographic research within the incidents of fire were completed, may not be universally held today after the housing crisis in 2008. Each community is more nuanced, and potentially why the CFAI afforded more flexibility in the risk assessment methods.

About the Author

Steven Knight, Ph.D., is the fire service practice lead and partner at the public safety consulting firm, Fitch & Associates. Contact him at sknight@fitchassoc.com.

Join the discussion

Copyright © 2020 FireRescue1. All rights reserved.