Journal of Safety Research 55 (2015) 99–103

Contents lists available at ScienceDirect

Journal of Safety Research

journal homepage: www.elsevier.com/locate/jsr

Assessing the number of fire fatalities in a defined population

Anders Jonsson, a,b,⁎ Anders Bergqvist, c Ragnar Andersson a
a
Division of Risk Management, Department of Health and Environmental Sciences, Karlstad University, Karlstad SE-651 88, Sweden
b
Evaluation & Monitoring Department, Swedish Civil Contingencies Agency, Karlstad SE-651 81, Sweden
c
Swedish Fire Protection Association, Stockholm SE-115 87, Sweden

a r t i c l e i n f o a b s t r a c t

Article history: Introduction: Fire-related fatalities and injuries have become a growing governmental concern in Sweden, and a
Received 12 February 2015 national vision zero strategy has been adopted stating that nobody should get killed or seriously injured from
Received in revised form 9 August 2015 fires. There is considerable uncertainty, however, regarding the numbers of both deaths and injuries due to
Accepted 1 October 2015 fires. Different national sources present different numbers, even on deaths, which obstructs reliable surveillance
Available online 23 October 2015
of the problem over time. We assume the situation is similar in other countries. This study seeks to assess
the true number of fire-related deaths in Sweden by combining sources, and to verify the coverage of each
Keywords:
Fatal fires
individual source. By doing so, we also wish to demonstrate the possibilities of improved surveillance practices.
Fire deaths Method: Data from three national sources were collected and matched; a special database on fatal fires held by
Fire-related mortality The Swedish Contingencies Agency (nationally responsible for fire prevention), a database on forensic medical
Fire injury surveillance examinations held by the National Board of Forensic Medicine, and the cause of death register held by the
Record linkage Swedish National Board of Health and Welfare. Results: The results disclose considerable underreporting in the
single sources. The national database on fatal fires, serving as the principal source for policymaking on fire-
prevention matters, underestimates the true situation by 20%. Its coverage of residential fires appears to be better
than other fires. Conclusions: Systematic safety work and informed policy-making presuppose access to correct
and reliable numbers. By combining several different sources, as suggested in this study, the national database
on fatal fires is now considerably improved and includes regular matching with complementary sources.
© 2015 National Safety Council and Elsevier Ltd. All rights reserved.

1. Introduction are differences in how individual countries define what is to be included

According to WHO estimates, globally, more than 300,000 people
are annually killed as a result of fires (WHO, 2004). In a global perspec-
tive, fires represent a significant share of deaths caused by injury,
equivalent to, for example, poisoning and drowning, with countries in
Africa, South-east Asia and the Eastern Mediterranean reporting almost
twice as high mortality rates compared to Europe and the United States
(WHO, 2004).
While mortality data are recorded in most countries, official records
generally suffer from great uncertainty. Nationally reported numbers on
fire fatalities usually originate from mortality data (WHO, 2015) or fire
brigade records (Geneva Association, 2014; European Fire Safety
Network, 2009), often with great disparities in numbers between the
two sources. One reason for diverging figures may be differences
in inclusion criteria between sources, but under-reporting, over-
reporting, and masked cases reported under other categories might be
complementary explanations. A recent report comparing different
European countries' fire statistics (DCLG, 2012) concludes that there
⁎ Corresponding author at: Division of Risk Management, Department of Environmental
and Life Sciences, Karlstad University, SE-651 88 Karlstad, Sweden. Tel.: + 46 54 700 2337.
E-mail address: anders.jonsson@kau.se (A. Jonsson).
in their statistics.
Internationally published research on fire-related deaths and
mortality usually address specific subcategories, such as dwelling fires,
certain risk groups, evaluate different prevention measures, and so
forth. Few studies attempt to map the entire spectrum of fire-related
deaths in a defined population. One common source for researchers to
identify and describe fire deaths is medical examiner reports (Runyan,
Bangdiwala, Linzer, Sacks, & Butts, 1992; Chernichko, Saunders, &
Tough, 1993; Barillo, & Goode, 1996; Marshall et al., 1998). Alternatively,
police reports (Leth, Gregersen, & Sabroe, 1998), official mortality
statistics/death certificates (DiGuiseppi, Roberts, & Li, 1998; Diekman,
Ballesteros, Berger, Caraballo, & Kegler, 2008; Yau, & Marshall, 2014),
or reports from fire investigators/fire service (Duncanson, Woodward,
& Reid, 2002; Holborn, Nolan, & Golt, 2003; Anderson, & Ezekoye,
2013) have been used. In some studies fire records are linked with
medical examiner reports (Istre, McCoy, Carlin, & McClain, 2002)
and in countries possessing coronial reports, these have shown to be a
fruitful source to characterize fire fatalities and the circumstances lead-
ing up to the fire onset (Brennan, 1999; Harpur, Boyce, & McConnell,
2013).
In Sweden, statistics reported by the fire brigade seem fairly compa-
rable with statistics from the Swedish cause of death register, both
showing approximately 100 fire fatalities per year (Swedish Civil

http://dx.doi.org/10.1016/j.jsr.2015.10.001
0022-4375/© 2015 National Safety Council and Elsevier Ltd. All rights reserved.
100 A. Jonsson et al. / Journal of Safety Research 55 (2015) 99–103
Contingencies Agency, 2014; The National Board of Health and Welfare.
2014). However, without a detailed study, at individual level, the degree
of overlap between the sources is unknown as is the level of under- and
over-coverage in each source.
Due to the existence of different sources reporting different
numbers, and due to the lack of validity assessments of these sources
with regard to fire-related deaths, the knowledge of how many people
actually die from fire-related events in Sweden remains unclear. There-
fore, the aim of this study was to assess the true number of fire-related
deaths in Sweden by combining sources and to verify the coverage of
each individual source.
2. Materials and methods
Information from three different Swedish data sources were collected
and matched; the database on fatal fires, held by the Swedish Civil
Contingencies Agency (MSB), the database on forensic medical exami-
nations held by the National Board of Forensic Medicine (RMV) and
the Cause of death register held by the Swedish National Board of
Health and Welfare (SoS). MSB is responsible for issues concerning
civil protection, public safety, emergency management, and civil
defense as long as no other authority has responsibility. Responsibility
refers to measures taken before, during, and after an emergency or
crisis. The agency serves as the national focal point for fire prevention
and besides the database on fatal fires the agency, for example, also
keeps records on all emergency turnouts reported from local fire
departments all over Sweden. RMV is an agency whose principal task
is to produce forensic reports required in legal cases. RMW investigates,
at the request of the police, the cause of death in cases of determined or
suspected unnatural deaths. SoS is the central government agency for
matters within the fields of social services, health and medical services,
environmental health, communicable disease prevention and epidemi-
ology. One of their official duties is maintaining health data registers
and official statistics.
We defined our target population as those deceased, within a
month, as result of a fire or explosive combustion process occurring in
Sweden between 1999 and 2007. A prerequisite for a successful match
between the three databases is that they all have the same unique
identifier for the victims to ensure that any particular individual is
only included once.
Personal identification numbers (PINs) were introduced in Sweden
in 1947. Since then all residents with a permanent address have been
assigned a unique PIN. These numbers are used widely in society in all
contexts where it is important not to mix up the identity of individuals.
Concerning the health service, the use of PIN is motivated by the need
for the unambiguous identification of patients to promote safe and
effective healthcare for the individual. In addition, the use of PIN also
makes it possible to use medical data for educational purposes, research,
and continuous improvement of the quality of healthcare (Ludvigsson,
Otterblad-Olausson, Pettersson, & Ekbom, 2009).
There are two legal frameworks to consider when authorities use
health data and PIN in Sweden, the Data Protection Act and the Public
Access to Information and Secrecy Act. The Data Protection Act provides
a very high level of protection for health data, but the use of health data
is permitted in research that has been approved by an ethical committee
and in the production of statistics. It should be noted that the Data
Protection Act only applies to living people. The act does not influence
how authorities use data on those who have died in fires, but does
apply when following up the status of people with potentially life
threatening injuries from a fire.
The Public Access to Information and Secrecy Act contains very strict
regulations concerning the protection of personal and economic data of
individuals, regardless of whether they are dead or alive. In normal cir-
cumstances it would be unthinkable for data from the health authorities
or RMV to be released to MSB since the data would be available for pub-
lic access there. However, the statistics function at MSB is able to protect
the data due to a clause on statistical secrecy in the Public Access to In-
formation and Secrecy Act. This clause mandates strict secrecy for per-
sonal and economic data in the production of statistics and similar
investigations, as long as they have been specifically named in govern-
ment regulations following from the Act. Investigations by the MSB into
accidents and rescue service responses are named in paragraph 7 of the
Public Access to Information and Secrecy Ordinance. Consequently all
the data are protected at the MSB and can only be released in identifi-
able form to research that has been approved by an ethical committee.
It then becomes the responsibility of the researcher to protect the
sensitive data. It is therefore unproblematic for the other agencies to
release their data to MSB and enable the triangulation of the data.
Both the database on forensic medical examinations and the cause of
death register use the Swedish personal identity number (PIN) as iden-
tifier. The database of fatal fires, which served as a starting point for the
building process of a new aggregate dataset, however, lacks in the
majority of cases information on the victims' PIN. Therefore, finding
the PIN was crucial and with help from RMV matching was achieved
by using background information such as date, location, age, sex, and
in some cases the first and last name of victim(s). Moreover, a
computer-based free text search on terms such as fire and smoke was
performed in order to identify possible new cases not known by MSB.
In the second step, the cause-of-death register was searched for
possible additional fire-related fatalities that were not present in the
database on fatal fires or the database on forensic medical examinations.
SoS keeps records and produces official statistics on all deaths among
Swedish residents, whether the deceased was a Swedish citizen or not
and whether the death occurred in Sweden or abroad. The causes
of death are classified according to the English version of WHO's
International Statistical Classification of Diseases and Related Health
Problems, Tenth Revision (ICD-10). We searched the register among
underlying causes of death, utilizing codes presented in Table 1, and
then manually excluded entries that did not meet our inclusion criteria.
Finally, our dataset was sent back to both RMV and SoS for
supplementary data on each valid case (based on PIN) from their
registries. A final check by looking at all data simultaneously ensured
that the inclusion criteria were met for all cases included. At the same
time, the fires were classified into either “residential fires,” “fires after
traffic incidents,” “fires in parked cars,” “fires in person” or “other
fires.” Other fires mainly consisted of fires in the open (e.g., grass
fires). The fires were also classified into either “unintentional fire” or
“intentional fire.” All cases coded as intentional self-harm or assault as
underlying cause of death were classified as intentional. The remaining
cases where the fire brigade assessed the fire to be deliberately set, and
cases where underlying cause of death were classified as undetermined
intent, were manually examined to determine intent.
In Table 2 we outline the concepts that we use. Source is equated
with database and when we use a selection/subset from a database/
source we call this a dataset.
3. Results
Altogether, 1,171 fire deaths were identified during the study period
(1999–2007), equating to an average of 130 deaths per year. Of the
1,171 fire deaths, 947 (81%) were registered in the database on fatal
fires. Out of these, one could not be found in any of the other two
sources. 1,110 (95%) were registered in the forensic database. Out of
Table 1
Underlying cause of death codes related to fire. (ICD-10, English version).
ICD-code
X00-X09 Exposure to smoke, fire and flames
X76 Intentional self-harm by smoke, fire and flames
X97 Assault by smoke, fire and flames
Y26 Exposure to smoke, fire and flames, undetermined intent
 A. Jonsson et al. / Journal of Safety Research 55 (2015) 99–103 101

Table 2
Explanation of used vocabulary and notation.

Term/notation Explanation

Target population Desired population (theoretically defined by inclusion criteria)

Assessed population Our estimate of the target population
Over-coverage Ineligible units (i.e. entries in a specific dataset not fulfilling the inclusion criteria of the target population)
Over-reporting or multiple records (i.e. a specific dataset contains duplicate entries)
Under-coverage Entries in our assessed population not found in a specific dataset
Under-reporting
Exclusive contribution Entries in our assessed population only found in a specific dataset (relative complement)
∩ Intersection (e.g. A ∩ B = entries in A and B)
U Union (e.g. A U B = entries in A or B, or both)

Database/source All cases Datasets matching our criteria

The database on fatal fires (X) (X′)

The database on forensic examinations (Y) (Y′)
The cause of death registry (Z) (Z′)

these, 12 could not be found in any of the other two sources. The cause
of death register included 1,130 (96%) cases, of which 32 could not be
found in any of the other two sources. Thus, 890 deaths (76%) were
identified in all three sources, while 236 deaths (20%) were found in
two sources and the remaining 45 deaths (4%) only in one of the sources
(Table 3, Fig. 1).
Table 4 provides a breakdown of the total number of fire fatalities
and the under-coverage in each individual dataset (X′, Y′, Z′) by type
of event. Of the 1,171 fire fatalities, residential fires were the most
common (77%), followed by fires related to a traffic incident (10%).
The proportion of fatalities as a result of intentional fires differs consid-
erably between the categories. In residential fires 92% of the fatalities
were as a result of unintentional incidents while 8% were due to
intentional incidents. However, in the categories “fires in parked cars”
and “fires in persons” the reverse relationship was observed, with
more intentional incidents than unintentional. Of the cases that were
not found in the fatal fire database (X′) (n = 224), 37% were due to
residential fires followed by transport accidents (26%) and other fires
(17%). This means, by type of event, that while X′ missed some deaths
from residential fires (82 out of 903 = 9%), more than half of all deaths
in the other event categories were absent. Of the cases that do not
appear in the forensic dataset (Y′) (n = 61), 61% were categorized as
residential fires and 38% as other fires. By type of event, Y′ missed
some deaths from dwelling fires (37 out of 903 = 4%), but 31% (23 of
75 fires) of other fires. Of the cases that do not appear in the cause of
death register (Z′) (n = 41), 46% were due to residential fires followed
by transport accidents (29%), meaning that Z′ missed very few residen-
tial fatalities (19 out of 903 = 2%), but about 10% of all deaths from
transport related fires and fires in persons (Table 4).
Table 4 also provides a breakdown of the exclusive contributions to
our assessed total number from each individual dataset (X′, Y′, Z′). The
database on fatal fires (X′) included one exclusive case, the forensic da-
tabase (Y′) included 12 exclusive cases and the cause of death register
included 32 exclusive cases.
Table 3
Estimates of fire fatalities in Sweden, 1999–2007, by source and from matching.
Number Exclusive Under-coverage
contribution
Dataset, fatal fires (X′) 947 1 224
Dataset, forensic examinations (Y′) 1,110 12 61
Dataset, cause of death registry (Z′) 1,130 32 41
X′ ∩ Y′ 918 28
X′ ∩ Z′ 918 28
Y′ ∩ Z′ 1,070 180
X′ ∩ Y′ ∩ Z′ 890
X′ U Y′ U Z′ 1,171
4. Discussion
This study shows that the annual number of fire-related deaths in
Sweden is considerably higher than previously reported when single
sources, such as reports from the fire brigades or the national cause
of death register, have been used. Based on the results in this study,
it can be shown that the regularly reported statistics systematically
underestimate the true situation by approximately 20–25%.
As was seen in the results, the under-coverage in single sources can
be substantial. There are some obvious reasons and natural explana-
tions why we do not find entries from all three registers for all the fire
deaths. Firstly, cases missing from the database of fatal fires could be
due to the method of injury surveillance used by MSB. MSB identify
and detect fire fatalities through media reporting and the municipal
fire brigades reporting system. Some of the fire-related deaths were
never reported to the emergency services and are therefore missed.
Fires in person, fires in parked cars, deaths in fire due to traffic incidents,
and other fires seem to be missing to a larger extent than residential
fires. For these categories, especially fires in parked cars and fires in per-
son, there are a greater proportion of intentional events. These events
more seldom lead to efforts from rescue services, which in turn may
be an explanation for why they are absent from the database. Further-
more, there are difficulties in determining the underlying cause of
death in car crash fatalities associated with fire (Viklund, Björnstig,
Larsson, & Björnstig, 2013) and this may cause an under-reporting in
the fatal fire database. Additionally, approximately twice as many
(30%) of the under-reported deaths died in hospital compared with
those reported (15%). This would indicate that a lack of communication
between healthcare and rescue services might be responsible for some
of the under-reporting. Secondly, in terms of the database on forensic
examinations and the cause of death register, the under-reporting is
likely due to either incomplete PIN due to registration problems or
non-Swedish citizens (cause of death register) or that no autopsy was
performed (database on forensic examinations). Among the under-
reported cases in the database on forensic examinations, the average
age was 81 and it may have been determined that an autopsy was not
required.
The under-coverage, in both cause of death (41/1,171 = 3.1%) and
forensic examinations (61/1,171 = 5.2%) (shown in Table 4), when
applying our approach with dependent selection using PIN, is in fact
small. The under-coverage in cause of death would be as large as 25%
if an independent ICD selection (Table 1, underlying cause of death)
had been used. This is due to the fact that not all of the cases we assess
as fire fatalities are coded with an external cause (as in Table 1) as the
underlying cause of death. Instead, the code may be specified as a con-
tributing cause or not coded at all. Coding errors, coding difficulties,
and the fact that different ICD code abstraction strategies generate
102 A. Jonsson et al. / Journal of Safety Research 55 (2015) 99–103

fatal fires (X') 1

28

28
890

32

12

cause of death registry (Z')

forensic examinations (Y') 180

Fig. 1. Venn diagram (non-proportional) showing overlaps between, and exclusive contributions from, three sources of fire fatality data in Sweden, 1997–2007.

different cohorts are reported from related areas (Jansson, 2005;
Wilhelms, Huss, Granath, & Sjöberg, 2010; Krishnan, Frew, Green,
Martin, & Dziewulski, 2013).
During our study we also observed other problematic issues in the
collection of data. We observed over-coverage, in both the database
on fatal fires and in the ICD selection from the cause of death register,
in relation to our inclusion criteria. In the database on fatal fires, 18
deaths were reported as fire fatalities but did not meet the inclusion
criteria (duplicates (n = 3), deaths before the fire event (n = 10) and
not fire-related deaths (n = 5)). In the cause of death register, 37 deaths
did not meet the inclusion criteria, primarily due to deaths occurring
outside of Sweden (n = 10), cases where victims died more than one
month after the fire incident (n = 21) and not fire-related deaths
(n = 6).
One could hypothetically assume the existence of additional cases
not captured by any of the sources. By using our method with three
different sources we believe this risk is minimal. One approach to
validate the results from our method (multi-source dependent linkage)
is to merge independent selections from the database of fatal fires and
ICD selection from the cause of death register, cleaning for ineligible
cases and eliminate duplicate cases. Since such an approach is likely to
be biased downwards a more appropriate method for validation
would be a capture–recapture analysis (Chao, Tsay, Lin, Shau, & Chao,
2001). Therefore, in order to additionally validate our assessment, we
performed a simple two-sided capture–recapture analysis by matching
the cause of death and fatal fire database (mutually independent
sources which is a methodological prerequisite). Our analysis (not
shown) resulted in totals very close to what we have presented from
our aggregate dataset therefore strengthening the trustworthiness of
our results.
While the results in this study highlight the difficulties in assessing
the true number of fire deaths in Sweden using the previously used
crude selections, this study also shows the importance of using different
data sources in order to obtain a more complete picture of the fatality,
especially from a preventative perspective. In order to obtain informa-
tion regarding the individual, the cause of death register and the data-
base on forensic examinations provides considerable information,
both through the information collected directly, and the information
that is possible to link through the PIN. However, no information
regarding the actual fire and its causes is possible to extract. The fatal
fire database, meanwhile, focuses on describing the mechanisms
behind, and the effects of, fires. However, due to a low level of registra-
tion of PIN the database rarely enables thorough descriptions of the
deceased persons, making it impossible to link this data with other
personal information. The method used in this study combines both
the individual and fire-related information meaning that unique oppor-
tunities are available in studying and analyzing fire characteristics,
personal characteristics, and the circumstances surrounding fatal fires
in Sweden further.
The results from this study can be seen as a starting point in future
fact-based prevention efforts in Sweden. Upcoming studies, resting on
a reliable multi-sourced data, will benefit the national fire service and
government departments in their efforts to understand and develop
prevention strategies and programs against fatal fires. While the
multi-sourced dataset is considerably more complete, there are also a
number of issues that need to be addressed in order to further improve

Table 4
Estimated numbers of fire fatalities in Sweden, 1999–2007, by type of event and degree of coverage/exclusive contribution from each source.

Residential fires Fires after traffic incidents Fires in parked cars Fires in person Other fires Total fires

Our total assessment 903 (77%) 115 (10%) 44 (4%) 34 (3%) 75 (6%) 1,171 (100%)
Unintentional 833 [92%] 100 [87%] 14 [32%] 3 [9%] 69 [92%] 1,019 [87%]
Intentional 70 [8%] 15 [13%] 30 [68%] 31 [91%] 6 [8%] 152 [13%]
Under-coverage
Dataset, fatal fires (X′) 82 (37%) 58 (26%) 25 (11%) 21 (9%) 38 (17%) 224 (100%)
Dataset, forensic examinations (Y′) 37 (61%) 0 (0%) 0 (0%) 1 (2%) 23 (38%) 61 (100%)
Dataset, cause of death registry (Z′) 19 (46%) 12 (29%) 1 (2%) 4 (10%) 5 (12%) 41 (100%)
Exclusive contribution
Dataset, fatal fires (X′) – – – – 1 1
Dataset, forensic examinations (Y′) 2 6 1 2 1 12
Dataset, cause of death registry (Z′) 12 – – – 20 32
 A. Jonsson et al. / Journal of Safety Research 55 (2015) 99–103
the data quality. Firstly, the data gathering process by MSB needs to be
performed in closer collaboration with police authorities, RMV and
SoS to establish a permanent sharing of data. Through co-operation,
the under-reporting is likely to be considerably less and by also includ-
ing police reports, for example, an improved description of the fire inci-
dent could be possible. Secondly, one of the largest problems with the
fatal fire database is the lack of registered PIN among many fatalities.
By ensuring that this information is included, as well as the fire service
code number on the incident report, the possibilities of conducting
deeper analyses in retrospective studies is greatly improved.
Furthermore, we would also suggest that an overview of the
inclusion criteria should be performed to also include people who
died of indirect causes of fire, for example deaths occurring during
evacuation or other related incidents (e.g., jumps from balconies, falls,
and falling building material). By broadening the inclusion criteria, a
better consistency with how a physician should classify underlying
cause of death would be achieved. According to WHO's guidelines, the
physician, when specifying the underlying cause of death, should state
the cause that initiated the chain of events and a similar viewpoint
may be relevant from the rescue service perspective. While the
proposed change in inclusion criteria does not solve the inconsistency
that exists for a fire in road accidents, as these should be (and usually
are) classified by the physician as traffic accidents, modified inclusion
criteria would improve the database on fatal fires. Lastly, compared to
other countries, differences exist regarding the cut-off time after the
fire (DCLG, 2012). In order to improve comparability and flexibility,
we would suggest increasing the cut-off time to 90 days after the fire
instead of the current 30 days.
5. Conclusions
In comparison to single sources, our multi-sourced assessment of
fire deaths in Sweden reveals both over- and under-coverage. While
the problem of over-coverage is minor and can be relatively easily
handled by carefully examining the data, entry by entry, the problem
of under-coverage is more complex to handle. We have shown a
method of assessing the number of fire fatalities through compiling
the database on fatal fires, the database on forensic examinations and
the cause of death register.
We have shown that previous regularly reported statistics from
Sweden using a single source has underestimated the problem. The
database on fatal fires underestimates the number by about 20% and
the cause of death register by 25%. By using a method of simultaneously
examining three different sources, a unique opportunity to assess how
many people actually die in fires, is possible. For a large proportion of
the deaths (75%), information could be obtained from all three sources,
producing a good basis for assessing that the inclusion criteria are met
and also an improved possibility of describing the deaths in more detail.
Acknowledgments
The authors' gratefully thank the Swedish Civil Contingencies
Agency for the financial support. Also, we would like to thank Marie
Ekman, The National Board of Forensic Medicine and Lars Age Johansson,
The National Board of Health and Welfare for their contributions. Finally,
we would to thank Colin McIntyre, Swedish Civil Contingencies Agency
for his contributions in the planning process and for providing language
help.
References
World Health Organization (WHO), Prentice, T., Beaglehole, R., Irwin, A (2004,). The World
Health report, 2004: Changing history. World Health Organization.
The World Health Organisation (WHO) (2015). WHO mortality database. (as of July 2014).
Retrieved from http://www.who.int/healthinfo/mortality_data/en/ accessed 2015–
02-11
103
Geneva Association (2014). World fire statistics bulletin no. 29, April 2014. Geneva: The
Geneva Association.
European Fire Safety Network (2009). Annual report 2009. not published.
Department for Communities and Local Government (DCLG) (2012). Comparison
of European fire statistics final report for the department for communities and local
government. Fire research report 1/2012. London, UK: Department for Communities
and Local Government.
Swedish Civil Contingencies Agency (2014). Räddningstjänst i siffror 2013 [Fire and Rescue
in Figures 2013]. (in Swedish).
The National Board of Health and Welfare (2014). Official statistics of Sweden, causes of
death 2013. in Swedish, summary in English.
Anderson, A., & Ezekoye, O. A. (2013). A comparative study assessing factors that
influence home fire casualties and fatalities using state fire incident data. Journal of
Fire Protection Engineering, 23(1), 51–75.
Barillo, D. J., Goode, R. (1996). Fire fatality study: Demographics of fire victims. Burns,
22(2), 85–88. doi:http://dx.doi.org/10.1016/0305-4179(95)00095-X.
Brennan, P. (1999). Victims and survivors in fatal residential building fires. Fire and
Materials, 23(6), 305–310.
Chao, A., Tsay, P. K., Lin, S., Shau, W., & Chao, D. (2001). The applications of capture-
recapture models to epidemiological data. Statistics in Medicine, 20(20), 3123–3157.
http://dx.doi.org/10.1002/sim.996.
Chernichko, L., Saunders, L. D., & Tough, S. (1993). Unintentional house fire deaths in
Alberta 1985–1990: A population study. Canadian Journal of Public Health, 84(5),
317–320.
Diekman, S. T., Ballesteros, M. F., Berger, L. R., Caraballo, R. S., & Kegler, S. R. (2008).
Ecological level analysis of the relationship between smoking and residential-fire mortality.
http://dx.doi.org/10.1136/ip.2007.017004.
DiGuiseppi, C., Roberts, I., & Li, L. (1998). Smoke alarm ownership and house fire death
rates in children. Journal of Epidemiology and Community Health, 52(11), 760–761.
Duncanson, M., Woodward, A., & Reid, P. (2002). Socioeconomic deprivation and fatal
unintentional domestic fire incidents in New Zealand 1993–1998. Fire Safety
Journal, 37(2), 165–179.
Harpur, A., Boyce, K., & McConnell, N. (2013). An investigation into the circumstances
surrounding fatal dwelling fires involving very young children. Fire Safety Journal,
61, 72–82.
Holborn, P. G., Nolan, P. F., Golt, J. (2003). An analysis of fatal unintentional dwelling fires
investigated by London fire brigade between 1996 and 2000. Fire Safety Journal,
38(1), 1–42. doi:http://dx.doi.org/10.1016/S0379-7112(02)00049-8.
Istre, G. R., McCoy, M., Carlin, D. K., & McClain, J. (2002). Residential fire related deaths and
injuries among children: Fireplay, smoke alarms, and prevention. Injury Prevention.
Journal of the International Society for Child and Adolescent Injury Prevention, 8(2),
128–132.
Jansson, B. (2005). Coding errors and underestimation of fall injury mortality. American
Journal of Public Health, 95(8), 1305–1306 author reply 1306. doi:AJPH.2005.067553.
Krishnan, P., Frew, Q., Green, A., Martin, R., & Dziewulski, P. (2013). Cause of death and
correlation with autopsy findings in burns patients. Burns, 39(4), 583–588.
Leth, P., Gregersen, M., & Sabroe, S. (1998). Fatal residential fire accidents in the
municipality of Copenhagen, 1991–1996. Preventive Medicine, 27(3), 444–451.
http://dx.doi.org/10.1006/pmed.1998.0295
Ludvigsson, J. F., Otterblad-Olausson, P., Pettersson, B. U., & Ekbom, A. (2009). The
Swedish personal identity number: Possibilities and pitfalls in healthcare and medical
research. European Journal of Epidemiology, 24(11), 659–667.
Marshall, S. W., Runyan, C. W., Bangdiwala, S. I., Linzer, M. A., Sacks, J. J., & Butts, J. D.
(1998). Fatal residential fires: Who dies and who survives? JAMA, 279(20),
1633–1637. http://dx.doi.org/10.1001/jama.279.20.1633.
Runyan, C. W., Bangdiwala, S. I., Linzer, M. A., Sacks, J. J., & Butts, J. (1992). Risk factors for
fatal residential fires. New England Journal of Medicine, 327(12), 859–863.
Viklund, Å., Björnstig, J., Larsson, M., & Björnstig, U. (2013). Car crash fatalities associated
with fire in Sweden. Traffic Injury Prevention, 14(8), 823–827.
Wilhelms, S. B., Huss, F. R., Granath, G., & Sjöberg, F. (2010). Assessment of incidence of
severe sepsis in Sweden using different ways of abstracting international classifica-
tion of diseases codes: Difficulties with methods and interpretation of results.
Critical Care Medicine, 38(6).
Yau, R., & Marshall, S. (2014). Association between fire-safe cigarette legislation and
residential fire deaths in the United States. Injury Epidemiology, 1(1), 10.
Anders Jonsson is a PhD student in Risk Management at Karlstad University. Part-time, he
is working as a statistician within the area of accident and injury analysis at the Lessons
Learned Section within the Swedish Civil Contingencies Agency.
Anders Bergqvist holds a Bachelor of Science (BSc) in Fire Protection Engineering from
Lund University. Currently, he is the CEO of the Swedish Fire Protection Association.
Ragnar Andersson has a combined background in engineering and public health. After
serving for fifteen years for the Swedish National Board of Occupational Safety and Health,
he took his PhD in social medicine at Karolinska Institutet, Sweden, in 1991 on occupational
injury prevention. Dr. Andersson is since 2001 holding a position as Professor of Risk
Management at Karlstad University, Sweden. His research is focused on accident and
injury analysis and prevention, injury surveillance, and macro-level determinants of risk
and safety. He is also an experienced teacher and supervisor, and publishes regularly with
his students and colleagues.