Investigating Smoke Alarm Effectiveness in Fatal Fires

BY JOSEPH M. FLEMING AND VYTENIS BABRAUSKAS

When the issue of smoke alarms and fire investigation comes up, we are reminded of Mark Twain’s wisdom, “It ain’t what you don’t know that gets you into trouble. It’s what you know for sure that just ain’t so.”¹ The fire service in the United States has been so “educated” about the effectiveness of smoke alarms that its members often assume that if someone died, the unit must not have operated. If a smoke alarm is found, then they assume the victim acted inappropriately or that the fire grew so fast that the smoke alarm could not provide adequate time to escape. Such assumptions may be seriously flawed.

Following are sections of National Fire Protection Association (NFPA) 921, Guide for Fire and Explosion Investigations, 2014 edition, that relate to the concerns addressed in this article (italics inserted by the authors).²

Section 4.4.1. Receiving the Assignment. The investigator should be notified of the incident and told what his role will be and what he is to accomplish. For example, the investigator should know if he is expected to determine the origin, cause, and responsibility; produce a written or an oral report; prepare for criminal or civil litigation; make suggestions for code enforcement, code promulgation, or changes; make suggestions to manufacturers, industry associations, or government agencies; or determine some other results.
Section 21.5. Determining Responsibility. After determining the origin, cause, and development of a fire or an explosion incident, the fire investigator may be required to do a failure analysis and to determine responsibility. It is only through the determination of such responsibility for the fire that remedial codes and standards, fire safety, or civil or criminal litigation actions can be undertaken.
Section 6.2.10.3.4. Smoke alarms should be taken into evidence when smoke alarm performance may be an issue. The alarm should be collected as evidence after being photographed in place and should not be altered by applying power, removing or inserting batteries, or pushing the test button. Alarms still on the wall or ceiling should be secured intact with mounting hardware, electrical boxes, and wired connections. Removing a section of wall material with the alarm may be needed to preserve the condition of the alarm and all electrical power connections.
Section 4.3.8. Expectation Bias. Expectation bias is a well-established phenomenon that occurs in scientific analysis when investigators reach a premature conclusion without having examined or considered all of the relevant data. Instead of collecting and examining all of the data in a logical and unbiased manner to reach a scientifically reliable conclusion, the investigators use the premature determination to dictate investigative processes; analyses; and, ultimately, conclusions in a way that is not scientifically valid.
Section 4.3.9*. Confirmation Bias. Different hypotheses may be compatible with the same data. When using the scientific method, testing of hypotheses should be designed to disprove the hypothesis (falsification of the hypothesis). Confirmation bias occurs when the investigator instead tries to prove the hypothesis.

* Explanatory information is contained in Annex A.

Despite this language, unless it is a major fire in terms of loss of property or life, investigations seldom go beyond the cause and origin phase. In fact, when it comes to evidence regarding smoke alarms, some fire investigators feel that it is their duty not to concern themselves with the smoke alarms, since it is a litigation concern. Many investigators seem to feel that their sole responsibility is to investigate potentially criminal issues. But if fire inspectors don’t collect this evidence, who will? Even when local and state fire officials investigate the smoke alarm status, they often do so with biases based on a mistaken understanding of the statistics relating to smoke alarms, the smoke alarm approval process, and published fire tests results. Prior papers have already raised awareness of this problem.^3,4 This article expands on the topic and offers possible solutions.

Bias in the Investigation of Fatal Fires

The need for fire investigators to avoid bias in investigating cause and origin issues has been discussed at length in many papers and in the media. The need for fire investigators to avoid bias when looking at smoke alarms, as well as other code-related issues, has for the most part been overlooked. Yet, the same logic and guidance should be applied to both aspects of a fire investigation. In many cases, the local fire chief or fire marshal wants to use the tragedy as an opportunity to reinforce a fire safety message. This well-intentioned motive can blind the investigator to key facts. After all, how do you teach the public about the benefit of smoke alarms after a fire in which the alarms operated?

In the first example below, the fire official held up the nonfunctioning smoke alarm, which survived the fire in the adjacent side of the duplex, and reported that disabled alarms were responsible for the tragedy. Despite the fact that no evidence existed to definitely determine the smoke alarm status in the side of the duplex where the fire started, it was assumed that the detector must have been nonoperational because that occupant died. The following excerpts are from stories that appeared in local papers after this fire.

10/05/04: “Blaze, reported at 1:33 a.m., kills 5 in duplex. A smoke detector without a battery was found in the unoccupied side of the duplex. No smoke detector was found in the charred side.”⁵
10/05/04: “Fire officials said there were no working smoke detectors in the fire apartment and fire may have raged for an hour. The officials are focusing on smoking. At some point, the parents awoke and tried to rescue the children.”⁶
10/05/04: “Fire officials said the lack of functioning smoke detectors was a key reason a young family of five perished in an early-morning house fire in Dennis yesterday.”⁷

The truth is that there was so much damage in the side of the duplex in which the fire started that it was impossible to know the status of the smoke alarm. Officials were basing this conclusion on the fact that a disabled alarm was found in the vacant side of the duplex. Shortly after this announcement, the landlord produced an affidavit showing that the alarms had been working a few months earlier. In addition, relatives reported hearing them operate as a result of smoke caused by cooking at a party weeks before the fire.⁸

Given that the smoke alarm produced a nuisance alarm related to cooking, it was likely an ionization smoke alarm. Given that the fire officials suspected smoking as a cause of the fire and hypothesized that the fire burned for an hour before being detected, it was likely a smoldering fire. In addition, since the parents were found after attempting to rescue their children, they were evidently alerted to the fire at some point. The investigators never considered the hypothesis that the smoke alarm may have operated too late to allow for safe egress. The investigators just assumed that the alarms didn’t work because the occupants died. We believe this happens hundreds of times every year in the United States.

Here is another example:

“The fire chief said a fire inspection in September noted three fire detectors in the house, and he could not explain why there was no evidence of them after the fire. He said that firefighters did a great job, but smoke detectors could have made a difference.”⁹

What is interesting is that the same news report noted that one of the survivors “told relatives that he heard a smoke alarm go off.” Why did the fire department ignore this statement? We contacted the local fire department and cautioned its members not to make assumptions. They were encouraged to keep an open mind and make a determined effort to find the smoke alarms. The next day, the fire department issued a new statement in which it was reported: “Two activated smoke detectors were present in the rowhouse where four people died in a fire Wednesday night.”¹⁰

In the example below, the local fire officials used the fact that occupants died with a working alarm to determine that the fire was caused by arson.

” ‘A working smoke detector that failed to rouse nine people killed in a pre-dawn house fire indicated the blaze moved more quickly than a normal fire and provided a critical clue in deciding the blaze was arson,’ authorities said Wednesday …. We had young, able-bodied people who we believe had a smoke detector warning and weren’t able to evacuate. I think that got our attention the most …. The initial investigation ‘could not find something that would lead us to a cause other than accidental. So, on an initial basis, it looked as if we had an accidental situation.’ “¹¹

The chief is assuming that most fires in which occupants die with working alarms are arson fires. Yet, according to the NFPA,¹² approximately 1,020 residential fire victims in the United States die with operating alarms, and approximately 330 die in arson fires.¹³ This means the vast majority of fire victims who die in fires in which the alarms operate die in nonarson fires. So, if the fire may well have been an arson fire, the fact that the smoke alarms operated is irrelevant to the issue.

Clearly, in these cases and most likely in many others, the investigators had preconceptions regarding the smoke alarms and then fit their hypotheses to those preconceptions. This type of analysis is not consistent with NFPA 921. (2) In these examples, the investigators appear to show expectation and confirmation bias. We do not believe these are isolated cases but the norm for most fatal fire investigations because fire officials are looking for an opportunity to educate the public with the message: “Smoke alarms work and save lives.” We have even seen this message after fatal fires in which the smoke alarm worked and someone still died.

Bias in the Analysis of the Available Data

Biases similar to those described above appear to affect also the analysis of the available data. The magnitude of the problem that this type of bias overlooks could be substantial. According to the U.S. Fire Administration (USFA), “When the ‘unknowns’ … are apportioned to the other three categories, alarms were not present in 52 percent of the fatalities in 1998; an additional 19 percent of the deaths occurred in homes where smoke alarms were present but failed to operate. In 29 percent of fire deaths, an alarm did operate-eight percentage points higher than in 1996. This is somewhat disturbing since there is a widespread belief that an operating alarm will save lives. In some of these cases, the alarm may have gone off too late to help the victim, the victim may have been too inebriated or feeble to react, or the fire may have been too close to the victim. Such cases merit further study.”¹⁴ Table 1 illustrates this disturbing trend throughout the 1990s.

There will always be a certain percentage of people who cannot be saved by smoke detectors-e. g., the handicapped, those intimate with the fire, and so on. There is no reason to believe that the number of those people quadrupled between 1988 and 2001.¹⁵ In addition, although the number of fires with working detectors increased approximately in proportion to the increase in the number of detectors installed, the increase in the percentage of fatal fires with working detectors has far exceeded it. Also, over this time period, because of improvements in codes, more smoke alarms were installed in each residence and more of them were hard-wired.¹⁶ The effectiveness of smoke alarms should have increased, but it appears to have decreased.

The failure of the USFA to study this problem is troubling. If it had studied it, it might have noticed that the increase occurred after Underwriters Laboratories (UL) and the smoke alarm manufacturers who sit on the UL217 Standards Technical Panel modified the UL217 Fire Tests in 1988 in an attempt to address the nuisance alarm problem. These modifications to the UL217 Fire Tests made it much easier to pass the smoldering fire test.¹⁷ This allowed less sensitive ionization smoke alarms to be sold. As a consequence, smoke alarms that were already relatively insensitive to the type of smoke that contributed to hundreds of deaths per year-smoldering smoke-were made even less sensitive. As the percentage of these desensitized alarms in U.S. homes increased, the percentage of fatal fires with operating alarms increased.

Reasons for Bias Relating to Smoke Alarms Chiefs Are Not Informed When Someone Dies with Working Alarms

The following statement appeared in a 2006 newspaper article: “There are ample statistics showing that fire detectors work, as well as plenty of empirical evidence. Fire officials across Central Massachusetts said they could think of no instances in which death occurred in a residence with a working smoke detector but could think of many instances in which a detector saved a life.”¹⁸

What makes this statement so interesting is that in the year in which the story appeared (2006), fires in Massachusetts in which the alarm status was determined revealed that 60 percent of the fatal fires were found to have operating alarms.¹⁹ If every time someone dies with operating alarms the media are not informed of that fact or, in many cases, are misled to believe that the alarm did not operate, the general public and other fire officials will never be made aware that it can happen. Then, when it happens in their jurisdictions, they assume that it is such a rare phenomenon that they ignore key evidence or find alternative explanations that blame the victim.

This is like a misinformation feedback loop that hides an inconvenient truth. Even on a nationwide basis, the percentage of fatal fires with operating alarms is almost 40 percent. (12) Yet invariably, when we ask fire officials to estimate this percentage, they typically guess anywhere from five to 10 percent. To compound this “public relations error,” the fire service is told that scientific studies show that smoke alarms provide adequate warning.

Mistaken Understanding of Results of Fire Tests

Fire officials and the general public continually hear good news about smoke alarms. “Smoke alarms of either the ionization type or the photoelectric type consistently provided time for occupants to escape from most residential fires …. Consistent with prior findings, ionization type alarms provided somewhat better response to flaming fires than photoelectric alarms, and photoelectric alarms provided (often) considerably faster response to smoldering fires than ionization type alarms …. Smoke alarms of either type installed on every level generally provided positive escape times for different fire types and locations.”²⁰

This statement is based on language in the Executive Summary of the National Institute of Standards and Technology (NIST) Home Smoke Alarm Study. (16) Although it is technically accurate, a more detailed analysis of the available data paints a very different picture. Although Table 3 does not appear in the report, it was provided in answers sent to the Boston Fire Department (BFD) in response to a series of questions about the report.²¹

As you can see, there will be many cases in which the ionization alarms will be providing little or no available safe egress time (ASET). In addition, since NIST inexplicably ignored the tenability along the egress path for living room fires, the results are actually much worse than indicated. NIST did admit in another document that “ionization detectors have been shown to sometimes fail to alarm in a smoldering fire even when visibility in the room is significantly degraded by smoke.”²² This result should not have surprised anyone. Fire tests conducted in the United States (1979), Great Britain (1997), and Norway (1991) all showed similar results.²³ It is true that in the fast flaming tests, the ionization alarm was quicker, but in those tests the photoelectric alarm was still providing on average two minutes ASET.

The well-documented ineffectiveness of ionization technology to detect smoke from smoldering fire may also be a factor in other scenarios in which the smoke has similar attributes to typical smoldering smoke-relatively larger particles, for example, as in the following situations:

Oxidative pyrolysis (nonflaming combustion)-a plastic item melted on a stovetop.²⁴
An electrical fire in the walls or attic space.²⁵
Aged (cold) smoke-Even smoke from a flaming fire may change as the particles agglomerate/coagulate.²⁶

The excerpt below, from a 1983 article in the NFPA Fire Journal reporting on a hotel fire, discusses “cold smoke.” The smoke entered the hotel rooms; it traveled from several floors away.

“The guest-room smoke detectors were the single-station, battery-operated ionization type. Many hotel occupants reported that the single-station, battery-powered detectors did not sound, even though smoke conditions were obvious by sight and smell in their rooms. However, laboratory examination of a sample of detectors by the Center for Fire Research, National Bureau of Standards, indicated there was no malfunction of the individual guest-room detectors tested. Cold Smoke Effect. The essential feature of smoke is its instability. As smoke travels away from a fire, and ages, the smoke particles in a cloud collide with one another and cluster. This process goes on continuously until the number of particles has been considerably diminished and the average size largely increased. Since the response of an ionization smoke detector is dependent on the particle concentration and size, some of the guest rooms might not have had a sufficient concentration of this aged smoke to operate the smoke detectors in the guest rooms.” (26)

Reinforcing the lack of awareness of these testing results is that the typical experience that one has with operating smoke alarms is when it responds to cooking aerosols, giving the impression that the alarm in question is not only working but is super sensitive. What most investigators do not understand is that smoke can vary in key characteristics. The same ionization alarm that is sensitive to cooking aerosol (e.g., 0.1-0.2 microns) is relatively insensitive to smoldering smoke (e.g., 1-2 microns) (Figure 1).

In Figure 1, A represents a photoelectric detector using a “scattered light principle,” a spot detector; B represents a photoelectric detector using “obscuration,” a beam detector; and C represents an ionization detector, a spot detector. Note that this chart assumes that the total mass of particulate stays constant for a given volume. This causes the number of particles to decrease as the size increases. It is actually the decrease in the number of particles that causes the ionization detector to become less sensitive to large particle smoke. Figure 1 helps explain why, in many cases, an alarm that does not respond effectively during a fatal fire that started in the smoldering (large particles) mode was reported to have operated just a few days or weeks earlier as a result of cooking (small particles).

Misunderstanding of UL217 Fire Tests (17)

It is interesting to note that sometimes smoke alarms checked to see if they are functioning properly are retested in the UL “Smoke Box.” This is based on the assumption that the percent of obscuration on the back of the alarms, typically one to two percent obscuration per foot (obs/ft.), provides an indication of how it will respond in a real fire. In actuality, this is obtained in a small box, with nonrealistic smoke, that is designed to be a “calibration” tool. It is not an indication of the level of smoke at which an alarm will respond in a real fire. It is merely testing an alarm to verify that it is within the same calibration levels that it was when it left the factory.

In the recent NIST tests, ionization smoke alarms rated at one to two percent in the UL 217 Smoke Box would respond at levels of obscuration as high as 22 percent obs/ft. in the smoldering scenarios.²⁷ This occurred despite the fact that in the single UL smoldering fire test, which uses white pine, the passing criteria is 10 percent. Real fires involving synthetic smoke produce the kind of smoke to which ionization alarms are less sensitive. These results, from the recent NIST tests, are consistent with smoke alarm studies for the past 30 years.

A new smoldering test that uses synthetic material commonly found in homes has been proposed, but the committee voted it down twice. Even if it is eventually approved, it will not be in force for several years, so the problem created by the lack of this “large particle” test, which has existed for more than 30 years, will continue for many years to come.

NFIRS Data Collection on Smoke Alarms

The statistics the NFPA uses to highlight the effectiveness of smoke alarms are based on data collected through the National Fire Incident Reporting System (NFIRS). Figure 2 shows four of the “blocks” relating to smoke alarms.

L1. What is meant by “in the area of the fire”? If the smoke alarm on the first floor, near the kitchen and living room, is disabled but the smoke alarm on the second floor operates and alerts the victims, how should the investigators answer this question? Alternatively, what if the smoke alarm on the second floor operates but not soon enough to provide adequate warning?
L1. As discussed earlier, many investigators do not appear to make the effort to find the smoke alarms because of flawed assumptions and the difficulty of searching through debris. But even in the cases in which the remnants of a smoke alarm are found, there is no means to collect these data.
L4. What if the smoke alarm in the apartment was disabled but the common area alarm system operated and saved the lives of several occupants?
L5. Since the hypotheses developed by the investigator are susceptible to “expectation bias,” how reliable is the conclusion of the investigator collected by L5?
L6. How is the investigator supposed to determine that the smoke alarm was “dirty,” “defective,” or “lacked maintenance”?

Three minor changes to the NFIRS collection system could improve the data collected on smoke alarms:

There should be two boxes labeled “L1-A” (for alarms within the unit/floor of origin) and “L1-B” (for alarms outside the unit/floor of origin). For each of these boxes, there should be supplemental boxes that ask the investigator to identify whether the alarm was ionization, photoelectric, dual, or undetermined.
In the box labeled “L4,” there should be a fifth category titled “Properly powered-operation undetermined.”
Since the main reason for disabling smoke alarms is repeated nuisance alarms, there should be a box to supplement “L6” that asks the investigator to estimate the distance to a cooking appliance and the bathroom.

To help find the detector, it may be helpful to look at adjacent apartments or townhouses. If they are constructed at the same time or if they have the same landlord, there is a possibility that the location of the detectors in the adjacent living unit can provide clues to the location in the unit of fire origin. In the absence of clues, it should be assumed that they are located where the local codes require them. More than once, the detector and the battery were found in the debris lying on the floor right under the spot on the ceiling where the investigator assumed the detector was located. Even though the plastic had melted, the metal parts of the detector were still recognizable.

There were also times when responders were told there were no smoke alarms but the responders were able to find them. This same phenomenon occurs when public investigators leave the scene once the cause is determined to be accidental and an insurance investigator later finds the smoke alarms.

Common Excuses to Explain Why the Alarm Was Not Effective

When a properly powered smoke alarm is discovered while investigating a fire in which an occupant died, a fire chief/investigator will often hypothesize explanations for which there is very little evidence. Although the investigator should take the following factors into account, they are not in and of themselves sufficient explanations.

“The smoke alarm was too old.” It is true that smoke alarms do not last forever, but most will still be operational after 10 years. A former industry executive has stated, “In theory, the electronic components in a smoke detector should last at least 30 years. But a smoke detector could fail at any time.” The Pittway Corp., one of the largest smoke alarm producers, recommends changing the detectors every 10 years because that provides a reasonable margin of safety. But, the industry executive adds, “I don’t know of any unit that has failed because of aging.”²⁸
“The children did not hear the smoke alarms.” Although there are many studies showing that some children will not awake to an alarm, why should that matter if their parents are with them? In addition, in many fires, it is the children who alert the parents.
“The occupant was impaired.” This is often mentioned after a fire with young adults as victims. Although it is true that being impaired can affect the ability of an occupant to respond effectively to an alarm, we have seen this explanation mentioned even in cases where the victim was found just a few feet from the door. In these cases, it is obvious that the victim responded and just needed a few more seconds to successfully escape. As a consequence, the extra 30 minutes or so provided by a photoelectric smoke alarm would provide more than enough time. In addition, the investigator should take into account the amount of carbon monoxide (CO) inhaled by the victims prior to the smoke alarm’s operating. Researchers have noted, “Both theory and experiment suggest that critical conditions threatening life safety (based on a 4.5 percent minimum dose criterion) tend to arise in 50 to 100 minutes after ignition.²⁹ When you compare this time frame to the time when furniture typically smolders-22 minutes up to 306 minutes with a mean of 88 minutes-it becomes apparent that the amount of CO potentially inhaled during the smoldering stage, which would cause confusion, is an important factor to consider.³⁰
“The smoke did not reach the detector.” A good example of this mistake is in a report issued by the USFA,31 which states: “Here is a case where a smoke detector was installed by the owner of a rental property, the rental occupant received smoke detector maintenance information, and the occupant practiced proper smoke detector maintenance. It is almost certain that the smoke detector was functional at the time of the fire. The fire department investigators were told by the occupants that the detector had alarmed previously when a towel had caught fire on the stove. In addition, approximately two weeks before the fire, the wife placed a new battery in the detector and tested the detector with a smoldering paper. The detector was found operational in this test. The failure here was that the detector was not installed in a recommended location.”

The problem with that theory is that a witness within a few feet of the front door describes heavy smoke on the first floor, including in the area of the smoke alarm, and did not hear the alarm sound. The investigators did not understand that a smoke alarm that responded effectively to a “flaming towel,” which would have produced small particles, might not effectively respond to a “glowing” (nonflaming) large-particle fire. Since the investigators were sure that any smoke alarm would have sounded if the smoke reached it, they ignored this eyewitness testimony and assumed that the smoke did not reach the alarm.

Potential Impact of Investigative Bias on Fire Statistics

Perhaps the most popular statement regarding smoke alarms is, “A working smoke alarm reduces your chances of dying in a fire by 50 percent.” Although just about every fire official and fire safety site highlights the fact that smoke alarms reduce your risk by 50 percent, few understand on what this statement is based. The source of this oft-cited claim is logic similar to the following: “The death rate per 100 reported home fires was more than twice as high in homes that did not have any working smoke alarms (1.18 deaths per 100 fires) either because no smoke alarm was present or an alarm was present but did not operate), as it was in homes with working smoke alarms (0.53 per 100 fires).” (12)

An in-depth analysis of these numbers tells a much more interesting story (Table 4):

The category “Non-confined Fires” was added in 2000 to the NFIRS reporting system. Prior to this change, these fires were typically categorized as “food on the stove.”³² According to the NFPA, there were no fatalities in this category, so they do not represent dangerous fires in which the alarm effectiveness is critical. So, a second row was created that included only “non-confined” fires.
If fire officials are occasionally miscoding fires as having nonoperational alarms when, in fact, they operated, this category would be expected to be artificially high. This might explain why the risk with nonoperational alarms is 2½ times higher than if there were no alarms. The risk should be the same. The third grouping estimates the number of miscodings by assuming the risk with nonoperating alarms is equal to the risk with no alarms and shifts the “extra” fatalities into the operating column. If the risk with nonoperational alarms was 1.24, then the total number of fatalities would be 242 (1.24*195). As a consequence, 348 (590 – 242) fatalities are shifted into the “Present and Operated” column.

It would appear that the benefit of operating smoke alarms is less than advertised and that part of that benefit may just be because of miscoding and statistical “interpretations.” When looking at just apartments, the numbers are even worse.

When looking at nonconfined fires, there is a 34 percent increase in risk with operating alarms. This may be because of fire officials coding the common area alarm as operating as opposed to the smoke alarm in the apartment in which the fire originated. NFIRS does not allow the official making the report to distinguish this important factor. The risk may also be reduced in apartments because occupants have fewer egress options than in one- or two-family homes. It appears that the reported benefit of smoke alarms is much less than commonly believed.

Experience in Boston and Massachusetts

Starting in the early 1990s, the fire marshal’s office of the BFD started to investigate fatal and nonfatal fires for the cause of the death, injury, or property loss in addition to the cause and origin of the fire. The first pattern that was noticed was the high number of people who were killed or injured in fires with disabled smoke alarms. In researching solutions to this problem, the BFD became aware of the difference in smoke alarm technology and the differences in various aerosols.

As a consequence, the BFD made a proposal to have the State Building Code mandate only photoelectric alarms within 20 feet of a kitchen or bath. The members of the board responsible for the code asked the BFD to make sure that photoelectric alarms, which obviously are less sensitive to nuisance smoke, were not also less sensitive to real fire smoke. While conducting this research, the BFD did not find any testing indicating a problem with the response of photoelectric smoke alarms; however, the BFD did become aware of the many studies that identified the ionization as having a response problem with smoldering smoke. As a consequence, in 1998, the Massachusetts State Building Code mandated the use of photoelectric technology in new or renovated construction.³³

Although no information exists regarding the current total percentage of homes in Massachusetts with photoelectric smoke alarms, it is not unreasonable to assume that more than 15 years after this code language was adopted, it is substantial enough to affect the fire statistics.

<b>(1-2)</b> The image on the right shows the remains of the melted alarm mounted on the wall (left image). Even though only some of the parts of a smoke alarm survive the fire, the parts that remain are often enough to identify the type of alarm. (Photo by Joseph Fleming.)

(1-2) The image on the right shows the remains of the melted alarm mounted on the wall (left image). Even though only some of the parts of a smoke alarm survive the fire, the parts that remain are often enough to identify the type of alarm. (Photo by Joseph Fleming.)

Looking at the data in Table 5, it would appear that the benefit was substantial. If Massachusetts had a reduction of 25 percent, like the rest of the United States, from the late 1990s to today, the death rate would be 7.0 instead of 4.4. If it was 7.0/million there would be approximately 17 (6.646 million * (7.0-4.4) extra deaths per year in Massachusetts. Alternatively, if the fire death rate in the United States dropped as fast as it did in Massachusetts, then the total death rate would be more than 1,000 less than currently estimated.

Boston has seen an even more drastic reduction. From 2009 to 2012, Boston had only four fatalities, despite being a large urban northeast city with older construction and a high population density. The Massachusetts State Fire Marshal’s Office, following the lead of the BFD, started a groundbreaking program to collect more detailed information on smoke alarms starting in 2011. A preliminary analysis of the data conducted by the BFD indicates that for the years 2011-2013 (in cases in which the smoke alarm information could be collected), 42 fatal fires occurred with ionization alarms and, in many of those cases, the fire was most likely smoldering or the alarms were disabled. In the five cases in which the alarm was photoelectric, most of the victims appeared to have medical or other issues that affected escape potential. Although the analysis is still preliminary, the data are encouraging.

The investigations of fires for potential code improvements, particularly as they relate to smoke alarms, have to be given as much emphasis as the investigation of fires for cause and origin. Collecting better data on the smoke alarm, particularly the type (ionization, photoelectric, or multicriteria) is not a panacea for the fire problem in the United States, but we strongly believe that it will help guide code officials to cost/benefit code improvements that could save hundreds of lives per year. More information is at http://www.interfire.org/features/smokedetector.asp.

REFERENCES

Brainyquote.com, “Mark Twain Quotes”, Retrieved 7/1/2014.
National Fire Protection Association, NFPA 921, Guide to Fire and Explosion Investigation, 2014 Edition.
Fleming, J., Smoke Detectors and the Investigation of Fatal Fires, Fire and Arson Investigator, May 2000.
Babrauskas, V., Some Neglected Areas in Fire Safety Engineering, Fire Science & Technology (Tokyo) 32:1, 35-48 (2013).
“Blaze in Duplex Kills Family of 5,” Boston Globe, Boston, MA, October 5, 2004.
“Smoking Eyed as Cause of Blaze that Killed Five,” Boston Herald, Boston, MA, October 5, 2004.
“Lack of Working Smoke Detectors Blamed in Deaths,” Cape Code Times, Hyannis, MA, October 5, 2004.
“Witness: Detector Worked at Dennis,” Cape Cod Times, Hyannis, MA, October 6, 2004, 2:00 AM.
“Family Killed in a House Fire,” Baltimore Sun, Baltimore MD January 22, 2010.
“Activated Smoke Detectors at Fatal Fire Site,” Baltimore Sun, Baltimore MD January 23, 2010.
“Smoke Detector Offered Clue in Ohio Fire,” AP (Associated Press) Online, June 2, 2005.
National Fire Protection Association, Smoke Alarms in US Homes, March 2014.
National Fire Protection Association, Intentional Fires, April 2014.
National Fire Data Center, “Fire in the United States, 1989-1998 (12th Edition), U.S. Fire Administration, Emmitsburg, Md., August 2001.
United States Fire Administration, Investigation of Fatal Residential Structure Fires with Operational Smoke Alarms, August 2006.
Peacock, R., Averill, J., Bukowski, R., Reneke, P., Cleary, T., Bryner, N., Walton, W., Kuligowski, E., “Performance of Home Smoke Alarms,” NIST Tech. Note 1455, 2003.
Underwriters Laboratories, Standard for Single and Multiple Station Smoke Detectors, UL 217.
“A shriek to heed as if your life depended on it”, Worcester Telegram, Worcester, MA, April 23, 2006.
“2006 Report of the Massachusetts Fire Incident Reporting System”, Department of Fire Service, MA.
Kidde.com, “Ionization & Photoelectric Sensing Technologies in Smoke Alarms”, Retrieved 8/4/2014
National Institute of Standards and Technology, “Questions and Answers Clarifying Findings of NIST Home Smoke Alarm Study,” March 2007.
National Institute of Standards and Technology, “Statement for the Record,” August 6, 2007.
Fleming, J.M., “Photoelectric v. Ionization Detectors-a Review of the Literature Revisited,” National Fire Protection Research Foundation Symposium, Orlando, Fla., Jan. 2005.
Characterisation of Fire Generated Particles, Per Blomqvist et al, SU BRANDFORSK Project 700-061, Fire Technology, SP Report 2010:01
FirstAlertStore.com, “Smoke Alarms FAQ”, retrieved 8/4/2014.
“Twelve Die in Fire at Westchase Hilton Hotel, NFPA Fire Journal, Jan. 1983.
“Validation of a Smoke Detection Performance Prediction Methodology-Volume 3. Evaluation of Smoke Detector Performance,” National Fire Protection Research Foundation, October 2008.
Meier, B., “How Long Do Smoke Detectors Last?” New York Times, September 29, 1990.
Quintiere J., Birky, M., McDonald F., Smith G., An Analysis of Smoldering fires in Closed Compartments and their Hazard Due to Carbon Monoxide, NBSIR 82 – 2556, U.S. Dept. of Commerce, July 1982.
Babrauskas, V., Krasny, F., “Upholstered Furniture Transition from Smoldering to Flaming,” Journal of Forensic Science, 1997; 42(6): 1029-1031.
Shapiro, J., Carpenter, D., Schlaenman, P., Stambaugh, H., Four House Fires That Killed 28 Children, U.S. Fire Administration, USFA-TR-020/December 1987.
National Fire Incident Reporting System, Complete Reference Guide, Jan. 2012. U.S. Fire Administration, National Fire Data Center.
Mass. State Building Code, 6th Edition, (1998).

JOSEPH M. FLEMING has been a member of the Boston (MA) Fire Department for more than 35 years. He has held the rank of deputy chief for more than 20 years. He was fire marshal for eight years. He is also a Professional Grade Member of the Society of Fire Protection Engineers. He has participated in task groups and testified before government committees across the Unites States regarding smoke alarm issues. Several states have based their smoke alarm regulations and recommendations on his research. He has been involved in hundreds of investigations, including a few involving firefighter fatalities, and in litigation involving fire investigations across the United States.

VYTENIS BABRAUSKAS was a long-time researcher at the National Institute of Standards and Technology, where he developed the two primary tools for measuring the heat release rate of fires-the cone calorimeter and the large-scale furniture calorimeter. He was awarded the first-ever Ph.D. degree in fire protection engineering from the University of California, Berkeley, and also has degrees in physics and structural engineering. He is the author of more than 300 articles and reports and three reference books-Heat Release in Fires, Fire Behavior of Upholstered Furniture and Mattresses, and The Ignition Handbook, which is the standard science reference work for fire investigators. He is working on a new book on electrical fires and explosions. Since 1993, his firm Fire Science and Technology Inc. has specialized in lending fire science support to fire investigations and litigations in addition to doing contract research for manufacturers, research institutes, and governmental bodies.

New IAFC Toolkit Helps Educate About Smoke Alarm Placement
NFPA: New Report Focuses on Smoke Alarm Effectiveness
Sounding the Alarm on Smoke Alarms