Skip to content
From Fire Safety Academy Wiki
Lisa (talk | contribs)
New FSE wiki article: Fire death rate (written by Claude Opus 5.5)
 
Linda (talk | contribs)
Add SVG figures fse-fire-death-rate-v1.svg, fse-fire-death-rate-v2.svg, fse-fire-death-rate-v3.svg, fse-fire-death-rate-v4.svg, fse-fire-death-rate-v5.svg (list-8 visual plan)
 
(One intermediate revision by one other user not shown)
Line 1: Line 1:
The '''fire death rate''' is the number of people killed by fire in a given period, usually a year, divided by the size of the population concerned. It is commonly expressed as deaths per million population.<ref name="AFK-3-79-80">Ahrens et al. (2008), pp. 3-79–3-80.</ref> Unlike a simple count of deaths, the rate takes account of changes in population, so it is used as a measure of fire risk and for comparisons over time and between groups.<ref name="AFK-3-79-80" /><ref name="AFK-3-80">Ahrens et al. (2008), p. 3-80.</ref> In 2004 the United States had 13 civilian fire deaths per million population, from an estimated 3,900 deaths and a resident population of 293,655,000.<ref name="AFK-3-80" />
The '''fire death rate''' is the number of people killed by fire in a given period, usually a year, divided by the size of the population concerned. It is commonly expressed as deaths per million population.<ref name="AFK-3-79-80">Ahrens et al. (2008), pp. 3-79–3-80.</ref> Unlike a simple count of deaths, the rate takes account of changes in population, so it is used as a measure of fire risk and for comparisons over time and between groups.<ref name="AFK-3-79-80" /><ref name="AFK-3-80">Ahrens et al. (2008), p. 3-80.</ref> In 2004 the United States had 13 civilian fire deaths per million population, from an estimated 3,900 deaths and a resident population of 293,655,000.<ref name="AFK-3-80" />


Some statistics, such as the U.S. death certificate series, are given per 100,000 population instead.<ref name="HC-3-6">Hall and Cote (2008), p. 3-6.</ref> Related measures divide deaths by the number of fires or households rather than by people, or by hours of exposure where such data exist.<ref name="AFK-3-80" /><ref name="AFK-3-68">Ahrens et al. (2008), p. 3-68.</ref><ref name="H-5-106">Hall (2008), p. 5-106.</ref> U.S. fire death rates fell steeply during the twentieth century. Yet in averages for 1993–2002 compiled by the International Technical Committee for the Prevention and Extinction of Fire (CTIF), they were still higher than in many European countries.<ref name="HC-3-8">Hall and Cote (2008), p. 3-8.</ref><ref name="AS-3-57">Ahrens and Stewart (2008), p. 3-57.</ref> Because fire deaths are rare, rates for small communities or single years vary strongly by chance.<ref name="AFK-3-87">Ahrens et al. (2008), p. 3-87.</ref> The values below come mainly from the 2008 NFPA handbook and describe the United States between 1977 and 2004 unless stated otherwise.
Some statistics, such as the U.S. death certificate series, are given per 100,000 population instead.<ref name="HC-3-6">Hall and Cote (2008), p. 3-6.</ref> Related measures divide deaths by the number of fires or households rather than by people, or by hours of exposure where such data exist.<ref name="AFK-3-80" /><ref name="AFK-3-68">Ahrens et al. (2008), p. 3-68.</ref><ref name="H-5-106">Hall (2008), p. 5-106.</ref> U.S. fire death rates fell steeply during the twentieth century.<ref name="HC-3-6" /> Yet in averages for 1993–2002 compiled by the International Technical Committee for the Prevention and Extinction of Fire (CTIF), they were still higher than in many European countries.<ref name="HC-3-8">Hall and Cote (2008), p. 3-8.</ref> Because fire deaths are rare, rates for small communities or single years vary strongly by chance.<ref name="AFK-3-87">Ahrens et al. (2008), p. 3-87.</ref> The values below come mainly from the 2008 NFPA handbook and describe the United States between 1977 and 2004 unless stated otherwise.


== Definition and calculation ==
== Definition and calculation ==


In [[fire statistics]], Ahrens et al. (2008) define a rate as a ratio of a measure of the size of a fire problem to a measure of the size of the group affected by it.<ref name="AFK-3-79">Ahrens et al. (2008), p. 3-79.</ref> For the fire death rate, the problem is the number of fire deaths in a period and the group is the population. The result is multiplied by a scale factor so that it can be read as a convenient whole number:<ref name="AFK-3-80" />
In [[fire statistics]], Ahrens et al. (2008) define a rate as a ratio of a measure of the size of a fire problem to a measure of the size of the group affected by it.<ref name="AFK-3-79">Ahrens et al. (2008), p. 3-79.</ref> For an annual fire death rate, the problem is the number of fire deaths in one year and the group is the population. The result is multiplied by a scale factor so that it can be read as a convenient whole number:<ref name="AFK-3-80" />


<math display="block">R = \frac{D}{P}\, k</math>
<math display="block">R = \frac{D}{P}\, k</math>
Line 14: Line 14:
| <math>R</math> || Fire death rate || deaths per million (or per 100,000) population per year
| <math>R</math> || Fire death rate || deaths per million (or per 100,000) population per year
|-
|-
| <math>D</math> || Number of fire deaths in the period, for example civilian deaths in one year || deaths per year
| <math>D</math> || Number of fire deaths in one year, for example civilian deaths || deaths per year
|-
|-
| <math>P</math> || Population of the area or group, for example the resident population || persons
| <math>P</math> || Population of the area or group, for example the resident population || persons
Line 22: Line 22:


As a worked example, the 2004 U.S. estimates give <math>R = 3{,}900 / 293{,}655{,}000 \times 10^6 = 13.3</math> deaths per million population, which Ahrens et al. (2008) round to 13.<ref name="AFK-3-80" /> The same value is 1.33 per 100,000. Converting between the two scales only requires a factor of 10. A death certificate rate of 1.5 per 100,000 for 2003 therefore corresponds to 15 per million.<ref name="HC-3-6" />
As a worked example, the 2004 U.S. estimates give <math>R = 3{,}900 / 293{,}655{,}000 \times 10^6 = 13.3</math> deaths per million population, which Ahrens et al. (2008) round to 13.<ref name="AFK-3-80" /> The same value is 1.33 per 100,000. Converting between the two scales only requires a factor of 10. A death certificate rate of 1.5 per 100,000 for 2003 therefore corresponds to 15 per million.<ref name="HC-3-6" />
[[File:fse-fire-death-rate-v1.svg|thumb|upright=1.5|alt=Diagram of a fire death rate as deaths divided by population times one million, with the 2004 U.S. example giving 13 per million|A fire death rate divides the number of fire deaths by the population and scales the result; the 2004 U.S. values give about 13 deaths per million.]]


The numerator depends on how a fire death is defined. Deaths are generally discussed separately for civilians and for the fire service, and civilians include everyone who is not a member of the fire service.<ref name="AS-3-49">Ahrens and Stewart (2008), p. 3-49.</ref> In the [[National Fire Incident Reporting System]] a fire death is anyone who dies as a result of the fire within one year of it. [[Smoke inhalation]] and burns are the most common causes of death, but a person fatally injured while jumping from a burning building also counts.<ref name="AS-3-49" /> Purser (2002) notes that toxic smoke products are recognised as the major cause of death in fires in the United States and the United Kingdom.<ref name="P-2-84">Purser (2002), p. 2-84.</ref> Van Rossum (2027) describes people dying because they are trapped by fire or smoke, overcome in their sleep or poisoned by inhaling toxic smoke gases, and notes that most fire deaths occur at home.<ref name="VR-7">Van Rossum (2027), §7.</ref>
The numerator depends on how a fire death is defined. Deaths are generally discussed separately for civilians and for the fire service, and civilians include everyone who is not a member of the fire service.<ref name="AS-3-49">Ahrens and Stewart (2008), p. 3-49.</ref> In the [[National Fire Incident Reporting System]] a fire death is anyone who dies as a result of the fire within one year of it. [[Smoke inhalation]] and burns are the most common causes of death, but a person fatally injured while jumping from a burning building also counts.<ref name="AS-3-49" /> Purser (2002) notes that toxic smoke products are recognised as the major cause of death in fires in the United States and the United Kingdom.<ref name="P-2-84">Purser (2002), p. 2-84.</ref> Van Rossum (2027) describes people dying because they are trapped by fire or smoke, overcome in their sleep or poisoned by inhaling toxic smoke gases, and notes that most fire deaths occur at home.<ref name="VR-7">Van Rossum (2027), §7.</ref>


National numerators come from two kinds of source. NFPA's estimates are projections from the [[NFPA fire experience survey]] and include deaths in arson, self-immolation and vehicle fires.<ref name="AFK-3-90">Ahrens et al. (2008), p. 3-90.</ref> Death certificate data may classify such deaths as homicide, suicide or vehicle accidents. Different organisations also use different combinations of the disease classification codes to identify fire deaths.<ref name="AFK-3-90" /><ref name="AS-3-49" /> Rates from different sources are therefore not identical, although their trends and patterns are consistent.<ref name="AFK-3-90" />
The comparisons here use two kinds of national numerator. NFPA's estimates are sample-based projections from the [[NFPA fire experience survey]] and include deaths from arson, self-immolation and vehicle fires.<ref name="AFK-3-91">Ahrens et al. (2008), p. 3-91.</ref> The National Safety Council's death-certificate series excludes some of these deaths, which may instead be classified as homicide, suicide or vehicle accidents.<ref name="AFK-3-91" /> Different organisations also use different combinations of disease classification codes to identify fire deaths.<ref name="AS-3-49" /> Rates from different sources are therefore not identical, although their trends and patterns are consistent.<ref name="AFK-3-91" />


== Rates and counts ==
== Rates and counts ==
Line 33: Line 34:
Over long periods the difference can be large. National Safety Council analysis of death certificates showed U.S. unintentional fire deaths in 2004 to be roughly 56 percent lower than in 1913. The 2004 rate, however, was 86 percent lower than the 1913 rate.<ref name="AFK-3-83">Ahrens et al. (2008), p. 3-83.</ref> Because of definitional changes, these percentages should be considered rough estimates. They nevertheless show how different progress can look when measured by numbers or by rates.<ref name="AFK-3-83" />
Over long periods the difference can be large. National Safety Council analysis of death certificates showed U.S. unintentional fire deaths in 2004 to be roughly 56 percent lower than in 1913. The 2004 rate, however, was 86 percent lower than the 1913 rate.<ref name="AFK-3-83">Ahrens et al. (2008), p. 3-83.</ref> Because of definitional changes, these percentages should be considered rough estimates. They nevertheless show how different progress can look when measured by numbers or by rates.<ref name="AFK-3-83" />


The choice of denominator can also change the conclusion. A fall in fire deaths in manufactured homes might mean that these homes became safer, or only that fewer people lived in them. Deaths per million people living in manufactured homes would separate the two explanations.<ref name="AFK-3-83" /> Adding a state to a database in the middle of a trend period can still bias such a rate if that state differs, for example in its share of manufactured homes.<ref name="AFK-3-83" />
The choice of denominator can also change the conclusion. A fall in fire deaths in manufactured homes might mean that these homes became safer, or only that fewer people lived in them. Deaths per million people living in manufactured homes would help distinguish the two explanations.<ref name="AFK-3-83" /> Adding a state to a database in the middle of a trend period can still bias such a rate if that state differs, for example in its share of manufactured homes.<ref name="AFK-3-83" />


Space heaters give a clear example of rates and counts pointing in opposite directions. For 1995–1999 fires and 1997 usage, portable liquid-fuelled heaters had 9.93–12.41 deaths per million households using them, against 5.95–7.11 for portable electric heaters.<ref name="H-5-106" /> In the unadjusted death counts for the same fires, however, portable electric heaters had 90 deaths and portable liquid-fuelled heaters 40. Sorting by death toll produced a ranking almost exactly opposite to the ranking by rates.<ref name="H-5-106" />
Space heaters give a clear example of rates and counts pointing in opposite directions. For 1995–1999 fires and 1997 usage, portable liquid-fuelled heaters had 9.93–12.41 deaths per million households using them, against 5.95–7.11 for portable electric heaters.<ref name="H-5-106" /> In the unadjusted death counts for the same fires, however, portable electric heaters had 90 deaths and portable liquid-fuelled heaters 40. Sorting by death toll produced a ranking almost exactly opposite to the ranking by rates.<ref name="H-5-106" />
Line 40: Line 41:


Long-term U.S. rates come from death certificates, which NFPA uses only for long-term trend analysis.<ref name="HC-3-4">Hall and Cote (2008), p. 3-4.</ref> On that basis the rate was 9.1 fire deaths per 100,000 population in 1913 and 9.9 in 1918. It fell to 3.1 in 1973 and was 1.5 in 2003.<ref name="HC-3-6" /> Classification changes in 1948, 1958 and 1968 limit the comparability of these values, and deaths in arson or post-collision vehicle fires may be omitted.<ref name="HC-3-6" /> Over the same nine decades the number of fire deaths fell by roughly two-thirds from its peak around the First World War.<ref name="HC-3-4" />
Long-term U.S. rates come from death certificates, which NFPA uses only for long-term trend analysis.<ref name="HC-3-4">Hall and Cote (2008), p. 3-4.</ref> On that basis the rate was 9.1 fire deaths per 100,000 population in 1913 and 9.9 in 1918. It fell to 3.1 in 1973 and was 1.5 in 2003.<ref name="HC-3-6" /> Classification changes in 1948, 1958 and 1968 limit the comparability of these values, and deaths in arson or post-collision vehicle fires may be omitted.<ref name="HC-3-6" /> Over the same nine decades the number of fire deaths fell by roughly two-thirds from its peak around the First World War.<ref name="HC-3-4" />
[[File:fse-fire-death-rate-v2.svg|thumb|upright=1.5|alt=Line chart of U.S. fire deaths per 100,000 population from 1913 to 2003, falling from about 9 to 1.5|U.S. fire deaths per 100,000 population from death certificates, 1913–2003; classification changes (1948, 1958, 1968) limit comparability.]]


In the early 1970s the report ''America Burning'' noted that the United States had one of the highest fire death rates in the industrialised world.<ref name="AS-3-43">Ahrens and Stewart (2008), p. 3-43.</ref> NFPA survey data put the U.S. rate at 34.4 deaths per million in 1977 and 13.1 in 2001, excluding the events of 11 September 2001. The rate in the United Kingdom fell from 15.1 to 10.1 over the same years.<ref name="HC-3-9">Hall and Cote (2008), p. 3-9.</ref> The United States and Canada tracked together, with rates twice as high as those of the United Kingdom and Japan in the late 1970s and nearly the same in the late 1990s.<ref name="HC-3-6" /> By the late 1990s the U.S. rate was lower than Japan's and sometimes lower than Sweden's.<ref name="AFK-3-64">Ahrens et al. (2008), p. 3-64.</ref>
In the early 1970s the report ''America Burning'' noted that the United States had one of the highest fire death rates in the industrialised world.<ref name="AS-3-43">Ahrens and Stewart (2008), p. 3-43.</ref> NFPA survey data put the U.S. rate at 34.4 deaths per million in 1977 and 13.1 in 2001, excluding the events of 11 September 2001. The rate in the United Kingdom fell from 15.1 to 10.1 over the same years.<ref name="HC-3-9">Hall and Cote (2008), p. 3-9.</ref> The United States and Canada tracked together, with rates twice as high as those of the United Kingdom and Japan in the late 1970s and nearly the same in the late 1990s.<ref name="HC-3-6" /> By the late 1990s the U.S. rate was lower than Japan's and sometimes lower than Sweden's.<ref name="AFK-3-64">Ahrens et al. (2008), p. 3-64.</ref>
[[File:fse-fire-death-rate-v3.svg|thumb|upright=1.5|alt=Line chart of fire deaths per million in the United States, Canada and the United Kingdom, 1977 to 2001, with the gap narrowing|Fire deaths per million population in the United States, Canada and the United Kingdom, 1977–2001; the 2001 U.S. value excludes 11 September 2001.]]


Average rates for 1993–2002, compiled by the CTIF Center of Fire Statistics, ranged widely. Russia had 108 deaths per million and Estonia 106, while the United States had 16, Japan 14, Sweden 13, Canada 12 and the United Kingdom 11. New Zealand had 10, Australia and Germany 7, and the Netherlands and Switzerland 5.<ref name="HC-3-8" /> The completeness of reporting may vary from country to country; Russian statistics, for example, jumped after the society became more open.<ref name="HC-3-6" /> Writing in 2002, Purser put U.K. and European rates at approximately 14 per million.<ref name="P-2-84" /> Ramachandran (2002) notes that the U.S. rate consistently ranked among the highest in the countries studied by the World Fire Statistics Centre.<ref name="R-5-80">Ramachandran (2002b), p. 5-80.</ref>
Average rates for 1993–2002, compiled by the CTIF Center of Fire Statistics, ranged widely. Russia had 108 deaths per million and Estonia 106, while the United States had 16, Japan 14, Sweden 13, Canada 12 and the United Kingdom 11. New Zealand had 10, Australia and Germany 7, and the Netherlands and Switzerland 5.<ref name="HC-3-8" /> The completeness of reporting may vary from country to country; Russian statistics, for example, jumped after the society became more open.<ref name="HC-3-6" /> Writing in 2002, Purser put U.K. and European rates at approximately 14 per million.<ref name="P-2-84" /> Ramachandran (2002b) notes that the U.S. rate consistently ranked among the highest in the countries studied by the World Fire Statistics Centre.<ref name="R-5-80">Ramachandran (2002b), p. 5-80.</ref>
[[File:fse-fire-death-rate-v5.svg|thumb|upright=1.5|alt=Bar chart of average fire deaths per million population, 1993 to 2002, for selected countries from Russia at 108 to Switzerland at 5|Average fire deaths per million population, 1993–2002, for selected countries (CTIF compilation); reporting completeness may differ between countries.]]


The handbook offers some explanations, mostly as hedged judgements. The United States has done well in holding down deaths per fire but poorly in preventing fires.<ref name="HC-3-8" /> Universal fire safety education, in schools and neighbourhoods, seems to be a major factor in the lower rates of western Europe and Japan.<ref name="HC-3-14">Hall and Cote (2008), p. 3-14.</ref> For Russia and much of Eastern Europe, an obvious explanation is that alcohol use and abuse appear to be higher, to the extent that statistics can be trusted. In the United States, two-fifths to one-half of adults who die in fire are legally intoxicated at the time.<ref name="H-5-107">Hall (2008), p. 5-107.</ref>
The handbook offers some explanations, mostly as hedged judgements. The United States has done well in holding down deaths per fire but poorly in preventing fires.<ref name="HC-3-8" /> Universal fire safety education, in schools and neighbourhoods, seems to be a major factor in the lower rates of western Europe and Japan.<ref name="HC-3-14">Hall and Cote (2008), p. 3-14.</ref> For Russia and much of Eastern Europe, an obvious explanation is that alcohol use and abuse appear to be higher, to the extent that statistics can be trusted. In the United States, two-fifths to one-half of adults who die in fire are legally intoxicated at the time.<ref name="H-5-107">Hall (2008), p. 5-107.</ref>
Line 54: Line 58:


Here <math>R_g</math> is the fire death rate of group ''g'', <math>R_{\mathrm{all}}</math> the rate of the whole population, and the relative risk <math>\mathrm{RR}_g</math> is dimensionless, with a value of 1 for the general population. For U.S. home fire deaths in 1992–2001, children aged 4 or younger had a rate of 22.9 per million and a relative risk of 1.74. Children aged 10–14 had the lowest rate, 5.2 per million, and people aged 85 or older the highest, 37.9 per million or a relative risk of 2.89.<ref name="MF-16-94" /> As an illustration, dividing 37.9 by 2.89 gives an all-ages home fire death rate of about 13.1 per million for that period.
Here <math>R_g</math> is the fire death rate of group ''g'', <math>R_{\mathrm{all}}</math> the rate of the whole population, and the relative risk <math>\mathrm{RR}_g</math> is dimensionless, with a value of 1 for the general population. For U.S. home fire deaths in 1992–2001, children aged 4 or younger had a rate of 22.9 per million and a relative risk of 1.74. Children aged 10–14 had the lowest rate, 5.2 per million, and people aged 85 or older the highest, 37.9 per million or a relative risk of 2.89.<ref name="MF-16-94" /> As an illustration, dividing 37.9 by 2.89 gives an all-ages home fire death rate of about 13.1 per million for that period.
[[File:fse-fire-death-rate-v4.svg|thumb|upright=1.5|alt=Bar chart of U.S. home fire deaths per million by age group, 1992 to 2001, highest for the oldest and youngest|U.S. home fire deaths per million population by age, 1992–2001, with relative risk (general population = 1).]]


The very young and older adults are the two high-risk age groups for fire death.<ref name="HT-1-8">Hall and Twomey (2008), p. 1-8.</ref> Gamache (2008) reports that children under 5 were 74 percent more likely to die in a home fire than the average person. People aged 65 and older had roughly twice the average risk, and those aged 75 and older nearly three times.<ref name="G-5-63">Gamache (2008), p. 5-63.</ref> Risk rises from about age 50 and is almost four times the average for people aged 85 and older.<ref name="G-5-67">Gamache (2008), p. 5-67.</ref> Gamache (2008) treats preschool children, older adults and people disadvantaged by income or education as the main [[high-risk populations]] for fire safety programmes.<ref name="G-5-63" /> In Japan, by contrast, young children were not a distinctly high-risk group, while the gap for older adults was much wider than in the United States.<ref name="H-5-106" />
The very young and older adults are the two high-risk age groups for fire death.<ref name="HT-1-8">Hall and Twomey (2008), p. 1-8.</ref> Gamache (2008) reports that children under 5 were 74 percent more likely to die in a home fire than the average person. People aged 65 and older had roughly twice the average risk, and those aged 75 and older nearly three times.<ref name="G-5-63">Gamache (2008), p. 5-63.</ref> Risk rises from about age 50 and is almost four times the average for people aged 85 and older.<ref name="G-5-67">Gamache (2008), p. 5-67.</ref> Gamache (2008) treats preschool children, older adults and people disadvantaged by income or education as the main [[high-risk populations]] for fire safety programmes.<ref name="G-5-63" /> In Japan, by contrast, young children were not a distinctly high-risk group, while the gap for older adults was much wider than in the United States.<ref name="H-5-106" />
Line 77: Line 82:
The Louisville, Kentucky, home smoke alarm campaign of the 1980s shows the problem. The city averaged 15 fire deaths a year in 1980–1984 and 11 a year in 1985–1988, a decline of more than one-quarter.<ref name="H-5-109" /> Treated as random variation around a constant risk, the difference between the two averages was only about 1.2 standard deviations. This was not yet statistically significant, as statisticians usually look for at least 1⅔ standard deviations in a one-sided test.<ref name="H-5-109" /> Recomputing from the nine annual tolls shown by Hall (14, 13, 19, 15 and 13, then 10, 9, 16 and 8) gives the same result of about 1.2 standard deviations. Because fires are 50 to 100 times as common as fire deaths, fire incidents show programme effects much sooner.<ref name="H-5-109" />
The Louisville, Kentucky, home smoke alarm campaign of the 1980s shows the problem. The city averaged 15 fire deaths a year in 1980–1984 and 11 a year in 1985–1988, a decline of more than one-quarter.<ref name="H-5-109" /> Treated as random variation around a constant risk, the difference between the two averages was only about 1.2 standard deviations. This was not yet statistically significant, as statisticians usually look for at least 1⅔ standard deviations in a one-sided test.<ref name="H-5-109" /> Recomputing from the nine annual tolls shown by Hall (14, 13, 19, 15 and 13, then 10, 9, 16 and 8) gives the same result of about 1.2 standard deviations. Because fires are 50 to 100 times as common as fire deaths, fire incidents show programme effects much sooner.<ref name="H-5-109" />


As an illustration of this randomness, annual death counts can be treated as Poisson counts, a distribution whose mean equals its variance.<ref name="Ha-1-188">Hall (2002a), p. 1-188.</ref> At the 2004 national rate of 13.3 per million, a city of one million people would expect about 13 deaths a year with a standard deviation of about 3.6, or 27 percent. A town of 10,000 people would expect 0.13 deaths a year, about one every seven to eight years, so a single death would change its rate enormously.
For a simple illustration, assume a stable underlying rate and independent rare deaths, so annual counts follow a Poisson distribution, whose mean equals its variance.<ref name="Ha-1-188">Hall (2002a), p. 1-188.</ref> At the 2004 national rate of 13.3 per million, a city of one million people would expect about 13 deaths a year with a standard deviation of about 3.6, or 27 percent. A town of 10,000 people would expect 0.13 deaths a year, about one every seven to eight years, so a single death would change its rate enormously.


Several other effects can mislead. Death rates vary randomly over time, often significantly. If a programme starts when such variation has pushed rates to a peak, a reduction can be expected even if the programme has no effect.<ref name="H-5-117">Hall (2008), p. 5-117.</ref> One very large incident can dominate a rate, so analyses covering 2001 must state whether the events of 11 September 2001 are included.<ref name="AFK-3-77">Ahrens et al. (2008), p. 3-77.</ref> The U.S. series for 1977–2001 quoted above excludes them.<ref name="HC-3-9" /> Groups defined too narrowly, such as households with a particular mix of habits and equipment, have no available denominator and too few fires for a reliable rate.<ref name="AFK-3-80" />
Several other effects can mislead. Death rates vary randomly over time, often significantly. If a programme starts when such variation has pushed rates to a peak, a reduction can be expected even if the programme has no effect.<ref name="H-5-117">Hall (2008), p. 5-117.</ref> One very large incident can dominate a rate, so analyses covering 2001 must state whether the events of 11 September 2001 are included.<ref name="AFK-3-77">Ahrens et al. (2008), p. 3-77.</ref> The U.S. series for 1977–2001 quoted above excludes them.<ref name="HC-3-9" /> Groups defined too narrowly, such as households with a particular mix of habits and equipment, have no available denominator and too few fires for a reliable rate.<ref name="AFK-3-80" />

Latest revision as of 18:30, 14 November 2026

The fire death rate is the number of people killed by fire in a given period, usually a year, divided by the size of the population concerned. It is commonly expressed as deaths per million population.[1] Unlike a simple count of deaths, the rate takes account of changes in population, so it is used as a measure of fire risk and for comparisons over time and between groups.[1][2] In 2004 the United States had 13 civilian fire deaths per million population, from an estimated 3,900 deaths and a resident population of 293,655,000.[2]

Some statistics, such as the U.S. death certificate series, are given per 100,000 population instead.[3] Related measures divide deaths by the number of fires or households rather than by people, or by hours of exposure where such data exist.[2][4][5] U.S. fire death rates fell steeply during the twentieth century.[3] Yet in averages for 1993–2002 compiled by the International Technical Committee for the Prevention and Extinction of Fire (CTIF), they were still higher than in many European countries.[6] Because fire deaths are rare, rates for small communities or single years vary strongly by chance.[7] The values below come mainly from the 2008 NFPA handbook and describe the United States between 1977 and 2004 unless stated otherwise.

Definition and calculation

In fire statistics, Ahrens et al. (2008) define a rate as a ratio of a measure of the size of a fire problem to a measure of the size of the group affected by it.[8] For an annual fire death rate, the problem is the number of fire deaths in one year and the group is the population. The result is multiplied by a scale factor so that it can be read as a convenient whole number:[2]

R=DPk

Symbol Description Unit
R Fire death rate deaths per million (or per 100,000) population per year
D Number of fire deaths in one year, for example civilian deaths deaths per year
P Population of the area or group, for example the resident population persons
k Scale factor: 106 for "per million", 105 for "per 100,000" –

As a worked example, the 2004 U.S. estimates give R=3,900/293,655,000×106=13.3 deaths per million population, which Ahrens et al. (2008) round to 13.[2] The same value is 1.33 per 100,000. Converting between the two scales only requires a factor of 10. A death certificate rate of 1.5 per 100,000 for 2003 therefore corresponds to 15 per million.[3]

Diagram of a fire death rate as deaths divided by population times one million, with the 2004 U.S. example giving 13 per million
A fire death rate divides the number of fire deaths by the population and scales the result; the 2004 U.S. values give about 13 deaths per million.

The numerator depends on how a fire death is defined. Deaths are generally discussed separately for civilians and for the fire service, and civilians include everyone who is not a member of the fire service.[9] In the National Fire Incident Reporting System a fire death is anyone who dies as a result of the fire within one year of it. Smoke inhalation and burns are the most common causes of death, but a person fatally injured while jumping from a burning building also counts.[9] Purser (2002) notes that toxic smoke products are recognised as the major cause of death in fires in the United States and the United Kingdom.[10] Van Rossum (2027) describes people dying because they are trapped by fire or smoke, overcome in their sleep or poisoned by inhaling toxic smoke gases, and notes that most fire deaths occur at home.[11]

The comparisons here use two kinds of national numerator. NFPA's estimates are sample-based projections from the NFPA fire experience survey and include deaths from arson, self-immolation and vehicle fires.[12] The National Safety Council's death-certificate series excludes some of these deaths, which may instead be classified as homicide, suicide or vehicle accidents.[12] Different organisations also use different combinations of disease classification codes to identify fire deaths.[9] Rates from different sources are therefore not identical, although their trends and patterns are consistent.[12]

Rates and counts

Rates can be used in any analysis where the size of the affected group may change.[2] NFPA estimated that annual U.S. fire deaths fell by 40 percent from 1980 through 2004, while the resident population grew by 29 percent. The rate of deaths per million therefore fell by 54 percent.[2] In terms of the definition, the new rate is 0.60 / 1.29 = 0.465 times the old one, a fall of 53.5 percent. Ahrens et al. (2008) conclude that increased fire safety may be best measured by the decline in the fire death rate, because the rate reflects changes in both deaths and population.[2]

Over long periods the difference can be large. National Safety Council analysis of death certificates showed U.S. unintentional fire deaths in 2004 to be roughly 56 percent lower than in 1913. The 2004 rate, however, was 86 percent lower than the 1913 rate.[13] Because of definitional changes, these percentages should be considered rough estimates. They nevertheless show how different progress can look when measured by numbers or by rates.[13]

The choice of denominator can also change the conclusion. A fall in fire deaths in manufactured homes might mean that these homes became safer, or only that fewer people lived in them. Deaths per million people living in manufactured homes would help distinguish the two explanations.[13] Adding a state to a database in the middle of a trend period can still bias such a rate if that state differs, for example in its share of manufactured homes.[13]

Space heaters give a clear example of rates and counts pointing in opposite directions. For 1995–1999 fires and 1997 usage, portable liquid-fuelled heaters had 9.93–12.41 deaths per million households using them, against 5.95–7.11 for portable electric heaters.[5] In the unadjusted death counts for the same fires, however, portable electric heaters had 90 deaths and portable liquid-fuelled heaters 40. Sorting by death toll produced a ranking almost exactly opposite to the ranking by rates.[5]

Trends and international comparisons

Long-term U.S. rates come from death certificates, which NFPA uses only for long-term trend analysis.[14] On that basis the rate was 9.1 fire deaths per 100,000 population in 1913 and 9.9 in 1918. It fell to 3.1 in 1973 and was 1.5 in 2003.[3] Classification changes in 1948, 1958 and 1968 limit the comparability of these values, and deaths in arson or post-collision vehicle fires may be omitted.[3] Over the same nine decades the number of fire deaths fell by roughly two-thirds from its peak around the First World War.[14]

Line chart of U.S. fire deaths per 100,000 population from 1913 to 2003, falling from about 9 to 1.5
U.S. fire deaths per 100,000 population from death certificates, 1913–2003; classification changes (1948, 1958, 1968) limit comparability.

In the early 1970s the report America Burning noted that the United States had one of the highest fire death rates in the industrialised world.[15] NFPA survey data put the U.S. rate at 34.4 deaths per million in 1977 and 13.1 in 2001, excluding the events of 11 September 2001. The rate in the United Kingdom fell from 15.1 to 10.1 over the same years.[16] The United States and Canada tracked together, with rates twice as high as those of the United Kingdom and Japan in the late 1970s and nearly the same in the late 1990s.[3] By the late 1990s the U.S. rate was lower than Japan's and sometimes lower than Sweden's.[17]

Line chart of fire deaths per million in the United States, Canada and the United Kingdom, 1977 to 2001, with the gap narrowing
Fire deaths per million population in the United States, Canada and the United Kingdom, 1977–2001; the 2001 U.S. value excludes 11 September 2001.

Average rates for 1993–2002, compiled by the CTIF Center of Fire Statistics, ranged widely. Russia had 108 deaths per million and Estonia 106, while the United States had 16, Japan 14, Sweden 13, Canada 12 and the United Kingdom 11. New Zealand had 10, Australia and Germany 7, and the Netherlands and Switzerland 5.[6] The completeness of reporting may vary from country to country; Russian statistics, for example, jumped after the society became more open.[3] Writing in 2002, Purser put U.K. and European rates at approximately 14 per million.[10] Ramachandran (2002b) notes that the U.S. rate consistently ranked among the highest in the countries studied by the World Fire Statistics Centre.[18]

Bar chart of average fire deaths per million population, 1993 to 2002, for selected countries from Russia at 108 to Switzerland at 5
Average fire deaths per million population, 1993–2002, for selected countries (CTIF compilation); reporting completeness may differ between countries.

The handbook offers some explanations, mostly as hedged judgements. The United States has done well in holding down deaths per fire but poorly in preventing fires.[6] Universal fire safety education, in schools and neighbourhoods, seems to be a major factor in the lower rates of western Europe and Japan.[19] For Russia and much of Eastern Europe, an obvious explanation is that alcohol use and abuse appear to be higher, to the extent that statistics can be trusted. In the United States, two-fifths to one-half of adults who die in fire are legally intoxicated at the time.[20]

Rates for groups and places

Ahrens et al. (2008) note that an individual wants to know his or her own risk, not that of an average American. Failing that, the person wants the risk for people as much like him or her as possible. Rates for specifically defined groups provide that information.[2] Australian performance-based design uses the concept of individual risk. The Fire Safety Verification Method handbook (2020) interprets it as the frequency at which an individual may be expected to sustain a given level of harm from a fire in the subject building.[21] Group rates can be compared through the relative risk, the group's rate divided by the rate for the general population:[22]

RRg=RgRall

Here Rg is the fire death rate of group g, Rall the rate of the whole population, and the relative risk RRg is dimensionless, with a value of 1 for the general population. For U.S. home fire deaths in 1992–2001, children aged 4 or younger had a rate of 22.9 per million and a relative risk of 1.74. Children aged 10–14 had the lowest rate, 5.2 per million, and people aged 85 or older the highest, 37.9 per million or a relative risk of 2.89.[22] As an illustration, dividing 37.9 by 2.89 gives an all-ages home fire death rate of about 13.1 per million for that period.

Bar chart of U.S. home fire deaths per million by age group, 1992 to 2001, highest for the oldest and youngest
U.S. home fire deaths per million population by age, 1992–2001, with relative risk (general population = 1).

The very young and older adults are the two high-risk age groups for fire death.[23] Gamache (2008) reports that children under 5 were 74 percent more likely to die in a home fire than the average person. People aged 65 and older had roughly twice the average risk, and those aged 75 and older nearly three times.[24] Risk rises from about age 50 and is almost four times the average for people aged 85 and older.[25] Gamache (2008) treats preschool children, older adults and people disadvantaged by income or education as the main high-risk populations for fire safety programmes.[24] In Japan, by contrast, young children were not a distinctly high-risk group, while the gap for older adults was much wider than in the United States.[5]

Sex and place also matter. In 1999–2002 the U.S. rate for males was 12.1 fire deaths per million, 38 percent higher than the 8.8 for females. The largest difference was at age 85 and over, with 58.5 deaths per million for men and 27.8 for women.[25] Rural communities with fewer than 2,500 people had an average rate of 30.9 per million in 1997–2001, excluding 11 September 2001. This was at least twice the rate of most other community sizes; communities of 2,500 to 5,000 had 18.4.[17]

Rates also differed by region and by income. In the early 1980s rates in the South were typically about 50 percent higher than in the Northeast and North Central regions and double those in the West. By 1997–2001 the South averaged one-fifth more than the Northeast and North Central regions, and the West half the South's rate.[17] Gamache (2008) notes that high fire death rates are often associated with poverty and low education, and attributes the higher rates in the South to rural poverty.[24][26] In 1997–1998 the rate for Native Americans and Alaska Natives was 2.1 times that for whites. In Canada the 2003 rate in First Nations communities was 3.0 per 100,000, 2.4 times the 1999 national average of 1.27.[27]

Other rate measures

A rate per fire measures how deadly fires are once they start, rather than how often people die. In reported U.S. home fires in 1999–2002, the death rate was 0.4 per 100 fires when a smoke alarm operated and 1.0 per 100 fires when no smoke alarm operated.[4] Fires started by smoking materials caused 3.4 deaths per 100 fires, against 0.3 for cooking equipment fires.[28] The death rate in 1989–1998 residential fires with automatic suppression equipment such as sprinklers was 78 percent lower than in fires without it.[17]

Per-fire rates are also used in design. The Data Sheets of the Australian Fire Safety Verification Method (2020) cite an NFPA study of U.S. home structure fires in 2009–2013. It found 5.3 deaths per 1,000 fires when a smoke alarm was present and operated, and 11.8 per 1,000 when no alarm was present or it did not operate.[29] The reduction of about 55 percent, (11.8 − 5.3) / 11.8, is offered as an alternative way to express smoke alarm efficacy. A U.K. comparison quoted there gives 8 deaths per 1,000 fires with no alarm operating against 4 with alarms that operated within 5 minutes of ignition, a reduction of 50 percent.[29]

Other denominators measure exposure. Comparing firefighting with other occupations needs on-duty deaths per hour of exposure, but no national statistics exist on the hours served by volunteer firefighters.[2] For homes and hotels, counts of persons do not work because the length of stay differs so much, and numbers of units might work better if such data could be obtained.[2] Van Rossum (2027) observes that, per hour of stay, the chance of dying in a fire is much greater in a residential building than in an assembly building. Dutch fire detection requirements are nevertheless stricter for assembly buildings.[30]

For cost-benefit studies of residential sprinklers, Notarianni (2002) defines a national residential death rate as residential fire deaths divided by the number of occupied residential units. NFPA published such rates by region, community size and house type, giving 4 × 8 × 3 = 96 values in that study.[31] Hall and Ahrens (2002) likewise describe usage and exposure data as the denominators of probability calculations, such as the numbers of heaters of each type in use.[32]

Statistical limits

Fire deaths are rare enough that rates for small populations are unstable. For total U.S. civilian fire deaths, year-to-year tracking is generally meaningful. For a single large city it is subject to considerable statistical noise, and towns and rural areas can only be tracked as part of a larger aggregation.[7] Hall (2008) advises a baseline of at least five years, more for communities of fewer than 100,000 people. Near 10,000 people, a community's own data are too few to track success by fire deaths in any reasonable period.[33]

The Louisville, Kentucky, home smoke alarm campaign of the 1980s shows the problem. The city averaged 15 fire deaths a year in 1980–1984 and 11 a year in 1985–1988, a decline of more than one-quarter.[33] Treated as random variation around a constant risk, the difference between the two averages was only about 1.2 standard deviations. This was not yet statistically significant, as statisticians usually look for at least 1⅔ standard deviations in a one-sided test.[33] Recomputing from the nine annual tolls shown by Hall (14, 13, 19, 15 and 13, then 10, 9, 16 and 8) gives the same result of about 1.2 standard deviations. Because fires are 50 to 100 times as common as fire deaths, fire incidents show programme effects much sooner.[33]

For a simple illustration, assume a stable underlying rate and independent rare deaths, so annual counts follow a Poisson distribution, whose mean equals its variance.[34] At the 2004 national rate of 13.3 per million, a city of one million people would expect about 13 deaths a year with a standard deviation of about 3.6, or 27 percent. A town of 10,000 people would expect 0.13 deaths a year, about one every seven to eight years, so a single death would change its rate enormously.

Several other effects can mislead. Death rates vary randomly over time, often significantly. If a programme starts when such variation has pushed rates to a peak, a reduction can be expected even if the programme has no effect.[35] One very large incident can dominate a rate, so analyses covering 2001 must state whether the events of 11 September 2001 are included.[36] The U.S. series for 1977–2001 quoted above excludes them.[16] Groups defined too narrowly, such as households with a particular mix of habits and equipment, have no available denominator and too few fires for a reliable rate.[2]

See also

Notes

  1. ↑ 1.0 1.1 Ahrens et al. (2008), pp. 3-79–3-80.
  2. ↑ 2.00 2.01 2.02 2.03 2.04 2.05 2.06 2.07 2.08 2.09 2.10 2.11 Ahrens et al. (2008), p. 3-80.
  3. ↑ 3.0 3.1 3.2 3.3 3.4 3.5 3.6 Hall and Cote (2008), p. 3-6.
  4. ↑ 4.0 4.1 Ahrens et al. (2008), p. 3-68.
  5. ↑ 5.0 5.1 5.2 5.3 Hall (2008), p. 5-106.
  6. ↑ 6.0 6.1 6.2 Hall and Cote (2008), p. 3-8.
  7. ↑ 7.0 7.1 Ahrens et al. (2008), p. 3-87.
  8. ↑ Ahrens et al. (2008), p. 3-79.
  9. ↑ 9.0 9.1 9.2 Ahrens and Stewart (2008), p. 3-49.
  10. ↑ 10.0 10.1 Purser (2002), p. 2-84.
  11. ↑ Van Rossum (2027), §7.
  12. ↑ 12.0 12.1 12.2 Ahrens et al. (2008), p. 3-91.
  13. ↑ 13.0 13.1 13.2 13.3 Ahrens et al. (2008), p. 3-83.
  14. ↑ 14.0 14.1 Hall and Cote (2008), p. 3-4.
  15. ↑ Ahrens and Stewart (2008), p. 3-43.
  16. ↑ 16.0 16.1 Hall and Cote (2008), p. 3-9.
  17. ↑ 17.0 17.1 17.2 17.3 Ahrens et al. (2008), p. 3-64.
  18. ↑ Ramachandran (2002b), p. 5-80.
  19. ↑ Hall and Cote (2008), p. 3-14.
  20. ↑ Hall (2008), p. 5-107.
  21. ↑ Australian Building Codes Board (2020b), p. 22.
  22. ↑ 22.0 22.1 Madrzykowski and Fleming (2008), p. 16-94.
  23. ↑ Hall and Twomey (2008), p. 1-8.
  24. ↑ 24.0 24.1 24.2 Gamache (2008), p. 5-63.
  25. ↑ 25.0 25.1 Gamache (2008), p. 5-67.
  26. ↑ Gamache (2008), p. 5-71.
  27. ↑ Gamache (2008), p. 5-76.
  28. ↑ Ahrens et al. (2008), p. 3-74.
  29. ↑ 29.0 29.1 Australian Building Codes Board (2020a), Data Sheet C3, pp. 109–111.
  30. ↑ Van Rossum (2027), §1.2.
  31. ↑ Notarianni (2002), p. 5-61.
  32. ↑ Hall and Ahrens (2002), p. 5-65.
  33. ↑ 33.0 33.1 33.2 33.3 Hall (2008), p. 5-109.
  34. ↑ Hall (2002a), p. 1-188.
  35. ↑ Hall (2008), p. 5-117.
  36. ↑ Ahrens et al. (2008), p. 3-77.

References

  • Ahrens, M., Frazier, P., & Kelch, G. (2008). Use of fire incident data and statistics. In A. E. Cote, C. C. Grant, J. R. Hall, Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 3-61–3-92). National Fire Protection Association.
  • Ahrens, M., & Stewart, S. (2008). Fire data collection and databases. In A. E. Cote, C. C. Grant, J. R. Hall, Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 3-43–3-59). National Fire Protection Association.
  • Australian Building Codes Board. (2020a). Data sheets: Fire Safety Verification Method (Handbook annex, Print version 1.1).
  • Australian Building Codes Board. (2020b). Handbook: Fire Safety Verification Method (Print version 1.1).
  • Gamache, S. (2008). Reaching high-risk groups. In A. E. Cote, C. C. Grant, J. R. Hall, Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 5-63–5-85). National Fire Protection Association.
  • Hall, J. R., Jr. (2002a). Probability concepts. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), SFPE handbook of fire protection engineering (3rd ed., pp. 1-182–1-192). National Fire Protection Association; Society of Fire Protection Engineers.
  • Hall, J. R., Jr. (2008). Using data for public education planning and decision making. In A. E. Cote, C. C. Grant, J. R. Hall, Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 5-103–5-118). National Fire Protection Association.
  • Hall, J. R., Jr., & Ahrens, M. J. (2002). Data for engineering analysis. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), SFPE handbook of fire protection engineering (3rd ed., pp. 5-65–5-78). National Fire Protection Association; Society of Fire Protection Engineers.
  • Hall, J. R., Jr., & Cote, A. E. (2008). An overview of the fire problem and fire protection. In A. E. Cote, C. C. Grant, J. R. Hall, Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 3-3–3-30). National Fire Protection Association.
  • Hall, J. R., Jr., & Twomey, E. R. (2008). Challenges to safety in the built environment. In A. E. Cote, C. C. Grant, J. R. Hall, Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 1-3–1-29). National Fire Protection Association.
  • Madrzykowski, D., & Fleming, R. P. (2008). Residential sprinkler systems. In A. E. Cote, C. C. Grant, J. R. Hall, Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 16-91–16-107). National Fire Protection Association.
  • Notarianni, K. A. (2002). Uncertainty. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), SFPE handbook of fire protection engineering (3rd ed., pp. 5-40–5-64). National Fire Protection Association; Society of Fire Protection Engineers.
  • Purser, D. A. (2002). Toxicity assessment of combustion products. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), SFPE handbook of fire protection engineering (3rd ed., pp. 2-83–2-171). National Fire Protection Association; Society of Fire Protection Engineers.
  • Ramachandran, G. (2002b). Measuring fire consequences in economic terms (J. R. Hall Jr., Rev.). In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), SFPE handbook of fire protection engineering (3rd ed., pp. 5-79–5-92). National Fire Protection Association; Society of Fire Protection Engineers.
  • Van Rossum, E. (2027). Introduction to Fire Safety Engineering. Fire Safety Academy.