Post on 22-Dec-2021
transcript
EUROPEAN COMMISSION EUROSTAT Directorate F: Social statistics Unit F-5: Education, health and social protection
Doc ESTAT/F5/ESAW/2021/4
Underreporting study
Item 4 of the agenda
Meeting of the Working Group
European Statistics on Accidents at Work (ESAW)
Virtual meeting
Luxembourg, 20 October 2021
Methodological study on under-reporting of occupational accidents in European Union
Abstract
3 Methodological study on under-reporting of occupational accidents in European Union
Abstract: Background: Occupational accidents or occupational injuries caused by accidents at work have
been target for prevention for well over 100 years. Fatal and serious accidents have been better
covered while less serious accidents may not be properly notified, reported, recorded and
compensated. Fatal accidents are clearly important as they cause a high number of years of lost life
and lower the life expectancy; the average age of a victim is in the range of 30–35 years (1).
However, the sheer high number of non-fatal cases is a cause of shorter or longer disability to work
and often considerably high number of losses to both the individual and the society and leads to
permanent reduction of workability as measured by Disability Adjusted Life Years. This is important
in terms of both ethical and economic losses to workers themselves, employers and the society as a
whole. A fatal accident is less frequent however; thousands of less serious injuries and incidents
usually precede it. These are good indicators for identifying risks and for the prevention of both fatal
and non-fatal accidents.
Objective: The objective of this study was to identify the level of under-reporting of non-fatal
accidents based on the assumption of better reported/recorded fatal cases in the European Union.
Based on existing numerical evidence a methodology could be established to work out a comparable
and adjustable mechanism for defining the magnitude and the number of non-fatal accidents in EU
Member States. This would be particularly interesting in the attempt of reaching the objective of the
study of providing a level of the existing underreporting.
Methods: Depending on several factors, such as the economic sector, gender, age and cultures in
EU Member States the reporting is varying widely at overall level and at sectoral fatal/non-fatal rates:
from 1 fatal/2 000 non-fatal reported cases to 1 fatal/10 non-fatal and an average EU rate in the
range of 1/1250. This wide range is reflected also in various sectors such as construction,
manufacturing and services. The particularities and characteristics of the sectors are reflected by
varying rates of fatal/non-fatal accidents. Further, certain countries have been selected as
benchmark countries and priority variables have been selected and have been used in consecutive
imputations with the purpose of replacing poor reported values until observed changes were
negligible.
Results: While there are genuine differences between EU Member States in reporting of fatal
accidents in overall and in different economic sectors, the adjusted numbers for less reported non-
fatal cases in EU are in reality much closer to the best reporting countries. Differences result from
selected technical processes, tools, methods of work, level of mechanization, automation, and the
level of labour intensive jobs in Member States. Administrative reporting practices from workplaces
are equally different. According to the available Eurostat data, we arrived to the result that the overall
number of non-fatal accidents goes from a range of reported 2.4 million cases to 6.9 million of such
injuries per year in question. Individual country under-reporting varies even more drastically. This
would have a major impact on the priorities for prevention and possibly could have significant effects
to insurance and social support systems for victims of occupational accidents. The relatively rare
fatal accidents will not provide a sufficient base of knowledge for heading for the long-term ‘vision
zero’ and ‘zero harm’ efforts.
Discussion: A way to check the level of reporting compared to the adjusted numbers could also be
the 2020 ad-hoc module of the labour force survey ‘Accidents at work and work related health
problems’ (2) even though the classification between formal reporting and ad-hoc labour-force
surveys are not entirely comparable. The average level and number of non-fatal injury cases in
relation to fatal ones tend to raise when economies are shifting from production based to service-
(1) Finnish Workers Compensation Centre, Statistics Report ‘Occupational accidents in 2009–2018’ -pdf -file, annex on
p.4 excel table in Finnish, (https://www.tvk.fi/tietopalvelu-ja-julkaisut/tilastot/tyotapaturmatilastot/) (2) https://ec.europa.eu/eurostat/statistics-explained/index.php?title=EU_labour_force_survey_-_ad_hoc_modules#Overview_of_the_ad_hoc_modules
Abstract
4 Methodological study on under-reporting of occupational accidents in European Union
oriented occupations as hazardous situations at workplaces decreases. This study provides
consistent data also in terms of rates of non-fatal accidents by 100 000 workers and the past ad-hoc
labour force survey (2013?).
Conclusion: Adjusting incidence of non-fatal accidents with the proposed methodology reflects the
actual risk level in Member States considerably better and provides a solid and research based
background for preventing risks at places of work.
Keywords: occupational accidents, accident-reporting systems, statistical methods, adjusting and
standardizing data.
Authors: Gianni Betti (3), Ivars Vanadzins (4), Hionia Vlachou (5) and Jukka Takala (6) (7) (8)
Acknowledgement: This work has been carried out under the supervision of the Eurostat project
manager Matyas Meszaros. Special thanks goes to Silvia Crintea-Rotaru for helpful comments.
(3) University of Siena; Department of Economics and Statistics, Italy, gianni.betti@unisi.it (4) Riga Stradins University, Institute of Occupational safety and environmental health, Ivars.Vanadzins@rsu.lv (5) Gopa GOPA Consultants, Senior Statistician, Hionia.Vlachou@gopa.lu (6) ICOH c/o INAIL Italy (7) Tampere University, Faculty of Engineering and Natural Sciences/Faculty of Social Sciences, Finland (8) Correspondence: ICOHPresident@icoh.it, jstakala@gmail.com
Table of contents
5 Methodological study on under-reporting of occupational accidents in European Union
Table of contents
Introduction and objectives ....................................................................................... 8
1. Introduction and objectives .............................................................................................. 8
1.1 Evidence for the existence of under-reporting ................................................................ 9
1.2 Objective ....................................................................................................................... 11
Materials and Methods ............................................................................................. 12
2. Materials and Methods .................................................................................................... 12
2.1 Materials and existing data ........................................................................................... 12
2.2 Description of methodology .......................................................................................... 14
Results....................................................................................................................... 19
3. Results .............................................................................................................................. 19
Discussion ................................................................................................................ 24
4. Discussion ........................................................................................................................ 24
Conclusions .............................................................................................................. 25
5. Conclusions ...................................................................................................................... 25
Annex ........................................................................................................................ 27
Abbreviations
6 Methodological study on under-reporting of occupational accidents in European Union
Abbreviations ESAW European Statistics on Accidents at Work
ILO International Labour Organization
AHMs Ad-Hoc Modules
BSN Baltic Sea Network countries
EFTA European Free Trade Association
ESAW European Statistics on Accidents at Work
EU European Union
Eurobase Eurostat's dissemination database
Eurostat Statistical office of the European Union
EU-OSHA European Union Agency for Safety and Health at Work
MS Member States
ISCO International Standard Classification of Occupations
LFS Labour Force Survey
NACE Statistical classification of economic activities in the European Community
WHO World Health Organization
ILO International Labour Organisation
SRMI sequential regression multivariate imputation
MAR Missing at Random
Country codes EU-28 European Union (28 countries)
EU-25 European Union (25 countries)
BE Belgium
BG Bulgaria
CZ Czechia
DK Denmark
DE Germany
EE Estonia
IE Ireland
EL Greece
ES Spain
FR France
HR Croatia
IT Italy
CY Cyprus
LV Latvia
LT Lithuania
LU Luxembourg
HU Hungary
MT Malta
NL Netherlands
Abbreviations
7 Methodological study on under-reporting of occupational accidents in European Union
AT Austria
PL Poland
PT Portugal
RO Romania
SI Slovenia
SK Slovakia
FI Finland
SE Sweden
UK United Kingdom
1 Introduction and objectives
8 Methodological study on under-reporting of occupational accidents in European Union
1. Introduction and objectives European Statistics on Accidents at Work (ESAW) is an administrative data collection. The legal
basis for ESAW data is the European Commission Regulation (EU) No 349/2011(9) as regards the
statistics on accidents at work. It defines the variables, the breakdowns and the metadata information
that the Member States have to deliver to Eurostat providing reliable and comparable information on
accidents at work. The data contains information about the reporting organisation, about the
characteristics of the victim, characteristics of the accident and certain causes and circumstances of
accidents.
According to the ESAW regulation mentioned above, an accident at work means a discrete
occurrence in the course of work which leads to physical or mental harm. The phrase in the course of
the work means whilst engaged in an occupational activity or during the time spent at work. This
includes road traffic accidents that occur in the course of the work but exclude commuting accidents
i.e. road accidents that occur during the journey between home and workplace.
Accidents at work are disaggregated into ‘non-fatal’ and ‘fatal’ and is defined as follows:
‘A non-fatal accident at work’ is an accident which a victim survives and may result in one or more
days of absence from work. A serious non-fatal accident at work is an accident at work resulting in
more than three days' of absence from work.
The Labour Force Survey modules on accidents at work and other work-related health problems
include self-reported data on all non-fatal accidents (including the possibility to exclude those
accidents with less than four days of absence).
The scope of the administrative data collection ESAW 'European Statistics on Accidents at Work' is
referring only to data on accidents with four days or more of absence as well as fatal accidents at
work. Only full calendar days of absence from work have to be considered, excluding the day of the
accident. Consequently, ‘more than three calendar days’ means ‘at least four calendar days’, which
implies that only if the victim resumes work on the fifth (or subsequent) working day after the date on
which the accident occurred should the incident be included.’ (10)
‘A fatal accident at work refers to an accident at work which leads to the death of a victim within one
year of the accident.’ (11)
These definitions are slightly different from the one provided by the ILO (1988), where: an
occupational accident is an unexpected and unplanned occurrence, including acts of violence,
(9) https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:32011R0349&qid=1556019536013&from=EN
(10) https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Glossary:Non-fatal_accident_at_work
(11) https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Glossary:Fatal_accident_at_work
1 Introduction and objectives
1 Introduction and objectives
9 Methodological study on under-reporting of occupational accidents in European Union
arising out of or in connection with work which results in one or more workers incurring a personal
injury, disease or death. As occupational accidents are to be considered travel, transport or road
traffic accidents in which workers are injured and which arise out of or in the course of work, i.e.
while engaged in an economic activity, or at work, or carrying on the business of the employer.
1.1 Evidence for the existence of under-reporting
Several studies have provided evidence for the existence of under-reporting of both fatal and non-
fatal work accidents in the European Union; in particular, the under-reporting in the case of non-fatal
work accidents could be of high level and differences of magnitude between EU countries significant.
Takala et al. (2017) showed that EU-28 rates are quite low compared to other countries globally
however within the EU the variation is relatively high, see Figure 1. which could signal potential
issues in the reporting of accidents at work. The implementation of systematic accident prevention
programs produced a positive impact and the fatal accident rates, in general, continue to decrease.
However, this trend is not observable for non-fatal accidents. An adjustment in order reach
comparability within the countries, is could be envisaged and would mean adjusting by the original
fatal accident/non-fatal injury rates. The adjustment will take into account the difference of the
economic structure composition and further other variables of the EU-28 as compared to their
original structure.
Actions on the prevention of fatal accidents arising from both work-related traffic injuries and all other
accidents has been challenging due to missing data e.g. from the UK for traffic injuries (12). When
only other accidents than traffic accidents are compared it has been easier. The countries with a low
incidence of accidents at work included major countries such as the UK with 0.74 fatal injuries per
100 000 employed and Germany with 0.9 per 100 000. Comparative global outcomes are based
further on ILO estimates (Hämäläinen et al. 2017) as a source for global data. Data including work-
related traffic accidents in the estimates significantly increases the rates including those in the EU-28
up to double of those presented in Figure 1.
Standardized incidence rates, in Figure 1, include adjustments based on average industry structures
in the EU. Countries that have a relatively high level of activity in high-risk industries, such as
construction, would otherwise show much higher rates as compared to those with a high service
industry component even though within each economic sector, their safety levels and rates would be
equal to those in an another country. While Eurostat rates have been standardized, the non-EU
country rates could not be adjusted due to lack of comparable data. Fatal injury rates in industrialized
countries are gradually decreasing (Takala 2019), partly due to a shift in the countries’ economic
structure from dangerous sectors to less risky ones, such as the services sector as well as
technological developments such as automatisation.
A number of scientific reports (Kurppa 2015, Hämäläinen 2017) showed that similar patterns should
be equally visible for non-fatal accidents.
(12) Health and Safety Executive , United Kingdom, 2011 , see further in Takala et al. 2017
1 Introduction and objectives
10 Methodological study on under-reporting of occupational accidents in European Union
Figure 1: Standardised incidence rates (per 100 000 workers) of fatal accidents at work for 2011 (13)
Takala et al. (2014) specified that accident prevention cannot start from the fatal injuries on top of the
pyramid, see Figure 2, without proper knowledge of the non-fatal injuries and hazards at work as
they represent the majority of accidents at work. The existing hazards — according to accident
causation theories — will end up in a more serious outcome when several contributing factors are
present simultaneously. Fatal outcomes are such rare events and can only be eliminated or reduced
when the individual hazards and exposures are known, and then are gradually and continually
eliminated or reduced, in a priority order, based on risk assessment.
Figure 2: Accidents at work for every fatal accident (14)
(13) Eurostat, Takala et al. 2017, Note: Singapore, Malaysia and World rates not standardised
(14) German figures are from German Statutory Accident Insurance System (DGUV), EU-28 and Singapore fatal injuries are taken from official statistics and non-fatal cases from ad-hoc labour force surveys, Source: Takala et al. JOEH: http://www.tandfonline.com/doi/pdf/10.180/15459624.20003.863131
1 Introduction and objectives
11 Methodological study on under-reporting of occupational accidents in European Union
1.2 Objective
The objective of the present project is to provide further evidence for the existence and the scale of
under-reporting of non-fatal and fatal accidents in the EU. Based on that the study aims at
developing and testing of a new methodology for a possible adjustment and standardisation of the
level of non-fatal accidents for all EU Member States.
Estimates of the magnitude of non-fatal accidents at work based on fatal work injuries are likewise
required to provide a quantitative base for calculating the economic costs of work injuries to
employees, employers, industries, insurances systems and the government. A prerequisite is that the
statistical information is reliable and comparable. Serious under-reporting of work injuries could result
in mistaken policy measures and heavily flawed economic estimations (Kurppa 2015).
2 Materials and Methods
12 Methodological study on under-reporting of occupational accidents in European Union
2. Materials and Methods
2.1 Materials and existing data
The data presently available in Eurobase (Eurostat’ dissemination database) demonstrates a wide
magnitude of difference between the EU Member States.
By inverting the numbers in Figure 2, reported here in EU Member States in Figure 3, it is evident
that the annual rates of fatal/non-fatal accidents vary from year to year significantly due to the fact
that fatal accidents are statistically rare events. In particular, in small countries a rolling average of at
least three years and sometimes more years should be used to avoid false interpretation based on
possible random annual fluctuation. The difference of the vertical scale from this Figure 3 as
compared to the next Figure 4 should be noted.
Figure 3: Number of fatal accidents per 1 000 non-fatal accidents (>3 days lost) in better reporting countries and selected other EU countries (1st group) for all NACE Rev. 2 branches A – U/UNK, 2008–14
The magnitude of difference in these rates are about 102 -fold. The annual fluctuation of the rates in
smaller countries compared to EU-15 total are large and often random. Data for EU-15 are used to
accent the significant differences in incidence rates as data from EU-28 would also include countries
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
EU15 ES DK DE FI
2008 2009 2010 2011 2012 2013 2014
Source: Eurostat (online data code: hsw_n2_02; hsw_n2_01)
2 Materials and Methods
2 Materials and Methods
13 Methodological study on under-reporting of occupational accidents in European Union
with suspected poorer reporting of non-fatal accidents at work.
Figure 4: Number of fatal accidents per 1 000 non-fatal accidents (>3 days lost) in better
reporting (15),and selected other EU countries (2nd group) for all NACE Rev. 2 branches A –
U/UNK, 2008–14
Another major difference of the fatal/non-fatal accidents can be seen when analysing various
sectoral accident rates shown in Figure 5. This means that adjusting the values should take into
account several variables such as economic sector (NACE), age, sex, rolling average of the years
etc.
(15) The country selection is simply due to level of reporting not necessarily between EU-15 and the rest. Also later
division of groups in Table 2 is based on WHO classification due to estimates made by ILO using WHO regions. WHO High Income includes Slovenia (not EU-15) higher income.
0
20
40
60
80
100
120
EU15 RO BG LT LV
2008 2009 2010 2011 2012 2013 2014
Source: Eurostat (online data code: hsw_n2_02; hsw_n2_01)
2 Materials and Methods
14 Methodological study on under-reporting of occupational accidents in European Union
Figure 5: Number of fatal accidents per 1 000 non-fatal accidents (>3 days lost) for selected NACE Rev. 2 divisions in EU-15, 2014 (ratio)
2.2 Description of methodology
The methodology proposed is based on applying of different methods in order to statistically
determine the size of under-reporting of non-fatal accidents in the EU.
The unique previous tentative effort for assessing the size of underreporting of non-fatal accidents
was carried out by Kurppa (2015), whose aim was to provide estimates in order to determine the
magnitude in size of occurrence and levels of reporting and underreporting of non-fatal work
accidents in the Baltic Sea Network (BSN) countries, including eight out of 28 EU countries. The
study conducted semi-quantitative data analyses for monitoring statistics that have been submitted
0
2
4
6
8
10
12A
03 -
Fis
hin
g a
nd
aqu
aculture
B06 -
Extr
action o
f cru
de p
etr
ole
um
…
A02 -
Fore
str
y a
nd
loggin
g
H50 -
Wate
r tr
ansport
B08 -
Oth
er
min
ing a
nd q
uarr
yin
g
H49 -
Lan
d tra
nsport
and tra
nsp
ort
via
pip
elin
es
A01 -
Cro
p a
nd a
nim
al pro
duction, h
unting a
nd…
F42 -
Civ
il eng
ineeri
ng
E37 -
Se
wera
ge
B05 -
Min
ing
of coal and
lig
nite
C14 -
Ma
nufa
ctu
re o
f w
eari
ng a
ppare
l
F41 -
Constr
uction o
f build
ings
D3
5 -
Ele
ctr
icity,
ga
s,
ste
am
an
d a
ir c
on
ditio
nin
g s
up
ply
E38 -
Waste
colle
ction, tr
eatm
ent a
nd d
isposal…
C20 -
Ma
nufa
ctu
re o
f che
mic
als
and
chem
ical pro
ducts
F43 -
Specia
lised c
onstr
uction a
ctivitie
s
J63 -
Info
rma
tion s
erv
ice a
ctivitie
s
UN
K -
Unknow
n N
AC
E a
ctivity
M6
9 -
Le
gal and
accounting a
ctivitie
s
C13 -
Ma
nufa
ctu
re o
f te
xtile
s
C16 -
Ma
nufa
ctu
re o
f w
ood a
nd o
f pro
ducts
of w
oo
d…
C18 -
Printing a
nd r
epro
duction o
f re
co
rded m
edia
N78 -
Em
plo
ym
ent activitie
s
C26 -
Ma
nufa
ctu
re o
f com
pu
ter,
ele
ctr
onic
and
optical…
C32 -
Oth
er
ma
nufa
ctu
ring
N81 -
Serv
ices to b
uild
ings a
nd
landscape a
ctivitie
s
C10 -
Ma
nufa
ctu
re o
f fo
od
pro
du
cts
G47 -
Reta
il tr
ade,
except o
f m
oto
r vehic
les a
nd
…
H53 -
Posta
l and
couri
er
activitie
s
C31 -
Ma
nufa
ctu
re o
f fu
rniture
Q87 -
Resid
ential care
activitie
s
Source: Eurostat (online data code: hsw_n2_02; hsw_n2_01)
2 Materials and Methods
15 Methodological study on under-reporting of occupational accidents in European Union
by national statistical authorities to the ILO during a 5-year period between 2003 and 2007.
Kurppa (2015), adopted two main schemes for providing estimates of the order of magnitude about
the size of occurrence and levels of reporting and underreporting of non-fatal work accidents:
In the first scheme, incidence rates of accidents at work of benchmark countries were applied to the
workforce of a country. Then the calculated numbers of accidents were compared with the numbers
registered in the official statistics.
In the second scheme, the expected numbers of non-fatal work accidents were calculated by
multiplying the registered number of fatal work accidents in a country by an external coefficient (ratio
between fatal and non-fatal accidents) of a benchmark country.
Based on previous analysis, Kurppa (2015) noted that the reporting of non-fatal work accidents is
nearly complete in Finland and Germany, where the compensation schemes encourage reporting by
providing positive incentives.
In the current paper, statistical indicators for Finland, Germany, and the EU-15 average were used
as benchmarks against the indicator data of work accidents of other countries were compared.
Germany and Finland show no or very low levels of underreporting; for those reasons the first step of
the proposed methodology consists in selecting Germany and Finland as ‘benchmark’ countries.
In Kurppa (2015) incidence rates have been estimated by gender and by branch of economic activity;
however, evidence shows that such rates may vary according also to other characteristics of the
labour market and due to administrative requirements. For these reasons, in the present
methodology, we aimed to investigate which could be the variables or characteristics having an
impact in the estimation of incidence rates. For meeting such an objective, an iterative imputation
procedure for correcting non-fatal accidents percentages/ratios was proposed. In synthesis, the
proposal is based on an iterative procedure of one model with three consecutive steps, as follows:
Step 1. Following proposal of Kurppa (2015), choosing ‘benchmark’ countries.
Step 2. Estimate linear regression models of non-fatal / fatal accidents ratios in such countries in
function of some selected explanatory variables.
Step 3. Impute non-fatal / fatal accidents ratios in other countries; consider such estimate as
Estimate [0]
Repeat Steps 2–3 in iterative way until convergence of the regression parameters in step 2 is
reached.
A suitable method is to use the ratio between fatal accidents and non-fatal accidents as long as
relatively reliable data for fatal cases is available. In general, fatal cases are much better reported
whereas reporting of the non-fatal ones have some major shortcomings. While in the ILO’s Global
Estimates by Hämäläinen et al. (2017) an overall estimate were used for all industries – and in some
cases in three major sectorial groups (agriculture, industry & construction, services) this could be
done at least for those economic sectors where sufficient data is available. Said that, the two main
decisions for the regression model in step 2 are questioned as follows:
1. Which type of model should be used linear /non-linear, including the dependent variable and
which regressors? The interpretation and estimation of linear models it is easier compared
to non-linear models. However, the dependent variable should have a non-skewed
distribution, although it will be difficult to get a normal distribution of the residual. For these
considerations, a logarithmic transformation of ratios of non-fatal versus fatal percentages
has been selected. The logarithm of ratios could be also seen as difference of logarithm of
percentages.
2. The number and type of statistical units; the list of regression could be the following:
2 Materials and Methods
16 Methodological study on under-reporting of occupational accidents in European Union
NACE groups (16);
gender;
age groups;
part of the body injured;
type of injury.
The number of statistical units depends on the number of years and the number of sub-groups to be
analyzed. There is a trade off in the choice of the number of the statistical units: on one side, the
higher the number (pattern ‘A’ in figure 6a), the higher the degrees of freedom, and the estimates
would be more efficient (lower variability of parameters); on the other side, the smaller the number
(pattern ‘E’ in figure 6a), the less biased is the calculation of non-fatal / fatal percentage ratio within
the sub-group. Since such sub-groups could be seen as a ‘pseudo-panel’ from a statistical point of
view, here it is proposed to adopt the method of Veerbek and Ninjam (1990, 1992, 1994) for
identifying the optimal number of statistical units.
The optimal solution consists in ‘cumulating’ information over 3 years; in particular, by adding a new
and fresh year when available, in this way new estimates might be calculated for every year (pattern
C2 in Figure 6b; this has also been suggested and applied by Betti et al., 2002 and Betti and
Gagliardi, 2018).
Figure 6a: Possible patterns of cumulation over years
(16) https://ec.europa.eu/eurostat/ramon/nomenclatures/index.cfm?TargetUrl=LST_NOM_DTL&StrNom=NACE_REV2&StrLanguageCode=EN&IntPcKey=&StrLayoutCode=HIERARCHIC
2010 2011 2012 2013 2014 2015 2016 2017
A X X X X X X X X
B X X X X X X X X
C X X X X X X X X
D X X X X X X X X
E X X X X X X X X
2 Materials and Methods
17 Methodological study on under-reporting of occupational accidents in European Union
Figure 7b: Possible patterns of cumulation over 3 years
Once the number of statistical units has been identified, the proposed imputation procedure is a
‘light’ and ad hoc version of the ‘sequential regression multivariate imputation’ (SRMI) approach
adopted by the imputation SAS software (IVE-ware). The later method, proposed by the authors of
the software (Raghunathan, Lepkowski, Van Hoewyk, and Solenberger, 2001), constructs the
imputed values by fitting a sequence of regression models and drawing values from the
corresponding predictive distribution, under the hypothesis of Missing at Random (MAR) mechanism,
infinite sample size and simple random sampling.
Taking into account all the aspects discussed above, the proposed model is described by the
equation (1):
𝑙𝑜𝑔 (% 𝑛𝑜𝑛−𝑓𝑎𝑡𝑎𝑙
% 𝑓𝑎𝑡𝑎𝑙)
𝑖𝑡= 𝛼 + 𝛽1𝑋1𝑖𝑡 + 𝛽2𝑋2𝑖𝑡 + ⋯ + 𝛽𝑘𝑋𝑘𝑖𝑡 + 𝜀𝑖𝑡 (1)
Where 𝑋𝑘𝑖𝑡 are the regressors to be included in the model, while 𝜀𝑖𝑡 is an error defined by two or
three variance components. This model is estimated for NACE groups, gender, age group, part of
the body injured, type of injury.
In particular, we propose to implement the ‘non-fatal accidents percentages/ratios’, estimated under
section 2, in order to estimate the level of incidence rates in Member States, where with the
proposed method it was identified to have under-reporting issues of non-fatal accidents at work.
Example of implementation of the proposed statistical methodology:
An iterative procedure will be performed for Steps 2 and 3. At every iteration, we estimate a set of
parameters 𝛽1, 𝛽2, … 𝛽𝑘; we compare such parameters with those in the previous iteration; when the
difference is less than a small value chosen, the iterative process ends. Then Task 3 is performed as
follows:
multiply such final set of parameters 𝛽1, 𝛽2, … 𝛽𝑘; by the corresponding set of
regressors/variables for a specific country, in a year;
in this way, we get the estimated log of the non-fatal/fatal ratio: 𝑙𝑜𝑔 (% 𝑛𝑜𝑛−𝑓𝑎𝑡𝑎𝑙
% 𝑓𝑎𝑡𝑎𝑙)
𝑖𝑡 (2)
calculate the non-fatal/fatal ratio by the exponential function;
multiply such ratio by the total number of fatal accidents in that country in that year;
2010 2011 2012 2013 2014 2015 2016 2017
C1 X X X X X X X X
C2 X X
X X X
X X X
X X X
X X X
X X X
X X X
2 Materials and Methods
18 Methodological study on under-reporting of occupational accidents in European Union
While there are several variables used – NACE groups, gender, age groups, body part, type of injury
— obviously not all have the same influence to the outcome. As a result, the biggest impact comes
from the sectoral variable — NACE groups. The order of the imputation process does not have an
impact on the outcome. The model is exactly the same for all variable treatments. The reason of the
selecting this method is explained above in point 2.2.
3 Results
19 Methodological study on under-reporting of occupational accidents in European Union
3. Results There is a wealth of information and reports that fatal accidents are much better reported than no-
fatal accidents in the majority of EU countries. This under-reporting study used several variables for
adjusting the poorly reported non-fatal accidents, as explained based on reporting from the
benchmark countries, using available variables, in particular, the sectoral rates of fatal and non-fatal
accidents, the age groups, the type of injury and the part of body injured. The regression results of
model in equation (1) in Section 2 are estimated for: NACE groups, age groups, part of the body
injured, type of injury and are reported in tables 1a and 1b. All four models are very good, and R-
squared ranging from 42 % to 79 %.
The selection of the two benchmark countries is based on the highest rank of reporting non- fatal
accident/fatal accident rates. Adding more Member States that have presumable lower level
reporting rates would dilute the outcome. On the other hand, for the selected benchmark countries
(Germany and Finland), there are no major reasons to assume over-reporting. Furthermore, there
may be still some level of under-reporting in these countries that is more likely for non-fatal minor
accidents as compared to fatal ones. The later variables such as part of the body injured and age
have actually very minor impact on the outcome.
The ILO Estimates (Hämäläinen, Takala et al. 2017) on the non-fatal number of accidents is the only
so far carried out similar effort worldwide. While it is based on a rough grouping of economic sectors
to only three: 1) agriculture, fishing, forestry, 2) industry and construction and 3) service industries, it
has the same baseline thinking than this study. Due to lack of global detailed data that outcome is
less reliable compared to that of the process in this study but points out to the same direction.
ILO 2014 data is the latest available outcome used for such process. Due to wide data collection
covering both occupational accidents and work-related diseases, the WHO data grouping available
was used by the ILO study. This is based on WHO regions called ‘WHO High Income’ and ‘WHO
EURO’ Region. As a result, the table lists first the WHO EURO Region Countries, followed by the
WHO High Income Region (starting by Austria). The WHO Regions are based on general health
factors in addition to occupational ones but it appears to be comparable also to accident reporting
due to differences in administrative practices in the two regions.
Table 1a: Results of regression models for NACE groups, age groups
Parameters Model NACE p-values Parameters
Model AGE p-values
Intercept 8 <.0001 Intercept 7 <.0001 NACE_A – Agriculture, forestry and fishing
0 <.0001 AGE18_34 0 <.0001
NACE_C – Manufacturing 0 1 AGE35_54 1 0
NACE_F – Construction 0 <.0001 AGEGE_55 0 1
NACE_H – Transportation and storage 0 <.0001
R-squared 1 R-squared 0
3 Results
3 Results
20 Methodological study on under-reporting of occupational accidents in European Union
The numbers estimated from the model in equation (2) in Section 2 are considerably different from
the raw reported numbers from countries, and in most cases also show significantly higher figures as
compared to ILO estimates based on rough country adjustments.
Figure 7 reflects the size of countries however also other factors clearly influence the outcomes of
this under-reporting study. It is demonstrating the outcome from the construction sector (NACE
rev. 2 - F). While reports show that the construction sector is not just a very risky one and also has a
high number of workers involved. The imminent risk from falling from elevated levels is a key cause
for high number of fatal injuries, but there are also a high number of non-fatal cases as well.
However, the high number of fatal cases causes a different rate between fatal and non-fatal
accidents. On other hand, the services sector and other selected industries have an opposite trend
as the fatal accidents are relatively rare events.
Figure 8: Number of All accidents, annual average in 2015–17, adjusted by NACE sectors
While for the full number of non-fatal cases it is important to estimate the magnitude and level of
risks, the comparison between countries needs to take into account their sectoral distribution and to
some level also their technological matureness and peculiarities of workforce. To compare just
consolidated numbers would give a picture that countries with high level of activities in construction,
manufacturing, mining, agriculture, forestry and fishing are more dangerous than those where less
risky jobs are more frequent even if the incidence rates in these industries are actually comparable.
Furthermore, the rate of fatal accidents/100 000 workers (see Figure 1) as well as the adjusted rate
of non-fatal cases/100 000 workers provide a relatively fair comparison between countries. However,
e.g. Luxembourg may have had some of the random fatal cases within the period under
consideration that may need more comprehensive number of years of adjustment. A best picture is
obtained with all these adjustments together including access to full data, see Figures 8 and 9.
0
200 000
400 000
600 000
800 000
1 000 000
1 200 000
AT BEBGCY CZ DEDK EE ES FI FRGRHRHU IE IT LT LU LV MT NL PL PTROSE SI SK UK
3 Results
21 Methodological study on under-reporting of occupational accidents in European Union
Figure 9: Number of non-fatal accidents in construction sector in EU-28 annual average, 2015–17 (NACE F), adjusted
Figure 10: Rate of non-fatal accidents/employment, annual average in 2015–2017, adjusted by NACE Sector and by best reporting countries, EU-28 divided into 2 separate WHO Regions. (%)
0
20 000
40 000
60 000
80 000
100 000
120 000
140 000
AT BE BGCY CZ DE DK EE ES FI FRGRHRHU IE IT LT LU LV MT NL PL PT ROSE SI SK UK
0%
2%
4%
6%
8%
10%
12%
14%
Bu
lgaria
Cro
atia
Czech
Repu
blic
Esto
nia
Hu
nga
ry
La
tvia
Lithu
ania
Po
land
Ro
mania
Slo
vakia
Au
str
ia
Be
lgiu
m
Cypru
s
De
nm
ark
Fin
land
Fra
nce
Germ
an
y
Gre
ece
Ire
land
Ita
ly
Lu
xe
mbou
rg
Malta
Ne
therla
nds
Po
rtug
al
Slo
ven
ia
Sp
ain
Sw
ede
n
Un
ited
Kin
gd
om
EU
28
3 Results
22 Methodological study on under-reporting of occupational accidents in European Union
Table 2: Comparative analysis of fatal and non-fatal accidents in EU-28 based on ILOSTAT,
EUROSTAT data (17) C
ou
ntr
y
To
tal
em
plo
ym
en
t
Occu
pa
tio
nal in
juri
es in
2014 r
ep
ort
ed
to
IL
O
Fata
l
Occu
pa
tio
na
l in
juri
es in
2014 r
ep
ort
ed
to
IL
O
No
n-f
ata
l
Accid
en
ts a
t w
ork
in
2014 r
ep
ort
ed
to
Eu
ros
tat
Fa
tal
Accid
en
ts a
t w
ork
in
2014 r
ep
ort
ed
to
Eu
ros
tat
No
n-f
ata
l
Glo
bal
esti
mate
s o
f o
ccu
pa
tio
nal
accid
en
ts
(A
t le
ast
fou
r d
ays a
bs
en
ce)
Lo
wer
lim
it (
Eq
. j)
(0.1
4)
ILO
2014
Glo
bal
esti
mate
s o
f o
ccu
pa
tio
nal
accid
en
ts
(At
least
fou
r d
ays a
bs
en
ce)
Up
pe
r lim
it (
Eq
. j)
(0.0
8 IL
O 2
014
Glo
bal
esti
mate
s o
f o
ccu
pa
tio
nal
accid
en
ts (
At
least
fou
r d
ays a
bs
en
ce)
Sele
cte
d I
LO
2014
No
n-f
ata
l accid
en
ts b
ased
on
th
e E
U28 U
nd
er-
rep
ort
ing
stu
dy,
all s
ecto
rs,
an
nu
al
avera
ge o
f
years
2015
-2017
EU-28 218 336 000 3 379 2 414 073 2 413 571 4 051 944 3 548 302 6 936 092
Bulgaria 2 981 400 115 2289 110 1 772 78 571 137 500 108 036 175 403
Croatia 1 565 700 36 13785 22 8 999 15 714 27 500 21 607 65 206
Czech Republic 4 974 300 45058 101 36 622 72 143 126 250 99 196 217 135
Estonia 624 800 16 4619 13 5 393 9 286 16 250 12 768 33 255
Hungary 4 100 800 78 19583 74 15 918 52 857 92 500 72 679 162 362
Latvia 884 600 39 1 409 27 857 48 750 38 304 55 425
Lithuania 1 319 000 60 3232 51 2 599 36 429 63 750 50 089 79 551
Poland 15 861 500 225 59 414 160 714 281 250 220 982 532 730
Romania 8 613 700 224 3351 253 3 101 180 714 316 250 248 482 494 258
Slovakia 2 363 100 39 7 365 27 857 48 750 38 304 93 244
Austria 4 113 700 119 52 968 85 000 132 222 108 611 221 047
Belgium 4 544 500 45 46 704 32 143 50 000 46 704 121 935
Cyprus 362700 5 1 613 4 1 359 2 857 4 444 4 563 7 173
Denmark 2 714 100 28 31 770 20 000 31 111 31 770 58 685
Finland 2 447 200 28 42 162 20 000 31 111 42 162 41 773
France 26 396 400 517 467 869 369 286 574 444 471 865 776 729
Germany 39 871 300 639 955 280 471 704 819 336 429 523 333 704 819 862 016
Greece 3 536 200 25 3 152 17 857 27 778 22 817 60 641
Ireland 1 913 900 45 13 103 32 143 56 250 44 196 88 028
Italy 22 278 900 396 305 246 459 251 769 327 857 573 750 388 929 983 300
Luxembourg 245 600 10 6 154 7 143 12 500 9 821 29 343
Malta 181 700 4 2 273 2 857 5 000 3 929 8 477
Netherlands 8 236 100 39 56 377 27 857 48 750 56 377 232 132
Portugal 4 499 500 148 111 134 105 714 185 000 145 357 286 253
(17) Adjusted by the EU-28 Under-reporting Study. ILO data is classified by WHO Regions (alphabetical in two groups).
Year 2014 is used as the latest available international reference.
3 Results
23 Methodological study on under-reporting of occupational accidents in European Union
Co
un
try
To
tal
em
plo
ym
en
t
Occu
pa
tio
nal in
juri
es in
2014 r
ep
ort
ed
to
IL
O
Fata
l
Occu
pa
tio
na
l in
juri
es in
2014 r
ep
ort
ed
to
IL
O
No
n-f
ata
l
Accid
en
ts a
t w
ork
in
2014 r
ep
ort
ed
to
Eu
ros
tat
Fa
tal
Accid
en
ts a
t w
ork
in
2014 r
ep
ort
ed
to
Eu
ros
tat
No
n-f
ata
l
Glo
bal
esti
mate
s o
f o
ccu
pa
tio
nal
accid
en
ts
(A
t le
ast
fou
r d
ays a
bs
en
ce)
Lo
wer
lim
it (
Eq
. j)
(0.1
4)
ILO
2014
Glo
bal
esti
mate
s o
f o
ccu
pa
tio
nal
accid
en
ts
(At
least
fou
r d
ays a
bs
en
ce)
Up
pe
r lim
it (
Eq
. j)
(0.0
8 IL
O 2
014
Glo
bal
esti
mate
s o
f o
ccu
pa
tio
nal
accid
en
ts (
At
least
fou
r d
ays a
bs
en
ce)
Sele
cte
d I
LO
2014
No
n-f
ata
l accid
en
ts b
ased
on
th
e E
U28 U
nd
er-
rep
ort
ing
stu
dy,
all s
ecto
rs,
an
nu
al
avera
ge o
f
years
2015-2
017
Slovenia 916 700 25 12
914 20 10 016 14 286 25 000 24 554 34 559
Spain 17 344 200 246 423 106 247 287 809 176 429 308 750 287 809 624 018
Sweden 4 772 100 41 30
319 36 21 343 25 714 45 000 40 268 74 986
United Kingdom 30 672 300 207 160 700 147 857 258 750 203 304 516 428
Table 2 shows the approximate outcomes related to the comparative analysis. The reference
numbers from the year 2014 from ILOSTAT and Eurostat will be more reliable for bigger Member
States while in small countries the random character of fatal accidents in any year could be balanced
by a taking average values also for the reference values of ILOSTAT/EUROSTAT. The ILO
Estimates (Hämäläinen, Takala et al. 2017) have already used an adjustment of the non-fatal cases
and these are already better than raw reporting data. The results of this under-reporting study show
clearly higher numbers. These are practically more than doubling the raw reported numbers from 2.4
million to 6.9 million of non-fatal accidents in EU-28 causing more than three days of absence from
work. The ILO Estimates for EU in 2017 is in the middle of the two different numbers or 3.6 million
occupational accidents with a lower and higher estimates and confidence interval of 2.5 – 4.1 million.
The authors of the ILO Estimate plan to use the methodology of this paper in near future.
The non-fatal accident numbers and level is important to highlight sectors, workplaces and jobs that
need special attention and provides much more information that the relatively rare fatal cases for
prevention. Furthermore, the overall costs to workers and their families, employers and the society
has been shown to be highest by those non-fatal injuries that are resulting in serious injuries,
extensive and permanent harm and lifetime inabilities to work. The level and numbers of those is
much higher than those of fatal cases and consequently results in highest costs and loss of Disability
Adjusted Life Years.
4 Discussion
24 Methodological study on under-reporting of occupational accidents in European Union
4. Discussion Different industries may show different patterns of work injuries because of industry-specific
circumstances. This is influencing also the ratio between fatal and non-fatal work injuries, which may
be inherently different across branches of economy (Kurppa 2015). Furthermore, stability of the ratio
of fatal/non-fatal accidents changes, in particular, for smaller countries where the number of fatal
accidents may vary significantly from one year to another. Therefore, it is essential to cover more
years, e.g. by using a rolling average of several years. The numbers of accidents, both fatal and non-
fatal, in most EU countries goes gradually down due to shift from more dangerous jobs to less
dangerous ones and general improvements in control of workplace risks, as increasing number of
workers are employed in service occupations that are less hazardous, in particular, for fatal
accidents risks.
We do not imply that the proposed approach and calculated numbers are precise due to 28 different
accident registration systems in EU. Different practices, administrative requirements and incentive
systems exist for workers reporting their accidents to their employer and further employers reporting
to authorities, and then finally countries reporting to EU. However, this method gives a much more
reliable and comparable data than reporting of just the raw data of today and thus could be used
both at national level and internationally for planning of effective and justified preventive measures.
Such methodology for calculations could also be used by governments to analyse the potential
effects of accidents on social security systems and economics in general.
4 Discussion
6 Annex
25 Methodological study on under-reporting of occupational accidents in European Union
5. Conclusions We are convinced that there is a solid and scientific solution to counter the obvious under-reporting
of non-fatal accidents (injuries) in EU Member State country statistics. The applied outcome of this
study is feasible and based on solid evidences. This study proposes a scientific solution and tries to
find an optimal method for estimating the level of under-reporting. It is a first comprehensive and
statistically tested effort based also on Kurppa 2015 and Hämäläinen, Takala 2017 and covers the
full EU level.
The system should take into account the economic structure and sectors of individual Member
States, the differing rates of fatal to non-fatal injuries in various sectors, and other parameters and
variables described earlier. Fatal outcomes are still rare events and the annual numbers in small
Member States need rolling average adjustments from year to year.
While fatal injuries are better recorded and reported than non-fatal ones, also the fatal cases may not
be reported in a uniform manner from original sources across countries. The quality and type of
various registration and compensation systems in Member States varies widely, and this may result
of under-reporting of fatal cases.
The target to standardize registration and compensation systems in EU is challenging as policy
makers mostly believe in their own country numbers. Convincing and explaining the need and
evidence for further adjustments and standardization is important to compensate for the EU-level
differences to provide reliable data.
The proposed model and methodology may also be used for calculation of so far non-covered
injuries, notably the continuously growing group of self-employed workers, non-wage farmers and
other non-wage worker groups. Estimations and calculation for accident level in the informal sector
should follow.
We have used selected country data notably those of Germany and Finland as references and
baseline for standardization. However, it should be kept in mind that Denmark has started currently
to work in the area of under-reporting, which was presented in the Luxembourg Working Group
meeting of last 16 October 2019. For this reason, in the near future, the reference base may be
broadened.
5 Conclusions
6 Annex
26 Methodological study on under-reporting of occupational accidents in European Union
References
Baltagi B.H. (1995), Econometric Analysis of Panel Data, John Wiley. 4th edition 2008.
Betti G. (2000), ‘Intertemporal Equivalence Scales’, International Review of Social Sciences (Rivista
Internazionale di Scienze Sociali), 25, pp. 21–36.
Betti G., D’Agostino A., Neri L. (2002), ‘Panel Regression Models for Measuring Multidimensional
Poverty Dynamics’, Statistical Methods and Applications, 11(3), pp 359–369.
Betti G., Gagliardi F. (2018), ‘Extension of JRR Method for Variance Estimation of Net Changes in
Inequality Measures’, Social Indicators Research, 137(1), pp. 45–60.
Hämäläinen P., Takala J., Tan B.K. (2017), Global estimates of occupational accidents and work-
related illnesses 2017, WSH Institute, Ministry of Manpower, ICOH et al. https://goo.gl/2hxF8x,
Accessed 20 Nov 2019.
Kari Kurppa and Finnish Institute of Occupational Health (2015): Severe Under-reporting of Work
Injuries in Many Countries of the Baltic Sea Region: An exploratory semi-quantitative study – What
goes unreported goes unfixed.
International Labour Organization (ILO), Resolutions Concerning statistics of occupational injuries
(resulting from occupational accidents), adopted by the Sixteenth International Conference of Labour
Statisticians (October 1998), pp. 2.
Nijman, Th.E., Verbeek, M. (1990), ‘Estimation of Time-Dependent Parameters in Linear Models
Using Cross-Sections, Panels, or Both’; Journal of Econometrics, 46, pp. 333–346.
Raghunathan T.E., Lepkowski J., Van Voewyk J., Solenberger P. (2001), ‘A Multivariate Technique
for Imputing Missing Values Using a Sequence of Regression Models’, Survey Methodology, 27, pp.
85–95.
Takala J., Hämäläinen P., Saarela K. L., Yoke Yun L., Manickam K., Tan Wee J., Heng P., Tjong C.,
Guan Kheng L., Lim S., Gan S. L. (2014), Global ‘Estimates of the Burden of Injury and Illness at
Work in 2012’, Journal of Occupational and Environmental Hygiene, 11, pp. 326–337.
Takala J., et al. (2017). ‘Comparative Analysis of the Burden of Injury and Illness at Work in Selected
Countries and regions’, Central European Journal of Occupational and Environmental Medicine, 23
(1-2), pp. 6–31.
Takala J. (2019), Burden of Injuries due to Occupational Exposures. In: U. Bültmann, J. Siegrist
(eds.), Handbook of Disability, Work and Health, Handbook Series in Occupational Health Sciences,
https://doi.org/10.1007/978-3-319-75381-2_5-1 © Springer Nature Switzerland AG 2019.
Verbeek, M., Nijman, Th.E. (1992), ‘Can Cohort Data Be Treated As Genuine Panel Data? ’,
Empirical Economics’, 17, pp. 9–23.
Verbeek, M., Nijman, Th.E. (1994), ‘Minimum MSE Estimation of a Regression Model with Fixed
Effects from a Series of Cross Sections’; Journal of Econometrics, 59, pp. 125–136.
6 Annex
27 Methodological study on under-reporting of occupational accidents in European Union
Classification of the different variables follow the Annex I of European Statistics on Accidents at
Work (ESAW)- Summary methodology, 2013 edition (18).
Figure 11: Results by NACE. Number of All accidents, sum of the years in 2015–17 and average annual number, adjusted by NACE sectors and by best reporting countries, and rate of under-reporting
(non_fatal_final = total estimate of the years, average 2015–17 = rolling average over the three years)
Country non_fatal fatal NACE non_fatal_final rate (%) of
underreporting average 2015–17
AT 12 307 144 A 51 908 76 17 303
BE 1 129 4 A 2 460 54 820
BG 161 17 A 15 384 99 5 128
CY 100 2 A 476 79 159
CZ 7 511 34 A 44 907 83 14 969
DE 179 040 216 A 179 040 0 59 680
DK 2 863 24 A 4 895 42 1 632
EE 429 4 A 5 175 92 1 725
ES 84 473 126 A 147 840 43 49 280
FI 11 329 17 A 11 329 0 3 776
FR 32 637 74 A 66 555 51 22 185
GR 276 7 A 5 142 95 1 714
HR 1 927 11 A 11 391 83 3 797
HU 2 210 36 A 19 748 89 6 583
IE 3 831 70 A 23 224 84 7 741
IT 87 978 233 A 351 325 75 117 108
LT 446 17 A 11 968 96 3 989
LU 396 2 A 1 949 80 650
LV 213 16 A 7 949 97 2 650
MT 62 0 A 321 81 107
NL 12 3 A 8 266 100 2 755
PL 3 731 48 A 29 486 87 9 829
(18) https://ec.europa.eu/eurostat/documents/3859598/5926181/KS-RA-12-102-EN.PDF/56cd35ba-1e8a-4af3-9f9a-b3c47611ff1c
6 Annex
6 Annex
28 Methodological study on under-reporting of occupational accidents in European Union
Country non_fatal fatal NACE non_fatal_final rate (%) of
underreporting average 2015–17
PT 17 143 74 A 44 179 61 14 726
RO 438 125 A 62 756 99 20 919
SE 1 617 24 A 4 013 60 1 338
SI 1 025 6 A 3 433 70 1 144
SK 1 494 12 A 17 569 91 5 856
UK 21 317 110 A 58 208 63 19 403
AT 41 388 25 C 137 849 70 45 950
BE 29 727 30 C 51 146 42 17 049
BG 1 800 50 C 135 817 99 45 272
CY 995 1 C 3 740 73 1 247
CZ 59 597 64 C 281 381 79 93 794
DE 627 460 223 C 627 460 0 209 153
DK 16 673 9 C 22 507 26 7 502
EE 3 540 9 C 33 722 90 11 241
ES 209 131 164 C 289 029 28 96 343
FI 18 109 11 C 18 109 0 6 036
FR 159 324 168 C 256 569 38 85 523
GR 2 435 17 C 35 827 93 11 942
HR 9 749 12 C 45 507 79 15 169
HU 25 327 39 C 178 717 86 59 572
IE 8 778 5 C 42 015 79 14 005
IT 198 877 250 C 627 148 68 209 049
LT 3 220 19 C 68 233 95 22 744
LU 2 134 4 C 8 293 74 2 764
LV 1 702 12 C 50 161 97 16 720
MT 1 212 2 C 4 949 76 1 650
NL 233 88 C 126 747 100 42 249
PL 83 431 171 C 520 672 84 173 557
PT 102 168 66 C 207 924 51 69 308
RO 4 400 124 C 497 831 99 165 944
SE 15 045 12 C 29 482 49 9 827
SI 12 727 13 C 33 657 62 11 219
SK 11 777 23 C 109 366 89 36 455
UK 98 424 75 C 212 232 54 70 744
AT 35 079 51 F 136 862 74 45 621
BE 24 041 42 F 48 453 50 16 151
BG 566 62 F 50 027 99 16 676
CY 578 2 F 2 545 77 848
CZ 7 923 67 F 43 819 82 14 606
DE 346 671 251 F 346 671 0 115 557
DK 16 186 14 F 25 594 37 8 531
6 Annex
29 Methodological study on under-reporting of occupational accidents in European Union
Country non_fatal fatal NACE non_fatal_final rate (%) of
underreporting average 2015–17
EE 877 12 F 9 786 91 3 262
ES 141 872 190 F 229 682 38 76 561
FI 19 340 14 F 19 340 0 6 447
FR 154 009 230 F 290 519 47 96 840
GR 1 172 26 F 20 200 94 6 733
HR 3 303 29 F 18 060 82 6 020
HU 2 327 62 F 19 235 88 6 412
IE 5 119 24 F 28 703 82 9 568
IT 90 362 337 F 333 794 73 111 265
LT 1 125 26 F 27 925 96 9 308
LU 6 583 9 F 29 969 78 9 990
LV 413 12 F 14 258 97 4 753
MT 858 7 F 4 104 79 1 368
NL 84 47 F 53 526 100 17 842
PL 16 042 179 F 117 274 86 39 091
PT 56 888 127 F 135 617 58 45 206
RO 1 438 170 F 190 588 99 63 529
SE 13 212 21 F 30 329 56 10 110
SI 4 034 11 F 12 497 68 4 166
SK 1 330 25 F 14 468 91 4 823
UK 72 775 147 F 183 821 60 61 274
AT 14 514 45 H 55 989 74 18 663
BE 23 197 40 H 46 225 50 15 408
BG 943 55 H 82 410 99 27 470
CY 404 1 H 1 759 77 586
CZ 12 725 61 H 69 585 82 23 195
DE 219 312 247 H 219 312 0 73 104
DK 11 024 15 H 17 235 36 5 745
EE 864 14 H 9 533 91 3 178
ES 88 711 204 H 141 999 38 47 333
FI 12 341 16 H 12 341 0 4 114
FR 116 911 182 H 218 054 46 72 685
GR 1 143 8 H 19 478 94 6 493
HR 2 656 12 H 14 359 82 4 786
HU 8 719 49 H 71 258 88 23 753
IE 3 809 8 H 21 118 82 7 039
IT 89 958 188 H 328 557 73 109 519
LT 1 141 28 H 28 003 96 9 334
LU 1 863 14 H 8 386 78 2 795
LV 740 29 H 25 259 97 8 420
MT 841 0 H 3 977 79 1 326
6 Annex
30 Methodological study on under-reporting of occupational accidents in European Union
Country non_fatal fatal NACE non_fatal_final rate (%) of
underreporting average 2015–17
NL 110 44 H 69 304 100 23 101
PL 18 702 143 H 135 179 86 45 060
PT 26 543 55 H 62 563 58 20 854
RO 1 181 103 H 154 762 99 51 587
SE 10 301 25 H 23 378 56 7 793
SI 2 934 12 H 8 987 67 2 996
SK 2 714 43 H 29 191 91 9 730
UK 85 087 205 H 212 499 60 70 833
AT 84 834 74 Other 280 533 70 93 511
BE 127 338 71 Other 217 520 41 72 507
BG 3 238 85 Other 242 572 99 80 857
CY 3 483 5 Other 12 998 73 4 333
CZ 45 164 107 Other 211 712 79 70 571
DE 1 213 566 356 Other 1 213 566 0 404 522
DK 78 958 28 Other 105 824 25 35 275
EE 4 393 12 Other 41 549 89 13 850
ES 775 053 273 Other 1 063 502 27 354 501
FI 64 081 35 Other 64 081 0 21 360
FR 937 234 537 Other 1 498 490 37 499 497
GR 6 933 35 Other 101 277 93 33 759
HR 22 937 36 Other 106 300 78 35 433
HU 28 280 63 Other 198 128 86 66 043
IE 31 357 28 Other 149 023 79 49 674
IT 418 115 500 Other 1 309 077 68 436 359
LT 4 873 32 Other 102 523 95 34 174
LU 10 219 16 Other 39 431 74 13 144
LV 2 346 16 Other 68 646 97 22 882
MT 2 980 4 Other 12 080 75 4 027
NL 812 174 Other 438 552 100 146 184
PL 128 399 276 Other 795 577 84 265 192
PT 202 158 117 Other 408 475 51 136 158
RO 5 135 236 Other 576 838 99 192 279
SE 70 806 33 Other 137 758 49 45 919
SI 17 178 11 Other 45 103 62 15 034
SK 11 837 40 Other 109 138 89 36 379
UK 412 227 255 Other 882 525 53 294 175
6 Annex
31 Methodological study on under-reporting of occupational accidents in European Union
Figure 12: Results by age groups, in 2015–17 adjusted by age of the victim and best reporting
countries
Country non_fatal fatal AGE non_fatal_final rate (%) of
underreporting
AT 6 746 0 Other 26 531 75
AT 73 206 55 AGE18 243 044 70
AT 84 171 147 AGE35 300 901 72
AT 23 999 137 AGEGE 92 665 74
BE 2 146 0 Other 4 278 50
BE 77 227 57 AGE18 129 949 41
BE 104 199 91 AGE35 188 796 45
BE 21 860 39 AGEGE 42 780 49
BG 8 0 Other 689 99
BG 1 672 43 AGE18 121 584 99
BG 3 395 155 AGE35 265 830 99
BG 1 633 71 AGEGE 138 106 99
CY 19 0 Other 82 77
CY 2 179 2 AGE18 7 941 73
CY 2 403 7 AGE35 9 430 75
CY 959 2 AGEGE 4 065 76
CZ 816 7 Other 4 456 82
CZ 48 076 65 AGE18 221 640 78
CZ 63 456 164 AGE35 315 005 80
CZ 20 572 97 AGEGE 110 302 81
DE 31 859 2 Other 31 859 0
DE 922 665 232 AGE18 922 665 0
DE 1 150 797 592 AGE35 1 150 797 0
DE 480 728 467 AGEGE 480 728 0
DK 2 713 3 Other 4 192 35
DK 37 959 18 AGE18 49 516 23
DK 59 139 38 AGE35 83 066 29
DK 25 892 31 AGEGE 39 281 34
EE 16 0 Other 176 91
EE 3 934 8 AGE18 36 448 89
EE 3 955 22 AGE35 39 456 90
EE 2 198 21 AGEGE 23 684 91
ES 1 575 0 Other 2 519 37
ES 361 823 156 AGE18 488 474 26
ES 758 533 593 AGE35 1 102 665 31
ES 177 308 208 AGEGE 278 395 36
FI 382 0 Other 382 0
FI 39 335 18 AGE18 39 335 0
FI 58 457 41 AGE35 58 457 0
6 Annex
32 Methodological study on under-reporting of occupational accidents in European Union
Country non_fatal fatal AGE non_fatal_final rate (%) of
underreporting
FI 27 026 34 AGEGE 27 026 0
FR 14 460 17 Other 26 921 46
FR 537 690 177 AGE18 845 063 36
FR 676 978 630 AGE35 1 145 661 41
FR 170 987 368 AGEGE 312 542 45
GR 2 0 Other 34 94
GR 3 502 18 AGE18 50 044 93
GR 7 038 58 AGE35 108 295 94
GR 1 417 17 AGEGE 23 550 94
HR 64 8 Other 343 81
HR 13 751 14 AGE18 62 297 78
HR 20 486 56 AGE35 99 935 80
HR 6 271 22 AGEGE 33 041 81
HU 386 0 Other 3 132 88
HU 23 814 45 AGE18 163 125 85
HU 32 248 116 AGE35 237 856 86
HU 10 415 88 AGEGE 82 973 87
IE 1 047 0 Other 5 833 82
IE 19 118 27 AGE18 89 907 79
IE 26 284 44 AGE35 133 097 80
IE 6 444 64 AGEGE 35 246 82
IT 1 913 4 Other 7 013 73
IT 206 386 231 AGE18 638 706 68
IT 496 255 765 AGE35 1 653 673 70
IT 180 736 508 AGEGE 650 509 72
LT 69 1 Other 1 687 96
LT 3 772 20 AGE18 77 847 95
LT 4 514 62 AGE35 100 313 96
LT 2 450 39 AGEGE 58 806 96
LU 42 0 Other 194 78
LU 6 030 6 AGE18 23 521 74
LU 12 758 28 AGE35 53 584 76
LU 2 365 11 AGEGE 10 729 78
LV 14 0 Other 477 97
LV 2 016 15 AGE18 57 965 97
LV 2 148 40 AGE35 66 501 97
LV 1 236 30 AGEGE 41 331 97
MT 38 1 Other 181 79
MT 2 261 5 AGE18 9 097 75
MT 2 726 4 AGE35 11 810 77
MT 928 3 AGEGE 4 342 79
6 Annex
33 Methodological study on under-reporting of occupational accidents in European Union
Country non_fatal fatal AGE non_fatal_final rate (%) of
underreporting
NL 26 5 Other 15 841 100
NL 335 53 AGE18 172 303 100
NL 544 171 AGE35 301 280 100
NL 346 127 AGEGE 206 972 100
PL 257 1 Other 1 825 86
PL 87 721 169 AGE18 525 881 83
PL 118 550 399 AGE35 765 260 85
PL 43 777 248 AGEGE 305 223 86
PT 7 949 7 Other 18 675 57
PT 120 513 62 AGE18 239 014 50
PT 213 790 238 AGE35 456 565 53
PT 62 647 132 AGEGE 144 503 57
RO 13 0 Other 1 702 99
RO 3 890 145 AGE18 429 814 99
RO 6 854 429 AGE35 815 454 99
RO 1 835 184 AGEGE 235 806 99
SE 129 2 Other 288 55
SE 34 271 24 AGE18 64 666 47
SE 49 502 47 AGE35 100 577 51
SE 27 080 42 AGEGE 59 428 54
SI 39 0 Other 119 67
SI 12 955 11 AGE18 33 430 61
SI 20 724 29 AGE35 57 583 64
SI 4 180 13 AGEGE 12 545 67
SK 73 0 Other 780 91
SK 10 520 35 AGE18 94 898 89
SK 13 693 69 AGE35 133 004 90
SK 4 866 39 AGEGE 51 051 90
UK 10 505 12 Other 26 105 60
UK 220 323 186 AGE18 462 184 52
UK 325 251 336 AGE35 734 680 56
UK 133 751 258 AGEGE 326 317 59
6 Annex
34 Methodological study on under-reporting of occupational accidents in European Union
Figure 13: Results by body part of the injured in 2015–17 and best reporting countries
Country non_fatal fatal BODY non_fatal_final underreporting
AT 23 701 225 BODY1_4_7 123 639 81
AT 561 2 Other 3 476 84
AT 100 958 3 BODY5_6 314 292 68
BE 36 039 61 BODY1_4_7 89 330 60
BE 3 141 59 Other 9 248 66
BE 95 578 3 BODY5_6 141 379 32
BG 1 321 181 BODY1_4_7 144 983 99
BG 1 0 Other 130 99
BG 3 198 7 BODY5_6 209 458 98
CY 996 6 BODY1_4_7 5 442 82
CY 11 0 Other 71 85
CY 2 653 0 BODY5_6 8 651 69
CZ 15 790 159 BODY1_4_7 114 642 86
CZ 827 64 Other 7 132 88
CZ 71 021 4 BODY5_6 307 716 77
DE 333 602 846 BODY1_4_7 333 602 0
DE 23 988 12 Other 23 988 0
DE 1 365 477 22 BODY5_6 1 365 477 0
DK 37 466 28 BODY1_4_7 68 240 45
DK 3 316 26 Other 7 175 54
DK 58 341 2 BODY5_6 63 414 8
EE 1 494 21 BODY1_4_7 21 469 93
EE 9 4 Other 154 94
EE 5 219 0 BODY5_6 44 756 88
ES 283 051 660 BODY1_4_7 557 909 49
ES 3 874 0 Other 9 071 57
ES 580 262 1 BODY5_6 682 536 15
FI 21 714 49 BODY1_4_7 21 714 0
FI 1 039 7 Other 1 039 0
FI 61 341 2 BODY5_6 61 341 0
FR 234 148 211 BODY1_4_7 504 295 54
FR 33 911 381 Other 86 757 61
FR 382 386 3 BODY5_6 491 473 22
GR 1 614 60 BODY1_4_7 36 192 96
GR 0 0 Other 0
GR 6 358 0 BODY5_6 85 080 93
HR 7 120 22 BODY1_4_7 46 642 85
HR 1 576 45 Other 12 264 87
HR 18 613 0 BODY5_6 72 764 74
HU 8 664 134 BODY1_4_7 92 765 91
6 Annex
35 Methodological study on under-reporting of occupational accidents in European Union
Country non_fatal fatal BODY non_fatal_final underreporting
HU 314 26 Other 3 994 92
HU 35 023 6 BODY5_6 223 781 84
IE 15 721 80 BODY1_4_7 99 889 84
IE 1 684 7 Other 12 707 87
IE 21 401 3 BODY5_6 81 149 74
IT 156 074 738 BODY1_4_7 721 651 78
IT 17 044 260 Other 93 613 82
IT 416 205 29 BODY5_6 1 148 436 64
LT 1 592 78 BODY1_4_7 51 170 97
LT 25 0 Other 955 97
LT 5 647 0 BODY5_6 108 317 95
LU 4 027 22 BODY1_4_7 23 390 83
LU 63 1 Other 435 86
LU 9 953 0 BODY5_6 34 499 71
LV 1 007 33 BODY1_4_7 43 294 98
LV 30 13 Other 1 532 98
LV 2 567 1 BODY5_6 65 860 96
MT 1 110 5 BODY1_4_7 6 566 83
MT 120 1 Other 843 86
MT 2 905 0 BODY5_6 10 255 72
NL 228 82 BODY1_4_7 145 841 100
NL 241 10 Other 183 117 100
NL 388 141 BODY5_6 148 109 100
PL 35 858 541 BODY1_4_7 334 829 89
PL 17 16 Other 189 91
PL 130 393 17 BODY5_6 726 598 82
PT 80 566 294 BODY1_4_7 230 024 65
PT 11 735 3 Other 39 799 71
PT 177 566 4 BODY5_6 302 541 41
RO 2 602 480 BODY1_4_7 420 668 99
RO 96 22 Other 18 436 99
RO 5 706 20 BODY5_6 550 512 99
SE 22 871 34 BODY1_4_7 60 898 62
SE 4 749 43 Other 15 019 68
SE 45 503 1 BODY5_6 72 303 37
SI 5 385 39 BODY1_4_7 21 589 75
SI 54 0 Other 257 79
SI 20 297 0 BODY5_6 48 560 58
SK 3 731 77 BODY1_4_7 52 736 93
SK 141 20 Other 2 367 94
SK 15 466 1 BODY5_6 130 457 88
UK 162 661 426 BODY1_4_7 485 920 67
6 Annex
36 Methodological study on under-reporting of occupational accidents in European Union
Country non_fatal fatal BODY non_fatal_final underreporting
UK 10 392 103 Other 36 876 72
UK 289 613 11 BODY5_6 516 301 44
6 Annex
37 Methodological study on under-reporting of occupational accidents in European Union
Figure 14: Results by injury type as adjusted by the best reporting countries, total EU data not available due to missing reporting by Member States
Country non_fatal fatal INJ non_fatal_final underreporting
AT 4 903 114 INJ70_120 31 933 85
AT 113 684 38 INJ10_30 373 917 70
AT 4 769 52 INJ40_60 23 591 80
AT 3 539 1 Other 17 870 80
BE 4 860 22 INJ70_120 15 622 69
BE 119 296 7 INJ10_30 193 647 38
BE 11 393 33 INJ40_60 27 814 59
BE 6 020 61 Other 15 002 60
BG 63 89 INJ70_120 8 986 99
BG 3 697 26 INJ10_30 266 300 99
BG 658 59 INJ40_60 71 284 99
BG 0 0 Other 0
CY 134 5 INJ70_120 951 86
CY 3 470 1 INJ10_30 12 440 72
CY 335 1 INJ40_60 1 806 81
CY 29 0 Other 160 82
CZ 536 47 INJ70_120 4 655 88
CZ 70 659 26 INJ10_30 309 885 77
CZ 4 526 23 INJ40_60 29 853 85
CZ 14 570 105 Other 98 098 85
DE 20 638 59 INJ70_120 20 638 0
DE 1 032 496 197 INJ10_30 1 032 496 0
DE 651 601 558 INJ40_60 651 601 0
DE 36 773 29 Other 36 773 0
DK 4 644 7 INJ70_120 10 833 57
DK 52 359 2 INJ10_30 61 682 15
DK 5 293 11 INJ40_60 9 378 44
DK 13 126 42 Other 23 740 45
EE 99 26 INJ70_120 1 853 95
EE 6 457 2 INJ10_30 61 040 89
EE 408 4 INJ40_60 5 801 93
EE 16 2 Other 232 93
ES 16 896 414 INJ70_120 44 974 62
ES 776 413 17 INJ10_30 1 043 644 26
ES 70 435 182 INJ40_60 142 396 51
ES 21 745 0 Other 44 873 52
FI 982 19 INJ70_120 982 0
FI 67 131 0 INJ10_30 67 131 0
6 Annex
38 Methodological study on under-reporting of occupational accidents in European Union
FI 12 591 25 INJ40_60 12 591 0
FI 2 427 14 Other 2 427 0
FR 90 636 79 INJ70_120 222 710 59
FR 334 430 14 INJ10_30 414 990 19
FR 217 257 49 INJ40_60 405 464 46
FR 107 346 454 Other 204 498 48
GR 49 12 INJ70_120 1 406 97
GR 7 453 11 INJ10_30 108 006 93
GR 598 41 INJ40_60 13 034 95
GR 110 1 Other 2 447 96
HR 1 582 25 INJ70_120 13 083 88
HR 20 740 5 INJ10_30 86 618 76
HR 1 489 8 INJ40_60 9 353 84
HR 3 616 32 Other 23 185 84
HU 741 55 INJ70_120 10 235 93
HU 42 693 28 INJ10_30 297 810 86
HU 2 206 49 INJ40_60 23 144 90
HU 377 31 Other 4 037 91
IE 370 52 INJ70_120 3 366 89
IE 30 671 1 INJ10_30 141 054 78
IE 959 19 INJ40_60 6 632 86
IE 4 214 14 Other 29 751 86
IT 25 105 101 INJ70_120 154 342 84
IT 529 221 592 INJ10_30 1 643 066 68
IT 17 239 14 INJ40_60 80 496 79
IT 18 563 258 Other 88 478 79
LT 121 47 INJ70_120 5 042 98
LT 6 895 11 INJ10_30 145 103 95
LT 475 17 INJ40_60 15 034 97
LT 27 2 Other 872 97
LU 182 14 INJ70_120 1 420 87
LU 12 528 5 INJ10_30 49 347 75
LU 908 6 INJ40_60 5 379 83
LU 218 7 Other 1 318 83
LV 21 22 INJ70_120 1 174 98
LV 3 098 2 INJ10_30 87 465 96
LV 285 8 INJ40_60 12 102 98
LV 301 27 Other 13 046 98
MT 35 1 INJ70_120 287 88
MT 3 456 0 INJ10_30 14 288 76
MT 163 7 INJ40_60 1 014 84
MT 10 0 Other 63 84
NL 241 10 INJ70_120 190 949 100
6 Annex
39 Methodological study on under-reporting of occupational accidents in European Union
NL 428 169 INJ10_30 171 252 100
NL 54 16 INJ40_60 32 496 100
NL 135 51 Other 82 927 100
PL 4 310 286 INJ70_120 50 942 92
PL 149 200 29 INJ10_30 890 552 83
PL 14 910 196 INJ40_60 133 849 89
PL 5 2 Other 46 89
PT 5 346 129 INJ70_120 20 339 74
PT 221 476 17 INJ10_30 425 530 48
PT 16 913 128 INJ40_60 48 873 65
PT 26 786 4 Other 79 010 66
RO 398 168 INJ70_120 79 461 99
RO 6 187 74 INJ10_30 623 799 99
RO 1 711 201 INJ40_60 259 455 99
RO 383 34 Other 59 283 99
SE 4 481 41 INJ70_120 15 164 70
SE 52 152 1 INJ10_30 89 124 41
SE 4 162 14 INJ40_60 10 696 61
SE 13 824 25 Other 36 269 62
SI 191 17 INJ70_120 795 76
SI 11 267 4 INJ10_30 23 696 52
SI 673 3 INJ40_60 2 129 68
SI 13 319 6 Other 43 003 69
SK 493 45 INJ70_120 8 762 94
SK 17 676 9 INJ10_30 158 651 89
SK 1 189 23 INJ40_60 16 050 93
SK 547 11 Other 7 537 93
UK 3 844 182 INJ70_120 15 554 75
UK 370 923 25 INJ10_30 757 999 51
UK 22 657 199 INJ40_60 69 635 67
UK 55 399 126 Other 173 803 68
6 Annex
40 Methodological study on under-reporting of occupational accidents in European Union
Commentary to the 4 annexed tables
Some clarification on symbols on the NACE groups used at the EU:
A Agriculture
C Manufacturing
F Construction
H Transport and Storage
Other Other economic sectorsBackground = theme colour 20%
The imputation based on the NACE Groups and age groups of the injured match perfectly well
overall. Summary data of injury type and injured body are provided for each country while total
numbers cannot be estimated due to some non-reported data.
As a whole the imputation method adjusting non-fatal accident numbers provides a reliable method
of estimating the level of under-reporting. The rate of under-reporting varies by country and in
average is 58.1 %. Instead of 2.9 million annual non-fatal accidents that are reported to EUROSTAT
we estimated that a realistic number covering also the non-reported cases was 6.9 million accidents
in average over the period of 2015–15.
See also comparison to earlier estimates made by the ILO, in 2017, based on latest data then
available of the reference year 2014 that indicated 3.55 million non-fatal accidents in EU-28 in 2014
with a range up to 4.0 million accidents. The method was a simplified one based on less reliable data
from other countries globally.
Based on global data other industrialised countries tend to follow the same tendencies in reporting –
and under-reporting – fatal and non-fatal accidents.
The variable non_fatal_final is built by the total estimate of the years 2015–17.
The variable average 2015–17 consists of the rolling average over the three years.
The variable age as classified
Code Label
00 Less than 1 year
01 1 year old
02 2 years
… etc
10 10 years
… etc
90 90 years
98 Above 90 years of age
99 Age unknown
is grouped as follows: group 1 — 18–34 , group 2 — 35–54, group 3 — 55, group 4 — other
The variable body ‘part of body injured’ as classified in Annex I is grouped into group body 5_6
(Upper Extremities, not further specified, Lower Extremities, not further specified), body 1_4_7
6 Annex
41 Methodological study on under-reporting of occupational accidents in European Union
(Head, not further specified , Torso and organs, not further specified, Whole body and multiple sites,
not further specified) and other including the remaining categories.
The variable injuries ‘type of injuries’ is grouped into:
INJ10_30 (010 Wounds and superficial injuries, 020 Bone fractures and 030 Dislocations, sprains
and strains)
INJ40_60 (040 Traumatic amputations (Loss of body parts), 050 Concussion and internal injuries,
060 Burns, scalds and frostbites)
INJ70_120 (070 Poisonings and infections, 080 Drowning and asphyxiation, 090 Effects of sound,
vibration and pressure, 100 Effects of temperature extremes, light and radiation, 110 Shock, 120
Multiple injuries)
Other (999 Other specified injuries not included under other headings)