I N J U R I E S
The impact of structural factors on the
injury rate in different European countries
KAR1N A. MELINDER, RAGNAR ANDERSSON •
Background: A previous study pointed to there being two kinds of injuries - those with a mainly social genesis and
those with a mainly environmental genesis. The aim of this study was to analyse how socioeconomic factors - such
as level of economic development, alcohol consumption and unemployment and more cultural factors - such as
education and religion - relate to kinds of injury. Methods: Motor vehicle traffic accidents were chosen to represent
injuries with a predominantly environmental genesis and suicides those with a mainly social genesis. Qualitative
comparative analysis (QCA) complemented by Pearson correlation was employed. The data come from 12 European
countries. Results: Four groups of countries emerged from the analysis. Group 1 was high on both kinds of injuries
and was also high on all the independent variables considered. Group 2 was low on social injuries and high on
environmental injuries; it had a low level of economic development, high alcohol consumption and a high proportion
of Roman Catholics. Group 3 was high on social injuries and low on environmental injuries; it had a high level of
economic development, low alcohol consumption and few Roman Catholics. Group 4 was low on both kinds of
injuries; the independent variables formed a similar pattern to those of group 3. Conclusion: The pattern for traffic
fatalities differs from that of suicides. There is also patterning with regard to structural factors; economic level,
education and religion seem to be more important with regard to injury rate differentials than alcohol consumption
or unemployment
Keywords: Europe, injury, social phenomenon, suicide, traffic accident
he current study stemmed from an earlier investigation 1
of trends in injury mortality in the Nordic countries. The
earlier study suggested a possible division of injuries into
two main causal categories: those with a predominantly
social genesis and those that are largely related to
technical and environmental haiards. Fatal injuries with
a predominantly social genesis were considered to include
suicides, homicides and accidental alcohol intoxications,
whereas examples of injuries with a primarily environmental cause included fire- and transport-related
accidents and drownings. On adopting this typology,
environment-related injury mortality was found to decline substantially in all Nordic countries, whereas that
for the more socially oriented category appeared stable
over time or even on the increase. Moreover, each of the
four Scandinavian countries was found to occupy a
category of its own with regard to the relative incidence
of the two kinds of injuries: Finland was high on both
kinds of injuries, Denmark was high on injuries with a
social genesis, Norway was high on environment-related
injuries and Sweden was low on both kinds of injuries.
* ICA. Me|]nder u , R. Andernon'
1 Karollraka Irotitutet, Department of Public Health Sciences,
Division of Social Medicine, Sweden
2 Sweden's National Institute of Publk Health, Stockholm. Sweden
Correspondence: Karin A. Mellnder, M.Sc, National Institute of
Public Health, SE-103 52 Stockholm, Sweden, tel. +46 8 56613542,
fax+46 8 56613601
These findings inspired the current study, which explored
possible structural determinants underlying inj uries in the
two categories among a larger number of European countries. The factors chosen for investigation encompassed
both socioeconomic variables (which are already known
to have an impact on injury rates) and more culturally
oriented ones (namely religion and education).
BACKGROUND
There is a varying amount of research - according to the
variables considered - on cultural, social and economic
factors, such as alcohol consumption, religion, per capita
income, unemployment and educational level and their
influence on injuries.2"8 There is also often a difference
in results between studies performed at individual and
aggregate levels,9'10 that is a particular factor might be
associated with injury to an individual, but the association
might not apply at aggregate level (and vice versa). One
example is that, at an individual level, unemployment is
associated with suicide, i.e. people who are unemployed
commit suicide more often than other people. At
aggregate level, however, countries with a high unemployment rate do not show higher suicide rates than
other countries.7'8
Suicide has been much researched. Indeed, Durkheim's1'
pioneering sociological study from 1897 was concerned
specifically with suicide. The study has been criticised on
methodological grounds, but its main conclusion - that
EUROPEAN JOURNAL OF PUBLIC HEALTH VOL. 11 2001 NO. 3
suicide is associated with low social integration - still has
some support.2 Although Durkheim's11 failure to find an
association between suicide and alcohol consumption has
been much debated, it should be remembered that most
studies of alcohol in relation to suicide have been
performed at individual level.3 We take each of our
independent variables in turn.
Alcohol consumption
A relationship was established at aggregate level between
alcohol consumption and suicide in Norwegian and
Swedish samples.12 However, this relationship was found
to be weaker in France and Portugal. Norstrom
suggested that the higher the per capita consumption in
a country, the weaker the impact of alcohol consumption
on suicide. Observations such as these have given rise to
the concepts of 'wet' and 'dry' cultures,3 the former being
defined as those with a high consumption of alcohol
(integrated into everyday living) and the latter as those
with a relatively low and sporadic consumption (only on
special occasions). In the case of Portugal, regional analysis has pointed to the importance of factors other than
alcohol consumption to the suicide rate, namely - in
accordance with Durkheim's11 theses - social integration
and religion.2
There are a vast number of studies on alcohol and traffic
safety showing - at least at individual level - that people
who drive when drunk are more likely to be involved in
road accidents.'' Knowledge of this association has
generated various traffic safety and preventive strategies
in many countries. In the Nordic countries, drink-driving
was seen as a problem at a very early stage and legislation
came into force in parallel with the arrival of the motor
vehicle. In some other countries, the connection between
drinking and motor vehicle accidents was acknowledged
much later.14'15
Religion
Associations have been found between religion and
suicide, but these have prompted questions concerning
the reliability of statistical records on suicide. For
example, certain religions, such as Roman Catholicism,
regard suicide as a sin. For this reason it has been suggested
that, within cultures where Roman Catholicism is the
predominant religion, death by suicide may not always be
recorded as such. More recently, however, religious belief
in itself has been shown to be associated with low
tolerance of suicide.5 Religion has also been found to be
associated with various health indicators on a more
general level. Although Sloan et al. questioned the
association between religion and health in a recent
review, their study had a specific individual-ethical focus
where the goal was to hinder physicians from promoting
religious faith in their patients.
Education
An association between level of education and injury
^ ^ incidence has been found at an individual level,'7 but has
HIM not been investigated at aggregate level.
Unemployment
The relations between unemployment and different kinds
of health problems were first recognised by Durkheim
as early as in 1897. However, studies of unemployment
and suicide show different pictures at individual and
aggregate levels, and there are even greater discrepancies
in results between spatial and time-series analyses.
Most individual-level studies, both cross-sectional and
longitudinal, point to an association between suicide and
unemployment. However, spatial, cross-sectional
aggregate studies do not show a higher incidence of
suicide in areas of high unemployment. On the contrary,
in areas of low unemployment, it seems that people who
become jobless display poorer health-related behaviour
than their counterparts in areas with a high unemployment rate. >°
In contrast, motor vehicle accidents have been found to
be inversely related to the unemployment rate - although
only when number of miles driven is not taken into
ID
account. °
Economic factors
With regard to traffic-related mortality and economic
factors there has been a change in view. Traffic-related
accidents were previously regarded as a problem of
industrialised countries, but it has now been shown that
there is a clear negative relation between a country's
economic status and accident incidence, particularly
when the number of motor vehicles is taken into
account.
Methodological concerns
Here, we consider two kinds of injuries - motor vehicle
traffic accidents and suicide - as representing inj uries with
environmental and social origins respectively.
Associations between injury types and various independent factors have been found, but there is not always
a consistent pattern. For example, relationships have
often been uncovered at an individual level but not at an
aggregate level (and vice versa). This holds for suicide and
alcohol and also for suicide and unemployment. Most
studies have simply taken one variable at a time (such as
religion or unemployment).
An exception worth mentioning in this context is a study
by Giesbrecht and Dick.19 They performed a regression
analysis relating what they called 'casualty type' (a particular kind of injury) to alcohol consumption in different
countries. No clear relations were found. In some countries associations were found for some casualty types but
not for others. In conclusion, the authors stated that 'in
order to attempt to explain the relationship it may be
necessary to use a systems perspective in order to examine
other factors or systems that may influence the relationship between alcohol consumption and casualties'
(Giesbrecht and Dick19 p. 875).
AIM
The aim of this paper was to analyse how socioeconomic
factors, such as per capita income, unemployment and
Structural factors and injuries
Structural
Determinants
Cultural
;
.
^ _ ^
s^~—'
. x
'
X
"
I
Social
Proximal
( Environ men talQ j
Determinants
^
—~
(individual level)
\
,/ /
~^^
J)
——
Suicide,
) Homicide,
Child.^—^^jyuries^— - ^ Alcohol into* ,
etc.
etc.
WorL ^~^
Figure 1 Model of the relationships between structural and proximal
(individual'level) determinants of injuries
alcohol consumption (which have been found to have an
impact on traffic fatalities and suicides in other studies)
are related to injury mortality and also to each other at
structural level. As well as these socioeconomic factors,
two culture-oriented factors were considered: religion and
education. Traffic injuries were chosen as the indicator
for environment-related injury and suicides as that for its
socially related counterpart.
METHODS AND MATERIALS
Conceptual frame of reference
As just stated, traffic fatalities and suicides are seen as
representative of injuries with primarily environmental
and social origins respectively. The distinction between
these two kinds of injunes is built on an individual
perspective, i.e. on consideration of the proximal
determinants of a specific individual encountering an
injury. Such proximal determinants can include anything
from a physical environment with many water hazards
(e.g. lakes) to a social problem faced by individual
determinants. The structural determinants that we investigated in this study operate at an aggregate level and
are both socioeconomic and cultural by nature.
The relations envisaged between the various variables
considered in this study can be incorporated into the
model illustrated in figure I.
As mentioned in the Background section, it cannot be
assumed that a relation that holds at aggregate level will
be replicated at individual level.
Methods
the various combinations of independent variables and
their links with the outcome under consideration (here,
death due to injury). According to the degrees of freedom
involved, the number of possible combinations can be
high. Generally, however, not all combinations are used
in practice and many can therefore be eliminated. It is
also possible that two or more variables show the same
pattern and can be combined into one.
In this analysis, QCA was complemented by Pearson
correlation. Pearson correlation was chosen instead of
partial correlation because it offers the opportunity of
relating all the variables investigated to one another.
The variables commonly employed in QCA are nominal
scale. Here, however, we used dichotomised interval scale
data. Each variable was allocated two categories and two
groups were created. The dividing line between groups
was set where it was most meaningful in relation to the
purpose of the study, i.e. where the groups were distinct
in a relevant respect (see Ragin's20 account of how QCA
functions).
Material
The injury information came from the Health for All
(HFA) database25 which contains both mortality
statistics and lifestyle and social indicator data. Agestandardised death rates (SDRs) per 100,000 inhabitants
for motor vehicle fatalities and suicides/fatal self-inflicted
injuries were employed as dependent variables. SDRs
were calculated directly from figures for the standard
population of the European countries in question. The
independent variables which were taken from the HFA
database were gross national product (GNP) in US dollars
per capita, unemployment rate in percent, years of
schooling and annual pure alcohol consumption (litres
per person aged 15 years and above). Information on
religion by country was obtained from Microsoft
Encarta 97.
In a previous validation of cause-of-death statistics
regarding injunes in the Nordic countries, it was
recommended that when suicides are compared between
countries, cases in the injury category 'undetermined
whether accidentally or purposely inflicted' should be
added to suicide. Raw figures in this category, which are
not included in the HFA, are covered by the World
Health Organisation' s (WHO's) mortality data. WHO
figures were used to establish the ratio between suicides
and undetermined cases in each country, thus providing
coefficients (i.e. the ratio of undetermined cases to
suicides + 1) for computation of SDRs for suicides
including undetermined cases.
Analysis was performed for the year 1990 for the 12
Western European countries so as to enable comparison
of results with those of earlier analyses of cultural values
in relation to suicides and fatal motor vehicle accidents.
Selecting a method for studying many variables at one and
the same time is not always straightforward. In this case,
a rather new technique, qualitative comparative analysis
(QCA), was the primary method chosen. (It is discussed
later in relation to more traditional methods.) The aim of
QCA is to svnthesise quantitative and qualitative approaches.20~24The analysis presented here is a spatial one.
QCA employs binary data, in the first instance - in this RESULTS
case - by assigning a country a value of 1 or 0 according
The values of the independent variables and age-SDRs
to whether or not it meets certain criteria. The next step
for the independent variables are shown by country in
is to arrange the data in a 'truth table' which summarises
table I.
EUROPEANJOURNALOFPUBLICHEALTHVOL.il 2001 NO. 3
variables (columns 5-8) in four categories. Each group is
taken in turn.
Group 1, which is high on both traffic fatalities and
suicides, contains Belgium and France which have
identical patterns with regard to their dichotomised
independent variable scores. Both Belgium and France
face potential problems in the forms of high unemploy-
Dichotomisation of the scores gave the pattern by country
shown in table 2.
A truth table - where the dependent variables, i.e. suicides and motor vehicle fatalities, are related to the independent variables - is displayed in table 3.
The truth table shows a pattern where the dependent
variables (columns 2—4) associate with die independent
Table 1 Mean years of schooling, unemployment rate, GNP, Roman Cadiolic population and annual pure alcohol consumption, in
relation to SDRs for motor vehicle traffic accidents, SDRs for suicides and self-inflicted injuries and suicides adjusted for undetermined
cases in selected European countnes in 1990
Country
Belgium
Denmark
Finland
France
Ireland
Italy
The Netherlands
Norway
Portugal
Spam
Sweden
GNP
(US dollars
per capita)
15,440
22,090
26,070
19,480
9350
18,709
17330
23,120
4,890
Annual pure
alcohol
Mean years
of
consumption' Unemployed
schooling
(1)
(%)
9.9
9.60
10.7
9.70
9.7
10.4
7.7
10.6
11.6
12.6
3.40
8.90
8.7
7.3
7.6
9.2
17.20
11.00
10.6
8.1
5.00
4.1
10.1
10.8
4J0
4.70
11.6
6.0
6.8
11.1
16.30
1.60
5.90
Roman
Catholic
(%)
SDRs for
motor
vehicle
traffic
acccidents
SDRs for
suicides
and
self-inflicted
injuries
SDRs for
suicides
adjusted for
undetermined
cases
84
17.16
<10
<10
10.86
17.53
22.43
19.98
26.92
12.33
16.83
29.09
18.95
33.74
13.23
14.36
10.67
6.80
8.07
7.43
26.92
15.14
8.51
76
94
84
33
<10
9.28
22.74
12.16
6.80
10.21
15.59
19.31
7.10
19.57
7.81
<10
20
8.16
9.06
15.88
7.79
21.44
11.69
94
97
UK
10,920
23,680
16,070
Mean
Median
17,577
18,709
9.7
10.6
8.2
7.65
5.90
-40
8.1
33
13.03
12.33
13.98
15.40
17.38
17.58
Lower group limit
for 1
15,440
10.40
9.2
8.90
76
13.23
15.14
19.57
11.5
5.5
7.8
a: Per person for those 215 years,
b: Per 100,000 Inhabitants.
Table 2 Dichotomised scores by country for mean years of schooling, unemployment rate, GNP, Roman Catholic population and annual
pure alcohol consumption, in relation to SDRs for motor vehicle traffic accidents, SDRs for suicides and self-inflicted injuries and suicides
adjusted for undetermined cases in selected European countries in 1990
Country
Belgium
Denmark
Finland
France
Ireland
Italy
GNP
(US dollars
per capita)
1
1
1
1
0
1
Annual pure
alcohol
Mean years
consumption' Unemployed
of
schooling
(1)
(%)
1
1
1
1
1
1
1
0
0
1
1
1
1
0
0
1
1
0
The Netherlands
Norway
1
1
1
1
Portugal
Spain
0
0
0
0
0
1
1
Sweden
UK
1
1
1
1
0
0
a: Per person for rhose ^15 years.
b: Per 100,000 inhabitants.
0
0
0
0
1
0
0
Roman
Catholic
(%)
1
SDRs for
motor
vehicle
traffic
acccidents
1
0
0
1
1
0
1
0
1
0
0
1
1
0
1
0
1
1
0
1
0
0
0
SDRs for
suicides
and
self-inflicted
injuries
1
1
1
1
SDRs for
suicides
adjusted for
undetermined
cases
1
1
1
1
0
0
0
0
1
0
0
1
0
1
0
1
0
0
0
0
Structured factors and injuries
ment and high alcohol consumption and are also high on
Of the independent variables, the proportion of Roman
what can be regarded as 'potentiality goods', i.e. length of
Catholics showed die strongest associations, being
schooling and GNP. They are also largely Roman
significantly correlated with all variables except suicides
Catholic countries (but not as predominantly so as
adjusted for undetermined cases (positively associated
Portugal, Spain or Ireland).
with motor vehicle traffic accidents, alcohol consumption and unemployment and negatively associated with
Group 2 which is high on traffic fatalities and low on
schooling and GNP). GNP was significantly correlated
suicides, generally has a pattern of a short period of
wirli mean years of schooling (positively) and both the
schooling, a high unemployment rate, a low GNP, high
proportion of Roman Catholics and unemployment
proportion of Catholics and high alcohol consumption.
(negatively), as well as with traffic accidents (negatively)
The exceptions are Ireland, with a relatively low alcohol
and suicides (positively). Schooling was positively
consumption, Portugal with a relatively low unemploycorrelated wirii borli GNP and suicides and negatively
ment rate and Italy with a high GNP.
correlated with bor_h motor traffic accidents and die proGroup 3, which is low on traffic fatalities and high on
portion of Roman Carliolics. Alcohol consumption was
suicides, shows the opposite pattern to group 2. These
positively and significantly correlated with traffic
countries generally have a lengthy period of schooling, a
accidents and the proportion of Roman Catholics and
low unemployment rate, a high GNP, few Roman
unemployment positively correlated with the proportion
Catholics and low alcohol consumption. The exception
of Roman Catholics and negatively correlated with GNP.
is Denmark which currently has high unemployment and
Suicides adjusted for undetermined cases were signia tradition of high alcohol consumption.
ficantly associated with unadjusted suicides only.
Group 4, which is low on both traffic fatalities and
With regard to the alcohol consumption variable, acsuicides, has low scores on all the independent variables
count should be taken of unrecorded consumption. In this
except schooling and GNP.
study, a check was made against unrecorded consumption
After the suicide rate was adjusted to include unestimates for die early to mid-1990s.27 Adjusting for undetermined cases, two countries changed value in the
recorded consumption increased levels in the Nordic
truth table. Portugal moved from having a low rate to a
countries,
but not to an extent diat changed any of die
high rate, whereas Norway shifted from a high rate to a
intercountry
relationships.
low rate.
The Pearson correlation coefficients are shown in table 4.
Table 4 shows that there are positive significant associ- DISCUSSION
ations between motor vehicle fatalities and both alcohol
Four groups emerged from the analysis. Group 1 had high
consumption and die proportion of Roman Catholics and
scores on all variables, many injuries of bodi kinds and
significant negative associations between motor vehicle
high levels on determining factors (both potentially good
fatalities and both GNP and mean years of schooling.
and potentially bad). Group 2 consisted of Catholic counSuicides show significant positive associations with GNP
tries witii many traffic accidents, few suicides, a low GNP,
and schooling and a significant negative association with
few mean years of schooling, high alcohol consumption
die proportion of Roman Catholics.
and a high unemployment rate. Groups 3 and 4 had
Table 3 Truth table for mean years of schooling, unemployment rate, GNP, Roman Catholic population and annual pure alcohol
consumption, in relation to SDRs for motor vehicle traffic accidents, SDRs for suicides and self-inflicted injuries and suicides adjusted for
undetermined cases in selected European countries in 1990
SDRs for
motor
vehicle
traffic
acccidents"
I
1
SDRs for
suicides
and
self-inflicted
injuries'
1
1
SDRs for
suicides
adjusted for
undetermined
cases
1
1
0
0
Portugal
Spain
Denmark
1
1
1
0
0
0
1
0
0
1
0
0
1
0
1
0
1
0
1
Finland and
Sweden
0
1
1
Norway
The Netherlands
0
0
1
0
0
0
1
1
1
UK
0
0
0
1
Country
Belgium
France
Ireland
Italy
a. Per 100,000 inhabitants.
b: Per person for those £15 years
GNP
(US dollars
per capita)
1
1
0
1
Annual pure
Mean years
alcohol
consumption Unemployed
of
schooling
(%)
(1)
1
1
1
1
1
1
1
0
0
1
1
0
Roman
Catholic
(%)
1
1
1
0
1
1
0
1
1
1
0
0
0
0
0
0
0
0
0
0
0
1
0
EUROPEAN JOURNAL OF PUBLIC HEALTH VOL. 11 2001 NO. 3
similar configurations of independent variables, e.g. a
high GNP and schooling and generally low levels on the
other vanables. The difference between the two was that
group 4 had low rates for both kinds of injuries, whereas
group 3 had relatively many suicides.
The main strength of the study is that more variables were
taken into account than has usually been the case
previously. Further, an effort was made to compensate for
uncertainty in suicide diagnoses. Methodologically, an
advantage of the study is that it provided a trial for a rather
new analytic method, QCA.
According to Coverdill et al, the principal advantage of
QCA lies in its emphasis on causal complexity and the
detailed features of cases, which compel the researcher to
consider the context in which specific causal relationships may hold. We regarded these features as well suited
to our desire to employ a method capable of identifying
complicated patterns in a data set.
QCA is usually employed with nominal scale variables,
in which case quantitative analysis is not productive.
Here, however, we did have quantitative data available
and could have chosen to use quantitative methods.
Although we had this opportunity, we found that QCA
still provided information that would otherwise not have
been generated. The zeros and ones in the trudi table may
give an impression of being rudimentary, but they make
it possible to capture more cases and variables at one and
the same time.
Although our choice of correlation as a complement was
initially prompted by a certain uneasiness with a new
method, we found that it helped in understanding how
the patterns identified by QCA arose. The truth table
generated by QCA showed a pattern. The computation
of Pearson correlation coefficients for the variables
demonstrated that some variables had a greater impact
than others. This helped us deal widi one of the major
disadvantages associated widi QCA, namely a failure
to provide information on which factors are most
important. Autocorrelation, which is regarded as a
problem in many statistical studies, is turned into an
advantage by QCA - although this also means that QCA
results are intrinsically uncertain with regard to causality.
Here, QCA was employed more as a 'driver' than a 'tester'
of theory, which is in accordance with how QCA is
envisaged to function.
Our decision to set die dividing lines for variable
dichotomisation at die values where die most meaningful
groups were created can of course be questioned. However, we found it preferable to using die mean or the
median, which results in unclear cases close to either one
of these measures. This emphasises die importance of
sampling. When groups are formed in die manner we
determined, outliers will do no harm. On the odier hand,
the patterns found cannot be easily extrapolated.
The study had an ecological design, but - since no
inferences were made from aggregate to individual level there was no nsk of ecological fallacy.
No contradictions were found between die results of tiiis
study and die findings of riiose we referred to as background. In diis sense, no novel findings are presented.
Nevertheless, in our view, die results of die current study
make diose of odiers more comprehensible. Further, two
rather seldom employed variables in diis context, i.e.
religion and education, were considered.
Table 4 Pearson correlation coefficients for mean years of schooling, unemployment rate, GNP, Roman Catholic population and annual
pure alcohol consumption in relation to SDRs for motor vehicle traffic accidents, SDRs for suicides and self-inflicted injuries and suicides
adjusted for undetermined cases in selected European countries in 1990
SDRs for
SDRs for
motor
suicides
vehicle
and
traffic
self-inflicted
acccidents0
injuries"
SDRs for motor
vehicle traffic
accidents"
SDRs for suicides and
self-inflicted injuries"
SDRs for suicides
adjusted for
undetermined cases
GNP (US dollars
per capita)
Mean years of
schooling
Annual pure alcohol
consumption (1)
Unemployment (%)
Roman Catholic (%)
SDRs for
suicides
adjusted for
undetermined
cases
GNP
(US dollars
per capita)
Annual pure
Mean years
alcohol
of
Unemployed consumption
schooling
(%)
(1)
1.0
-0.211
0.005
-0.718^
-O.735 d
0.307
-0.220
1.0
0.929 d
0.677*
0.533 c
-0.329
-0.044
-0.502*
0.005
0.92^
1.0
0.488
0.388
-0.456
0.026
-0.444
-0.718**
O.677*1
0.488
1.0
0.716^
-0.500*
-0.406
-0.790^
-O.735 d
0.533 c
0.388
0.716 d
1.0
-0.462
-OJ58
-O.708 d
0.026
-0.456
-0.4O6
-0.500*
-0.358
-0.462
0.455
1.0
0.629*
1.0
0.455
0.691 d
-0.444
1
0.629*
1.0
0.685 d
0.307
0.784 d
-0.044
-0.329
-O.5O2
c
a: Per 100,000 inhabitants.
b. Per petson for those 215 years.
c- Correlation significant at the 0.05 level (two-tailed test).
d: Correlation significant at the 0.01 level (two-tailed test)
-0.79O"
-O.7O8
d
0.69 l
0.685 d
Roman
Catholic
(%)
d
0.784 d
Structural factors and injuries
Of interest regarding the idea that GNP and religion are
Traffic fatalities showed a rather expected pattern in that
the most influential variables is die relationship found
they were positively associated with alcohol consumption
previously between alcohol consumption and traffic
and negatively associated with GNP. Suicide however
accidents which is strong at the individual level.8 The
seemed to be a more complex phenomenon. There was a
case of Denmark which, despite a high alcohol consumppattern in the trudi table, but there were few associations
tion, belongs to the group with a low rate of traffic
between variables.
accidents (group 3) may indicate that GNP/schooling
A similar point can be made concerning alcohol con(taken as one) is an important mediator of the relationsumption and unemployment. Both have been subject to
ship between alcohol and traffic accidents. Indeed, GNP
considerable research in connection with injuries, but
and schooling have very similar relationships with die
they showed the fewest significant associations with
other variables. Whether schooling has a role of its own
outcome from among the independent variables
to play, which is distinct from GNP is a question for future
(although they do fit into the general pattern). The
research.
independent vanables that seemed to have the greatest
impact were the percentage of Roman Catholics and
Traffic accidents and suicides were tentatively chosen as
GNP. They also had a strong negative correlation with
representative of injuries with primarily environmental
each other.
and social origins respectively. At first glance, diis does
not seem to hold, as a relationship between alcohol conThe importance of religion and, in particular, the
sumption (a social factor) and traffic accidents (environnegative association found between the proportion of
mental injuries) has been found. However, die different
Catholics and GNP might be related to Weber's28 classic
roles played by alcohol in 'wet' and 'dry' countries must
idea that the spirit of capitalism is rooted in Protestantbe taken into account. Wet countries have a high alcohol
ism. Although Weber's28 book was first published in 1904,
consumption but one that is integrated into society. Such
it still retains its relevance today.
countries have many traffic accidents and few suicides and
No significant relationship was detected between alcohol
are also largely Roman Catholic. Wheriier religion plays
consumption and suicide in this study; according to the
a role in relation to the way alcohol functions in these
truth table, high alcohol consumption could go together
countries needs to be studied further.
with either a high or a low suicide rate. Nevertheless, a
significant relationship has been found earlier between
alcohol consumption and suicide.2|12'1-'The studies con- CONCLUSION
cerned were primarily conducted in die Nordic countries,
A relatively new technique, QCA, complemented by
which have relatively high suicide rates and relatively low
Pearson correlation, was employed in order to investigate
alcohol consumption. In addition, comparisons have
the impact of various factors on environmentally and
been made with France and Portugal, countries that do
socially related injuries. It was shown that traffic fatalities
not show the same relation between suicide and alcohol
show a different pattern from suicides. It was also possible
as some other Catholic countries.
to detect patterning among different variables. Of the
independent vanables, GNP, schooling and religion apWith regard to unemployment it has already been repear to be more important at an aggregate level dian
ported that there is no clear relationship between
alcohol consumption and unemployment. The roles of
unemployment and suicide at aggregate level; rather, the
die latter seem to vary according to country according to
nature of the relationship seems to be dependent on
their GNP, schooling practices and predominant religion.
contextual factors. The information provided by the truth
table presented here suggests that countries with a high
The research was supported by the National Institute of Public
level of unemployment have many traffic fatalities, but
Health in Sweden.
many suicides only when economic and educational
An earlier version of this paper was presented at the Third Nordic
levels are high.
Safe Community Conference, Reykjavik, Iceland, 25-28 August
Our findings make it reasonable to suppose tentatively
1999.
that GNP and religion (proportion of Catholics) are the
most influential factors with regard to injury and interpret
the roles played by alcohol and unemployment more as a
1
Melinder KA, Andersson R. Stable and dynamic changes
consequence of GNP and religion.
in injury mortality in the Nordic countries. What do they say
Portugal and Norway fell into different groups when the
about inherent national characteristics with regard to risk? Scand
J Public Health 2000,28:16-22.
suicide rates were adjusted to include undetermined cases
2
Skog OJ, Teixeira Z, Barrias J, Moreira R. Alcohol and
of injury. In the case of Norway, this did not change the
suicide: the Portuguese experience. Addiction 1995;90". 1053-61.
overall patterning, since groups 3 and 4 were formed by
3
Rossow I. Alcohol and suicide: beyond the link at the
individual level. Addiction 1996;91:1413-6.
the same dependent variable values. Portugal, however,
4
Evans L The fraction of traffic fatalities attributable t o
showed a different dependent variable pattern to other
alcohol. Accident Anal Prevent 1990;22:587-602.
members of group 1 when the suicide rate was adjusted.
5
Neeleman J, Halpern D, Leon D, Lewis G. Tolerance of
suicide, religion and suicide rates; an ecological and individual
Whether this was because there is something specific
study in 19 Western countries. Psychol Med 1997,27:1165-71.
about Portugal or whether the procedure adopted was a
6
Levin JS. Religion and health: is there an association, is it
poor way of validating suicides cannot be determined
valid, and is it causal? Soc Sci Med 1994;38:1475-82.
widiout further research.
7
Jin RL, Chandrakant PS, Tomislav JS. The impact of
EUROPEANJOURNALOFPUBLICHEALTHVOL.il 2001 NO. 3
unemployment on hearth: a review of the evidence. Can Med
Assoc 1995;153:529-40.
8
Platt S. Unemployment and suicidal behaviour: a review
of the literature. Soc Sci Med 1984;19:93-115.
9
Robinson WS. Ecological correlations and the behavior of
individuals. Am Sociol Rev 195O;15:351-7. In: Lipset SM, Smelseer,
editors. Sociology: the progress of a decade. Englewood Cliffs:
Prentice-Hall, Inc., 1961:145-51.
10
Scheucti EK. Cross-national comparisons using aggregate
data: some substantive and methodological problems. In:
Merrit RL, Rokkan S, editors. Comparing nations: the use of
quantitative data in cross-national research. New Haven and
London: Yale University Press, 1966:131-67.
11
Durkheim E. Suicide: a study in sociology. Glencoe:
Free Press, 1958.
12
Norstrom T. Alcohol and suicide in Scandinavia. Br J
Addict 1988;83:553-9.
13
Norstr6m T. Alcohol and suicide: a comparative analysis
of France and Sweden. Addiction 1995;90:1463-9.
14
Petridou E, Trichopoulos D, Sotirou A, et al. Relative
and population attributable risk of traffic injuries in relation to
blood alcohol levels in a Mediterranean country. Alcohol
1998,33:502-8.
15
Alvarez FJ, del Rio MC, Prada R. Drinking and driving in
Spain. J Stud Alcohol 1995;56:403-7.
16
Sloan RP, Bagiella E, Powell T. Religion, spirituality, and
medicine. Lancet 1999,353:664-7.
17
Murray A. The home and school independent of young
drivers involved in traffic accidents. Accident Anal Prevent
1998,30:169-82.
18
Soderlund N, Zwi AB. Traffic-related mortality in
industrialized and less developed countries. Bull WHO
1995:73:175-82.
19
Giesbrecht N, Dick R. Societal norms and risk-taking
behaviour, inter-cultural comparisons of casualties and alcohol
consumption. Addiction 1993,88:867-76.
20
Ragin C The comparative method, moving beyond
qualitative and quantitative strategies. Berkeley: University of
California Press, 1987.
21
Coverdill J, Finlay W, Martin JM. Labor management in
the southern textile industry: comparing qualitative, quantitative,
and qualitative comparative analyses. Sociol Methods Res
1994;23(1):54-85.
22
Griffin L, Ragin CC Some observations on formal methods
of qualitative analysis. Sociol Methods Res 1994;23(1):4-21.
23
Hicks A. Qualitative comparative analysis and analytical
induction: the case of the emergence of the social security state.
Sociol Methods Res 1994;23(1):86-113.
24
Amenta W, Poulsen JD. Where to begin: a survey of five
approaches to selecting independent variables for qualitative
comparative analysis. Sociol Methods Res 1994;23(1):22-53.
25
Health for all. Statistical database. Copenhagen- WHO
Regional Office for Europe, 1998.
26
Melinder KA, Andersson R. Differences in injury mortality
between the Nordic countries: with special reference to
differences in coding practices. Scand J Soc Med 1998;26(3): 190-7.
27
Harkin AM, Anderson P, Goos C. Smoking, drinking and
drug taking in the European Region. Copenhagen: World Health
Organisation Europe, 1997.
28
Weber M. The Protestant ethic and spirit of capitalism.
London: Unwin Paperbacks, 1904/1985.
29
Gerhards J. Religion and the spirit of capitalism, a
comparison of attitudes towards work and the economic order
in the USA and Spain. Berliner J Soziol 1996;4:541-9.
Received 26 October (999, accepted 15 May 2000